I love OpenRouter, long time user. Stripe will hopefully be a good custodian.
I just want to point out some features of OpenRouter that make it more than just a model selection and routing endpoint and that I find incredibly useful:
0/ Default routing is to the cheapest provider, but they're usually not the most performant. I'd guess 99% of OpenRouter integrations never tweak the default routing. Here you can setup cheapest with performance minimums:
1/ Using broadcast you can push all your analytics to clickhouse / s3 / snowflake and a bunch of other compatible destinations. Setup a clickhouse server ($5 VPS[0]) and send all your traces to it:
there is also PII detection. This will show up in observability as rejections/blocks etc.
There are so many model routing solutions (same with observability, security etc.) but they're all 80% solutions - OpenRouter really rounds out with well implemented features that you need when deploying models at any scale and I gladly pay the toll.
[0] not sure if these exist any more but clickhouse is resource efficient
> Model router is also a natural home for llm security - OpenRouter has the beginnings of prompt injection detection
Maybe I'm biased from the perspective of a "harness provider", but I think Model routers often have too little context to act as well-informed prompt injection prevention. e.g. it lacks context of where which part of the message(s) originates from and sanitization/safeguards were already performed on the application layer.
Something like OpenRouter's "flag" mode is fine, but usage of auto-redact or auto-block should really only be used if there is significant risk exposure through your harness or otherwise they are a constant source of bugs.
They might be popular for indie developers, but nobody doing serious AI or in a corporate environment is using them.
And if they are, their compliance team is about to strike them down. The VCs forcing this acquisition do know this.
- Why would you add a penalty of 50 ms at a minimum? And that is not the p95...
Just run LiteLLM in house and you dont even really need that.
- Their capacity pools are shared across the whole user base, a massive batch processing by another of their customers and think what that means for your response time...
- So instead of negotiating corporate rates with OpenAI or Anthropic, you would be using an intermediary and topping up the corporate credit card... for a 5% markdown ? Really?
- They can see all your critical corporate data on the in and out
- They present some pink SOC 2 promises but then wash their hands and defer to
you and the providers. Its just the Bolt and Uber model the drivers are not
our employees....
- They are a man in the middle proxy that is a massive security liability
for your corporation
- They have no support for private cloud points
- No geofencing guarantees
- No intellectual property legal indemnification unlike what AWS or Microsoft or Google offers
- Its a provider roulette inconsistent with hosts providing different quantization levels causing random shifts in response quality
- Support via a Discord server...
The only reason they were not shutdown yet by Anthropic or OpenAI is because they have the same VCs, as those two. That would mean said VCs investment would go to zero. Oh and those are the same VCs that own Stripe...
Just setup a private proxy tier using something like LiteLLM, even if you really dont need it. Just code your enterprise apps to have have fallback loops on the core hyperscaler providers like AWS Bedrock or Azure Foundry...
The other thing OpenRouter gets you is: Access to loads of models via a common API/sign-up/prepayment mechanism.
I want to know which vendor/model does best at my extracting-facts-from-text task? Which does best at my OCR-a-text-document task? Which can deal with a safe-for-work beach photo without a censorship system false alarm?
OpenRouter lets me run my tests against openai and anthropic and google and x and bytedance and qwen and llama, with a single sign-up and a single payment.
Nobody would trust one of the labs that have their own models with this. They’d assume they’re taking all the data, and biasing the routing to favour themselves. xAI would probably be the least trusted, but the same would apply for any of them.
I think it is fitting because they provide „all“ models independent from the provider. Open in this sense means to me being open to switch the model, speed and provider at any time.
I used to demand the same but a word can have multiple meanings depending on the context, eg OpenMind means something specific. Open{$IT-term} does not necessarily mean Open Source {$IT-term} and OpenAI might mean an AI available to all people, not implying anything about cost. What could be illegal is the word Free, which a for-profit company would never use in the first place.
Turns out even a proxy can be worth $8bn with the right business model behind it.
Users get an array of providers competing behind a single API, meaning they have to compete on price and quality not vendor lock-in. This encourages users to join OpenRouter over specific model vendors.
Providers get easy access to revenue (and data) and new customers with little to no ad spending, encouraging them onto the platform too.
Someone noted they pass through 100 billion annually if I remember correctly, but they also take a 5.5% cut, so Stripe secures them as a “customer” while getting their money back over time. It also to me feels like an adjacent business aligned with sort of what Stripe already does to begin with, being someone elses middle man.
Who is that someone? And unnamed employee that told somebody else?
They make less than 140 million in revenue, what today is like 5 signing bonus for OpenAI :-) They are not profitable and see through less than 2 billion in revenue.
I've been using OpenRouter since relatively early (I build evvl.ai - an eval platform on top of it) and think that they're selling at a good time.
An underpinning of their model is that API calls / inference are similar across providers allowing for commodity tokenization cost comparisons, but the providers are beginning to shift to non commodity features that don't easily shift.
For instance, calling Gemini with "search grounding" isn't something that OpenRouter can do (they sub their own web search in), last I checked they weren't doing real time voice models, etc.
Your main point is still right, the providers are beginning to shift to non-commodity features.
Another example might be all the managed agent features, where you get a VM + model, right now it's mostly Google and Anthropic that have this offering.
The service tier is another, where Google, Vertex, OAI, Anth offer it but only OAI and Google offer flex service tier and not all models from them get it.
OpenRouter for me is an abstraction over that complexity, they will have their work cut out for them.
I can see Stripe adding extra services on top. However, in their main business its "easy" for them because getting into payment processing is hard intially. But whole search, harness, VMs, agent loops are not hard to find somewhere else.
I think it has to be more than just a 'proxy' what Stripe bought.
Stripe and every other big company could build this without any issues.
It either is just a really stupid business decision or it is about the name. The only model proxy i know is OpenRouter despite plenty of other model proxies existing.
Awesome will check this out. EUrouter is too cost prohibitive for me, and now that open router has been acquired by stripe it’s only a matter of time until it’s unusable
OpenRouter is great business I agree, but I'm not still convinced how the integrity of providers' models is ensured. In other words, can't the provider serve DSv4 flash advertising it as DSv4 Pro?
I'm aware that OpenRouter checks response quality onboarding, and does further checks occasionally, but I'm concerned that it's basically a cat-and-a-mouse problem between the scammers and the detectors. For example, there could be a signal that a specific pattern of requests are from OpenRouter's quality testing bots. Or, they can just route 1% of requests to an inferior model and benefit a small gain, hoping it fits into the statistically allowed margin.
How is that a problem that's unique to openrouter? Who's to say when I query the Anthropic API directly, wanting to use Opus, but they determine it doesn' need Opus, route it to Sonnet instead and pocket the difference? The only difference with openrouter is that it's another layer where this type of fraud could occur.
Yes, it is awfully inconsistent today. They run some tests (accuracy table buried on model page) but they are sparse and only capture a single moment in time. I would love to see OpenRouter take this more seriously.
AI products are going to have to deal with accounting. An agent performs some work. It uses some models and perhaps some other metered services. Someone has to meter that activity. They have to attribute costs, apply the product's pricing rules, bill the customer, collect, reconcile with vendors, and maintain the ledger.
Stripe can use OpenRouter to build the financial and accounting infrastructure for every product that sells metered AI work.
I think the analogy is ADP. Payroll for all the work that's going to be done by AI agents.
A single policy location for multiple models with usage limits/rate limits, spend limits all configurable using a common syntax + deep integration with account levels and high visibility into subscription/payments related to meter usage to help understand whether or not your pricing model is underwater as costs change is exceptionally valuable. AFAIK each product consuming model providers without an openrouter style proxy has to do that work themselves, and if you are already using stripe for subs/payments, the direct integration is a huge value add for customers, especially if you can alert on "upcoming price changes from your provider changes your unit economics for XYZ packages/sub tiers".
Similar to how Stripe is a middleman for payments across (fragmented) banks, they want to be a middleman for (fragmented) AI models as tokens are the new currency.
As AI agents/harness/human are spenders of tokens, enabling them to derisk from being locked to a specific model provider & allowing to (re)route to any model at anytime for better leverage in a single API, similar to how they are doing for payments.
paying for data access, for "deterministic computing", for letting the AI use online services. Maybe even micro payments for using tested and verified skills/mcps.
Example: access to real-time stock trading data, access to weather information, letting it make stock trades, etc.
At first I was confused why Stripe bought OpenRouter, but I think this makes sense.
Not OP, but I could see the vision. Right now OpenRouter is developer-paid with no way to pass costs onto the end user. Stripe could help bill and meter at the user level for per-user billing. I would also think more SDKs/easier ways to add AI features (paid for via openrouter) into apps and websites.
Open router supports per-request usage metering, as well as per-key. Building user-level metering from this is trivial, and there are many implementations of this out there. Supporting per-user first-class is trivial considering Open router already has per-request and already supports per-request attribution. How important that is, is a big question, but definitely not something that requires Stripe to build.
Could there be use cases where you want to run inference for users in your app, but don't want to host your own backend, user authentication, and billing?
The user could log into his personal OpenRouter account, authorize your application, and optionally set a budget, all in something like a Stripe payment screen.
> The user could log into his personal OpenRouter account, authorize your application, and optionally set a budget, all in something like a Stripe payment screen.
Good point, just needs an user friendly flow to sign up and set the budget in a way that is more similar to Stripe checkout than regular account creation.
Perhaps stripe thinks metering is a lovely big market? A nice complement to their existing service. Plus a little AI buzz likely helps their valuation. OpenRouter could be a step rather than a goal.
As user agents become more common it's only natural that they will be used for taking the heavy lifting out of e-commerce purchases. There will be a big need for digital payments to verify and reconcile these purchases.
With LLMs that are plugged into digital payments we will essentially have buying agents in our pocket that can find us exactly what we want for the cheapest price and the quickest delivery.
I'm always surprised at how many people in the tech sector don't think that things can and will change. In the next couple of years? Sure. 5+? Who knows. Necessity is the mother of invention.
I'm always surprised at how many people in the tech sector assume overcoming fundamental LLM weaknesses is always “just a matter of time,” and that time will arrive quickly enough to be relevant. If you haven’t seen any of the videos of people trying to get frontier LLMs to count consistently, you might think we’re further from LLM accountant than you realize. It’s not a data problem— I’m pretty sure they’ve encountered numeric progressions between 1 ABs l and 100 before.
I'm not just talking about LLMs though. People thought the internet and email were just a fad that would never catch on. People thought that the idea of home computing was absurd. All because they only saw what was in front of them at the time.
Also, how many accountants do you know that are truly happy with their work? I'd like to think many of them would love to do something other than crunch numbers all day.
> People thought the internet and email were just a fad that would never catch on. People thought that the idea of home computing was absurd.
We're far removed from "we've sold all the computers the world can buy" (back from when computers were the size of a building).
This is a disingenuous argument. The web and email were basically instant hits and people realized it. Similar story for home computing once the computing power caught up.
> I'm always surprised at how many people in the tech sector assume overcoming fundamental LLM weaknesses is always “just a matter of time,” and that time will arrive quickly enough to be relevant.
Lots of techies are tech-optimists ("tech always improves quickly").
Lots of people also have dollar signs in their eyes (or related, such as increased visibility and scope).
Things can easily change for the worse, such as a stock market crash. Who knows? Will we get AGI or will we see more homeless? History says... AGI for sure!
It's very difficult for this sort of thing to change for the worse. It's basically impossible on the software side. If next year's model is worse then keep using this year's model. Hardware technology can potentially go backwards, but things have to get really bad for that to happen.
LLMs are inherently unreliable. We do already work with unreliable technologies for many things but there's this engineering principle called "use the right tool for the job".
And it's a lot more likely that LLMs can't be changed to become unreliable, it's just how they work. So we would need more basic research, that doesn't grow on trees and for which the timelines are basically open ended. Maybe tomorrow, maybe right after cold fusion hits mass adoption.
I'm not happy about this. Having used OpenRouter a lot and enjoying the experience, I need to say that corporate consolidation is never good for consumers.
YouTube there’s no real alternative because hosting unlimited amounts of video for free is a money pit that’s impossible to turn into a viable business.
If some big corporation wasn’t willing to subsidise it for some ulterior motive, it simply wouldn’t exist
curious as a non-user, but what does openrouter have to do with stripe? Ignoring the corpo-speak about mission statement.. the missions don't really seem aligned in reality.
I don't think they have much to do with stripe. I'm guessing they want to round out their business more and include payments for AI since OpenRouter is basically just an AI traffic router that takes your money and provides it to the APIs you use.
My guess: your LLM agent makes buying decisions for you. Providers of agents can integrate witg stripe. Those using openrouter click a button for micropayment support.
That's a rule but then Stripe won the market by having very friendly interface and minimizing the pain of dealing with paperwork for merchants. When PayPal (the worst offender, just hostile company) acquired BrainTree it was clearly a terrible news. When Stripe acquires OpenRouter I wouldn't be so sure.
Happy for the OpenRouter team. Been using the platform since early stages and the ability to run any AI model with the same api key and same request has been great to experiment with new models and being able to switch models in prod with minimal effort.
They also support fallback by default so you don’t have to write wrappers and logic to choose models, it just works with their SDk using config.
It's a mostly equity offer. Private capital is always more risky than publicly traded shares so probably less than $7bn risk adjusted, which makes sense for Stripe.
I manage our company's model spend and build tools around it.
The good thing about the developer experience is that it is dead simple to create keys that are time-bound, model restricted, and have budgets against them in an easy, programmatic way (also in bulk). Costs are guaranteed to be on par or lower than any model provider, so it's an easy sell to the finance department.
To look at it from the other side, Gemini is the absolute worse when it comes to key management. The labyrinth of actions that one has to navigate to generate budgeted keys for a team is not worth it and I still haven't figured out how to manage the keys at a granularity I need.
OpenAI and Anthropic both have what I would consider to be middling DevEx. Most things work, but only on their platform, and there are some weirdnesses, like the fact that anthropic disabled its admin API key creation, so you can't programmatically create keys against a budget.
> Gemini is the absolute worse when it comes to key management.
I saw another comment thread on this recently. What I gathered was they hid it in the Cloud Platform, so for those who jumped the cloud hoops already, maybe it's alright. (However when I read what Openrouter offers, as you just laid out, I feel Google can't hold a candle to Openrouter DevEx.)
However, if there's anyone who is aware, was there a brief point where getting a Gemini API key was simple (enough)? I'm asking because, in true Google fashion, the AI product went through a lot of changes.
Not OP, but you can't set a per api key limit in OpenAI Platform, you also can't really easily see spending going on per key, and its also in a far off page separate from the keys, really not easy to see what is going on with spend in real time.
You're supposedly worth a trillion dollars, can't you know... pay for your own market research rather than forcing the public to provide it for you? Is this what they seriously teach devs there?
I am working on my own agent implementation (https://usehax.dev/), and I can praise OpenRouter for being seemingly the only provider endpoint that actually gives proper API for gathering available models and their capabilities (context length, image support, pricing, etc), reports exact routing details and total cost in the response, and provides an endpoint for obtaining current available credits (for "/usage"-like command). So for OpenRouter specifically I could deliver a really good experience out-of-the-box (always fresh models and metadata, exact pricing, etc).
You would think that these things would be standard fare, but they really are not. Other providers need metadata from a separate catalog, don't report pricing, don't really report available credits / usage either.
Speaking for myself (I'm not the GP comment): The platform 'just works', has great reporting interface to help understand excatly what usage happened when, and integrates well with lots of different harnesses.
Hint: It's not 15%. And there are many reasons, such as not having to keep up-to-date billing details in 70 providers, and not wasting money because most providers want you to prepay a balance that gets stuck in there if you switch to another provider.
It also has way better uptime than the underlying platforms, even for proprietary models like Claude. When Claude APIs are having issues, OpenRouter Claude still keeps working because they can route to AWS Bedrock instead of Anthropic etc. This effect is even bigger with open-weight models because they typically have 5-10 providers.
OR I can keep paying 15% markup till next month and then jump to newest / cheapest model with one-liner change instead of being locked to a model/provider.
That requires that you create accounts and separately manage each provider you want to use. OpenRouter removes that - if a new provider gets added, you can start using them immediately.
They get paid the same either way, Openrouter takes their cut and passes it onto the consumer. And anyone with an Openrouter account has already escaped from their proprietary moat. So for those who have expressed a clear preference for provider-agnostic model infrastructure, do you want to just wall yourself away from them forever?
And per the rankings [0] 18.2% of the tokens sold through went to OpenAI [1]. 3.8% for Anthropic though. Google's up there at 23.4% primarily because GCP/Vertex/whatever they're offering is a billing nightmare and the only way to set a hard cap and sure you won't wake up the next day with a $10k bill is via OR.
[1] Although gpt-oss 20b and 120b are indexed as 'openai' because they're the creator even though they don't serve gpt-oss via openrouter. But neither are in the top 20 of models this month, while Luna and Sol are.
Read my sentence again. With your human eyes and brain:
> **Google's** up there at 23.4% primarily because GCP/Vertex/whatever they're offering is a billing nightmare and the only way to set a hard cap and sure you won't wake up the next day with a $10k bill is via OR.
OpenAI absolutely lets you set a hard or soft cap. I've set them and they've worked, I think one time I saw a $-0.01 balance. Their docs were probably written by ChatGPT. Later in that article it says:
> A spend limit can monitor spend without enforcement, or you can enforce it as a hard limit so API requests fail after spend reaches the limit.
And they still allow the most certain limit, which Google does not: have $0 on your deposited balance and don't have auto-reload enabled, you get 4XXed instead of going into debt (or at least by at most a couple cents).
Although for fun, go ask Gemini how you can set a cap on GCP spend. It'll try to write you a pub-sub client.
They already sell API access to the model on Google/Microsoft/Amazon cloud. OpenRouter is the same deal. Note that the labs are not giving out model weights to OpenRouter. They are just selling tokens.
It's a land grab. Right now none of them have a dominant position. Much better to get mind share and market share now. They can always squeeze the value chain later. And whoever wins surely will.
Quite frankly, the proprietary company's products are no longer good enough for them to be able to enforce a moat. The scene is popping and nobody is satisfied staying with one provider. If they stopped selling via Openrouter, the 18.2% of Openrouter's revenue that currently goes to OpenAI or the 23.4% of their revenue that goes to Google would probably dry up. Some people would switch back, but those who are happy to juggle multiple budgets and endpoints and API key management infrastructure are already doing so to avoid Openrouter's 5-10% cut. This is really evident at any $bigco where you can wait for weeks for a purchase order, needing leadership's approval and accounting's "I'll get to it when I get to it."
> OpenAI/Anthropic/etc don’t establish the direct customer relationship.
But they also don't have to pay to support and maintain the direct customer relationship and all the billing, support, liability, etc. The credit card merchant fee (a not insubstantial % at scale) is paid by Openrouter (or Stripe now) and the bill Openrouter pays to OpenAI is transferred probably via wire or something with a lot less fees. Meanwhile if you're paying OpenAI via credit card directly, OpenAI is paying Stripe or whoever.
> And OpenRouter can steer traffic away from those labs over to their own model.
The inverse is also true. If you've deposited money on a Openrouter account, you've taken a very big step across their proprietary moat. Would the proprietary companies rather lose them as a customer forever, or be able to entice them back, which can be done with a single change of a JSON parameter?
> The major labs don’t get personalized data about the user to train on, since it’s aggregated.
(Edit: Looked it up, Openrouter does send a pseudonymous but stable hashed userid upstream, stable across different API keys in the same account [0]. That makes de-identification a lot easier.)
First, a lot of the corporate use is getting ZDRed, so there's nothing to track without going back on zero data retention.
De-identification was trivial for normal web traffic before LLMs. Think of all the crap people's harnesses and claws send without de-identifying. Many coding agents can't help themselves from scanning your home directory.
Sure, the type of person to comment on HN probably has good hygene, but the typical vibe coder doesn't even know what they're leaking.
When the market settles down, most of the money is going to be in bulk inference, for batch jobs. In 5 years everyone's laptop will be able to run Qwen 3.8 27B for coding tasks, but businesses will still need to run inference 24/7. Very few people will still need SOTA models once you can run an Opus 4.6 class model on your laptop.
It’s a marketplace with a markup on every token they sell. And I’d rather go to this shop, than sign up individually at the 70+ different providers they broker access to - even if it comes with a price.
Frontier model providers (Anthropic et al.) gate new accounts to impractically low rate limits and spending caps until clients unlock higher limits with cumulative usage, make equivalent cash deposits up front, or talk to a Sales department to work out some other arrangement. They are handicapped by the postpaid billing model.
It's an administrative burden to sign up very every model provider, and there are many independent inference providers now that serve only open source models.
OpenRouter provides a useful service by allowing easy prepaid model access with much higher rate limits, and they also aggregate different model providers to route queries by price, latency, etc.
I don’t understand it, but good for OpenRouter. Personally, I think OpenRouter’s value is somewhat perishable. In the early AI market where there are a jillion models and anew one every week, it’s great to be able to try them all without having to create accounts and keys for each one. OpenRouter, which I use, makes this easy for developers to do. Add some cost controls and other “management” knobs and it works great. But the market is not going to be this frothy forever. As things settle down and commoditize, the value of switching on a dime diminishes as people lock into their favorite models. And with the OpenAI and Anthropic APIs being defacto standards for how to talk to models, it’s easy enough to switch to another model every so often, as long as you aren’t doing it multiple times per week. The other alternative is that OpenRouter stays in the mix but its pricing gets ground down and down. If it’s (nearly) free and still adds some value, fine. But that doesn’t justify $7B. So, yea, I don’t get it from Stripe’s perspective.
> As things settle down and commoditize, the value of switching on a dime diminishes as people lock into their favorite models
I can imagine just the opposite outcome from the same scenario: as people settle into their favorite but commoditized models, competition for marginal inference cost will take over. A company like OpenRouter that promises the cheapest tokens by the minute becomes essential on the low-cost margin.
I think that OpenRouter and equivalents get pushed out of the market only if the froth calms down (as you posit) and winning models stay proprietary, perhaps with their own unique API surfaces.
> A company like OpenRouter that promises the cheapest tokens by the minute becomes essential on the low-cost margin.
Okay, I can see that, but if their value is just cost optimization, their ability to mark up the tokens becomes increasingly difficult as well. Or, people will build a router themselves to avoid paying the markup, possibly with reduced features, but someone will open source it. Heck, Claude or GPT can probably one-shot it these days. Either way, I think the whole OpenRouter model is going to struggle unless the market stays frothy.
First, cost optimization directly contributes to how much markup they can charge. Say on average they deliver savings of 20%, that is excess value OR/Stripe can mostly capture.
Also, you don't need to mark up tokens much if you're a commodity volume business. Think of Costco and their margins & membership fees. Not everything has to be high margin, not everything has to be a SaaS subscription.
Read Ben Thompson to understand aggregation theory. Many of the largest internet companies, like Google and Amazon, are simply aggregators. OpenRouter is an aggregator of AI tooling. Stripe itself was just a convenience layer on top of merchant gateways (another aggregator). OpenRouter is so big they can negotiate special contracts with OpenAI for special rates.
Just because some Higher Ups in Stripe bought this, doesn't mean its that valuable.
It can easily be, that Stripe can just afford it and think that they are part of the big boys and thats just what companies cost today.
It could be that Stripe really really like the brand name and all the positive notion of it.
It could be that Stripe doesn't want to build this from scratch in a timefrime of 1 year or 2 because Stripe might be too corporate to be able to do startup stuff.
Stripe has quite a high motivation to leverage agents thoguh because they are preparing for Agents which will buy through stripe. They already provide the SKU backend and support the agent payment stuff.
It's a simple marketplace and they control a ton of the demand side who trust them to find the best price and providers that work. The value is huge to users - one place, multiple models, providers bid for the traffic and the user doesn't have to think about anything. The value they hold hostage against providers is gargantuan - nice model there it'd be a real shame if none of our users used it.
I don't see how stripe adds any value here (and I've had such terrible experience with stripe automatically breaking my stuff I am worried I can't trust openrouter now) but I can see stripe wanting to be in the middle of any two people giving each other money on the internet and this is squarely (lol) that.
Ha that’s a really interesting comparison! I guess the difference is that OR directly sends revenue to model providers, so maybe they’re more likely to continue working together? But I can totally see it go the other way once a provider feels confident enough their users won’t switch away. At that point the OpenRouter tax, however small, will be a problem to solve.
You can save money and get higher uptime. Without it using a cheap provider for open models would be risky because they might go down a lot, but routers can detect that and instantly switch to a different provider. They basically take away all your exposure.
The other thing is convenience and centralized security from using one gateway to access everything. It's a lot better than having to deal with N accounts with separate limits and monitoring. And giving your payment details to one company instead of 20 is obviously safer.
Investors like them because the pricing is inherently usage based so there's zero risk of clients using more tokens than what they paid for. Guaranteed profit as long as they can keep a modest amount of customers.
On acquisition pricing: Why smarter AI models could drive up compute prices 10x (at least temporarily). [0]
This seems mind blowing, but the big boys seem to be behaving as if it's directionally true.
If compute is constrained and expensive, OpenRouter is what you'll use to get around the constraints at individual providers.
OpenRouter allows you to have 1 API key to access
hundreds of models across dozens of providers.
Can’t explain the valuation since everything in this space is rationally overvalued, but I don’t think OpenRouters valuation is that surprising, all things considered.
So much so that right now GPT-5.6 Sol tokens are half price from OpenRouter compared to OpenAI directly. Which I guess means enough people are using OpenRouter, that OpenAI are more concerned about getting those users to switch (from, presumably, Anthropic) than they are about encouraging OpenRouter to exist and long term reduce margins for them by competition / lack of lock-in.
If you stick "AI" anywhere in your product pitch you get to stick a few extra zeroes on your valuation that has been pulled from your ass which you then sell to feckless VCs looking to swallow the world before the bubble pops
PE is desperately competing other PE to get into the promise of some AI thingy NOW?
Just guessing. The frenzy around popular, good already, and successful services with the corporate crap flowing from this acquisition announcement too is appalling. The "what's best for you, the user" heavy emphasis when this would be inherently evident in any honest service forecasts the opposite.
Some highlights from one of my agents asked for a no bullshit evaluation:
"By buying OpenRouter, they own the routing layer that decides which model gets called and at what price."
"Stripe wants to be the economic infrastructure for AI — meaning they want to be the toll booth and the traffic cop for the entire AI economy."
"... insider market intelligence that OpenAI, Anthropic, and Google don't fully have. Stripe would now have it."
"The $7B+ price is absurd on any financial metric, but this isn't a financial acquisition — it's a strategic positioning play to own the platform layer of the next computing paradigm."
"140x revenue multiple: At ~$50M revenue, this is not a financial valuation. It's a strategic land grab — buying the chokepoint before someone else does."
I do not feel a particularly strong smell of 'best for the user' here for some reason... More like the usual 'how do we squeeze out more for our PE folks from this' kind of scent.
Tokens are just a type of currency. Stripe is a middleman for currency.
Using something like OpenRouter (or any of the AI Gateways) is better than tying yourself to one LLM provider that can rug pull on pricing or change models in a way you don't like.
The value is in the network effect I think. OpenRouter is popular and has a good head start over anything Stripe could build internally.
to a certain extent, openrouter did manage to turn tokens into a currency. The fungibility is there, as much as you don't care so much if you're getting your token from novita or some other player, these become fungible.
They aren’t trading tokens, they are a market for compute. The difference matters quite a lot. What you have in openrouter is the ability to exchange money for compute at the vendor of your choice.
We are basically back to 2020 trying to get people to understand that an NFT is NOT the underlying asset it abstractly represents…
You spend $100 to get tokens from Claude. Then go to your friend and tell them you are willing to sell the tokens you got for $50, it’s a bargain! They look at you, tell you that makes absolutely no sense, and why the fuck would they want to buy your tokens, ie something that only makes sense in your very specific prompting context.
Tokens don’t have an inherent value and have no other properties that would make they tradeable.
Another way to think about it: a company that spends $1M on tokens is burning that money hoping they will be able to recoup by generating a better product. If it would be tradeable, they could spend $1M to generate tokens, produce absolutely no product, and just resell all those tokens to get their money back. That’s obviously nonsense, that’s not at all what tokens are. And they cannot transform the tokens they got into compute, you need to consume hardware and energy to mint tokens, you cannot convert tokens into hardware
Tokens are minted then consumed. You cannot trade them, by definition. What would it even mean to trade a token? You can maybe trade a voucher that allows you to then redeem it for some token generation (aka compute), but the tokens themselves don’t make sense to be tradeable. And because they aren’t fungible you cannot swap a token for another one, that’s just not what LLMs are
I think you are reading too much into the "fungible" term.
No one is talking about trading tokens.
The idea being that getting your tokens from Provider A is no different than Provider B, especially if they both offer the same model. You can change one value in a request to openrouter and suddenly be hitting a different provider but offering the same tokens, because they offer the same model and the same settings.
> The idea being that getting your tokens from Provider A is no different than Provider B, especially if they both offer the same model.
You're describing a commodity. However, the commodity isn't the tokens, but the compute capacity, i.e., serving a model. And compute capacity isn't a currency — at least not, until you can acquire compute capacity from one party and exchange it with another party.
dgellow is absolutely right: tokens aren't a currency, nor can you trade them, nor are they fungible. The original claim that "tokens are just a type of currency" [1] makes no sense.
Yeah, I mean, it looks like a network effect, but anybody can aggregate the models; rather, it is the switching costs in the logs and work, and the cost savings, etc.
Surely stripe if anyone have learned to harness the cash flowing through their system. Hell, they could be emitting bonds on expected token consumption bills!
It's also very representative of the HN comment section vibes overall.
But such is normal with normies providing startup critiques, if you just look at every announcement and say "this will fail", you'd be right 99% of the time.
I mean, the AI skepticism is still so prevalent today in HN, despite it solving decades old math problems, hacking into companies, making software engineer no longer code, etc. (all this just in the last year).
So nothing really has changed in this community's vibes.
If you think HN is anti-AI, you are in deep. There isn't many places on the internet (sans X/Twit which i don't check) that is as AI-positive as HN. Sure, holdouts remain but the frontpage has a 60% constant vibecoded project submission.
You are mixing supporting AI (which HN is mostly against, but more supportive than other places), and thinking AI is dumb, which HN has been pretty persistent on for years.
I've not used Open Router but I'm wondering what are the reasons to route LLM queries at large scale?
Isn't the trend increasingly to let an orshestrator agent decide which model to use in terms of problem difficulty vs. size and thinking budget?
Once you fix those parameters there's little to optimize over, no? I mean you want to find the cheapest one across providers but that's not a recurring decision.
I suspect only a very small subset users are sophisticated/interested enough to do that for themselves on a regular basis. More will likely hop between agents based on cost and the perceived nature of the task, without hard data (or an orchestrator model) to support that decision.
In the second case OpenRouter is very useful; in the first case, maybe it’s an area they move into in time - providing and owning/optimising the orchestrator?
Convenience. Paying only one provider, but get access to different models hosted around the world. Need a model for a project where latency matters? EU hosted and ZDR for another customer? Cheapest for personal use?
One thing that always worries me with these acquisitions: how are they gonna handle the impending 5.6M+ Stripe Connect users who rely on OpenRouter's APIs? Will they integrate it into their core infrastructure or leave it as a separate product?
I rather have protocols be built and less middlemen PaaS.
One great example is Open Banking. I would support an Open Router that mirrors Open Banking, unfortunately the Open here doesn’t carry the same connotation.
Kudos to the team for succeeding in executing around a clear DX issue during the chaos of providers, though long term I am not really a fan of it staying as is even if Stripe gets to be the one carrying it forward.
What open banking? I can't even get a read-only token for my personal use to see the balance on my account and I sit two team calls away from people who cooked the API
Where it is already operational and available elsewhere worldwide. Unfortunately complex lobbying and Wall Street not being a fan of it as they’re the biggest loser makes it slow to get it done in the US.
With vibe coding, you can make a bot that pretends to be human accessing the banks website, and be able to hit the bot with an API, and have the bot interact with the bank for you.
Every single Open Banking implementation I know also requires everyone to go through a useless certified middleman (eg. Plaid, GoCardless, ...) instead of letting people use the APIs directly
Open banking is slightly different, as it's more about account access (although with read+write access, you can kind of build it into a payment scheme as well, although it's usually clunky).
When you say rest of the world, what countries exactly do you mean? I'm from the Czech Republic and the process is also insurmountable for a regular company. You either go through a middleman or you can apply for a license (only 4 companies in the entire country ever got it and it costs upwards of $50k in legal fees just to file the application).
Can you just link us to the relevant pieces? I rather not search on my own and read wall street propaganda on why it's bad (which is what I'm finding and not interested in).
One problem here with the protocols is the field isn't yet so stable. Most people are copying OpenAI's endpoint, and so you have "OpenAI compatible", but there are always little hairs on it. Case in point, OpenAI itself no longer offers multiple basic inference options on their new models. Can't pass temp, get back logits, etc. on anything after GPT 4.1. It's not a total zoo and Open Router is pretty easy but everything is just enough a special snowflake that models aren't a commodity. Maybe later though.
Agreed, though that's not a reason to discourage building it or putting it forward. That's more the reason why a protocol would be beneficial, we are ending up with price wars and middlemens everywhere because we're racing.
Feels the same - businesses from pizza shops to eCommerce sites will be arguing with not their credit card provider about transaction fees, but their AI provider (who supports all 3 major networks, including Discovai, Clankers Club Aierican Express
Stripe acquiring the biggest model gateway makes sense when you think of inference as the next payments. Both are developer API infrastructure where routing and provider abstraction matter.
It is amazing how we watch from the sidelines these aquisitions and think .... why such big bill ? But it takes a good team to sell and everytime someone is successful it is because probably they have great sellers who believe in their value and demand a price. It is certainly not an easy task to get to the finish line for a startup and pay back their investors.
Counterpoint: you can't just go to Visa/Mastercard or a merchant acquirer out there and set up an account on the same terms that Stripe can.
On the other hand, you can sign up to any LLM provider and get API access on terms that are the same or better (since I'm sure they don't appreciate having a middleman and would benefit from incentivizing direct usage) than OpenRouter gets.
Openrouter is fragile, one good competitor and they are at risk.
Or if one of the provider suddenly decide to forbid openrouter from using their api. Why would they do that tho ?? Well it's not like execs never take dumb decisions.
In my humble opinion, it's extremely overpriced. But then, it's just the standard with AI currently. Divide every ai company valuation by 1000 and you might get it's real value.
Yes, you can get better pricing if you do it yourself. But the true advantage of OpenRouter is that, in a space where there's a new model being released every week, you can easily switch to whatever model is best at any given time without having to set up accounts with multiple providers. Or you can just experiment with the latest release. Their product is the convenience. Of course if you decide to only use a specific provider or two, then you don't need OpenRouter.
But it's not really that difficult to make a clone of OpenRouter's service. What they do isn't really that original.
Their only value comes from the fact that the currently have lots of traffic. And I dkn't think that their cumstomers are really bound to theur servuce. They could switch to a competitor without too much hassle.
Yes, but as l9ng as there's no clone, they're good. On the potential clone side, I guess many will be put off by the fact that "there's already OpenRouter". Also, a clone would need to find a way to make existing OpenRouter customers to switch, which isn't easy.
To add to this; payment infrastructure requires a lot of heavy lifting. There's a lot of regulations you need to adhere to, different payment systems in different countries, settlement, chargebacks, etc.
There is a reason why not doing your own payment processing is a thing.
OpenRouter may have some interesting things in streamlining the process of switching LLM providers, but it is indeed something easy to replicate in comparison to payment processing.
Now that's the value. OpenRouter has the power to step on the air hose of any provider they don't like.
That's Google's real power. Works for them. Between search, ads, and the "app store", they can crush most companies. OpenRouter's power is only in one area. For now.
Having worked on web apps that processed online payments before and after Stripe I totally agree, this was an area of real pain that became suddenly extremely simple because of Stripe. The alternatives were terrible–100 page Word docs of SOAP API docs for Authorize.net, massive PCI compliance requirement specifications, horrible legacy merchant services businesses.
On the other hand, I operate an app that talks to (and logs prompts/meters costs) to many different LLM API providers, and I do not consider it painful at all. I have an AI agent to deal with any integration quirks, if needed. Mostly they provide OpenAI-compatible APIs anyhow. It's basically a no-brainer to go direct with the providers and save 5%, the great majority of the cost of an AI-powered app is no longer dev time implementing the integration, it's the tokens themselves.
Stripe operates in a very regulated industry where new entrance is difficult
OpenRouter is commodity stuff, I've never used it personally and picked alternatives. The sense I got from people who said they used it was that they are on average penny pinchers. That does not seem like an ideal user base
OpenRouter is already the entrenched go-to router tbh. There's competitors (Vercel) but they're already well known. Stripe can also make them more competitive because they can waive / profit off the transaction cost. Considering stripe takes like a 3% fee, and the openrouter markup is 5%, that's a pretty big boost.
Now is that enough revenue to earn back 7B.... I'm not too sure. I guess they're betting realy big that these sorts of model routers will gain explosive token usage in the future
I don't get the 'transaction fee reduction' arguments. Yeah they can reduce the fee, which in turn reduces profit margin... any other product/service in a similar position can do the same... it's not some sort of free win
I presume they are sitting on a lot of usage data and will be able to roll out efficient auto-routing based on it. There is a huge demand for this as most users are overwhelmed by provider options and just want to route to the best/cheapest model based on the task.
I think it's kind of like what GitHub has despite getting rid of the quota system for Copilot: existing customers that are already setup to deal with them.
In this case I imagine that OpenRouter's moat is going to be that businesses that already use Stripe will have a lower bar to choosing OpenRouter to provide the AI access, and clean integration with Stripe to pay for stuff like AI support bots.
The flexibility is kind of what I'm pointing to, if I understand you right. Instead of dealing with Anthropic, OpenAI, XAi etc directly and having systems to handle user assignment, budgeting etc for each, companies just deal with GitHub, which they've probably been dealing with anyway.
They have enough customers to have leverage against providers. As an example, if they decide they can be "punished" if they claim ZDR but lie about it.
As with Cursor it's the data - all the prompts and responses that can be used to train AI models. Getting a head start on that at the pace of current development costs.
They are the leading LLM API routing tool. Stripe is the leading payments API. They are betting on OpenRouter being as important to developers as payments. Both let them take a small percentage of a shitload of transactions. Being the goto service for APIs and executing well is the moat. Someone else would have to beat the volume leader on price or beat them on execution. Good luck. The only realistic competitors would be Amazon, Google or Microsoft and none have shown any interest so far.
I don't understand why one would let a router handle their LLM request (other than for routing it to the cheapest inference provider for open models).
When I build an application that uses LLMs, it's tuned and tested for a specific model only. LLMs are not really drop-in replacements of each other, even different versions of the same series.
The pertinent question isn't why OpenRouter but why a router company is worth 7B.
You can use one key for multiple models, you can have many keys with a budget on each, you can manage all your keys in one piece of glass, you don't have to deal with the upstream vendors directly just to name a few. These are all things that may be important to developers. I can't even get an OpenAI API key because those morons can't get their phone verification to work properly. OpenRouter is the only way I can use their models.
OpenRouter’s “moat” is currently “why would I use anything else?”.
Stripe has presumably only purchased them because they think there’s consumer-surplus to monetize here, which presumably will soon be giving me lots of reasons to use something else.
It isn't clear to me what advantages OpenRouter has over running a similar router-aggregator locally for ~free. I guess they have stronger bargaining power than a solo engineer or small startup, and can therefore negotiate volume discounts with model providers. The question is if those discounts let them offer cheaper rates than a local router-aggregator even after the markup. And if they can do that, then OpenRouter is undercutting the providers themselves on price, which I would imagine the providers would not want to see.
If you meant that generally: because productivity is increasing; the world is getting richer. Not all of it flows down, but some does; top SV salaries are up too. By a lot.
edit: Inflation does not account for the observed growth.
We are certainly seeing higher monetary velocity in certain sectors. That usually follows from real economic expansion but could also be due in part to inflation.
Corollary to this is that whenever companies acquire smaller companies and lay off the small company's sales teams, they deserve what they get. Looking at you, big company that turned my last one from nearly a billion valuation into a $20M fire sale in 3-4 years.
I've seen three or four others building routers. How hard is that really? The folk I spoke to needed only a few millions to build it (they claim). I've seen prototype of some too that looked pretty simple. Maybe the whole thing is driven by enterprise wanting the external vendor support?
The software is "easy", it's the business and operations that are hard: scaling the infrastructure reliably; finding revenue sources with secure margins; deflecting regulatory and risks responsibility; billing and collecting on massive low-margin high-volume transactions; enterprise sales; etc
It's a business line more suited to finance, law, sales, and accounting people than tech people and a pretty laborious one. That's often the case when the tech looks "easy" but the sector only seems to have a few big winners.
- They support 400 different models, 80 different providers, and an unlimited number of new custom agents. Every single model, provider, and agent, has its own weirdness that has to be accounted for. Tool calls change by model. Effort changes by model. Backend APIs (messages, responses, etc) change by provider. There are thousands of specific tweaks, fixes, hacks, that need to be implemented to make this thing "just work". And you have to keep updating it all, weekly.
- On top of that, they support providers running in multiple countries, which increases the legal, logistical, financial, and networking complexity.
- On top of that, they need to figure out pricing and deals with providers, and ensure the providers allocate the necessary networking and compute capacity. They need to work with providers directly to troubleshoot and fix all kinds of issues, from networking to application.
- On top of that, they implement multiple kinds of request routers. An auto router to route your request to the best general model, a fusion model to attempt the request on multiple models and pick the best response, a pareto router to route requests to the best coding model for your request.
- On top of that, they build custom features that businesses and users want, like Data Loss Prevention (which I'm blown away they actually provide for free). Lots of very useful business functionality for managing not only what model and provider to use, but also limits on usage, filters, etc. They also implement SSO, prompt injection guardrails, logging/auditing, workspaces, etc.
- On top of that, you have to be very good at just implementing HTTP APIs. Most people aren't good at it. API design is hard, HTTP is way more complicated than it seems, network traffic shaping is a black art.
- On top of that, running applications and infrastructure, and scaling it 10x every year, is a subtle yet critical skill. No matter how good your code is, it's pointless without working hardware/running apps.
- On top of that, they provide a generous free tier, which has to be subsidized, and is probably only partially subsidized by providers.
Why pay for this rather than build one? Because there's no sense in building one. There's a reason restaurants don't build their own factories to manufacture their own pots and pans. Their business isn't selling pots and pans, it's selling people cooked steak. It's a bad idea to waste time, effort, and money trying to make things you aren't going to sell.
Vibe-coded content & design aside, you haven't implemented all their features, you've got some basic security holes and missing networking, and your pricing is the same as OpenRouter despite your company being 5 months old. It might not be as easy as it seems.
My surface-level assessment is that building a router is easy, but routing efficiently is difficult. For the latter problem, traffic provides signals which you can use to improve routing rules, creating a feedback loop. More traffic => more feedback => better routing. So a lot of the value of Openrouter would be that they are the most adopted router (I think?) and so likely have the most data to (potentially) produce the best router. I'm only speculating.
Probably they want to use the userbase, not the technology? There was a heavy emphasis - too heavy for my taste so it bounces me back - about that they will do what is the best for the user. I think that was already there, Stripe was not needed for that. This is more for the benefit of others.
> I like Stripe's long-stated mission ("To increase the GDP of the Internet").
Also corporate speak. Their mission is take their cut of the Internet's GDP. Nothing against them, they provide real value for that cut. But that's what the actual mission is.
That feels unfair to read it that way. The mission as stated is that they will not only take a cut of the pie but grow the pie. If the total value they add is greater than the value they capture, everyone wins.
You might not believe them, but that’s what the words are supposed to mean.
> You might not believe them, but that’s what the words are supposed to mean.
It's corporate bullshit not because I don't believe they can grow the pie, but because that's not their motivation.
In fact, a nice formula for coming up with your corporate mission statement would be:
1. Ignore your company's ground-truth financial incentives
2. Find a positive 2nd order effect E (or a plausible one) that you don't really care about, but maybe you genuinely think is nice
3. Claim your "passion for E" is what drove you to start your company, and keeps you working, because, of course, all great companies must have a "deeper purpose"
It's not mutually exclusive. I remember the Bad Old Days of trying to make the PayPal button work; if Stripe makes it easier for people to take payment online then it increases the GDP of the Internet.
> Stripe last said businesses on its platform generated $1.9 trillion in payment volume in 2025, up 34% year over year.
More transaction volume = more enterprise value potential. More GDP, total rake goes up of said economic activity they facilitate. Middleman Moat, accumulating volume (present via Paypal, future via OpenRouter).
It's particularly stupid given that if you look at where the tokens are being spent, the top application overwhelmingly is Hermes Agent. It's more than the next six clients combined.
Hermes Agent is AI slop "ai assistant" software that is being entirely developed by a bunch of AI agents. They are stuck in a constant whack-a-mole bug fix session because the agents keep breaking things.
Openrouter is being entirely propped up by garbage-tier software that is only popular because it's being endlessly hyped by idiot youtuber AI Agent Bros who are so uncreative and incapable of critical thought, they're using AI agents to tell them what content to make.
When people finally realize how bad Hermes is, OpenRouter is pretty fucked. And then when all the weebs writing furry porn get bored and move on, they'll be fucked even more.
What's really bad about this: Stripe is so tightly tied to the credit card industry that said industry's obsession with morality policing will force Stripe to start censoring openrouter.
Probably a mixed bag at OpenRouter today. Some employees probably just became very wealthy. Others will see what it's like being owned by a different company with different goals. Layoffs tend to accompany most acquisitions within 6-12 months these days. A $7B price tag probably comes with some hefty expectations.
the toll booth on the AI highway just got acquired by the toll booth on the payments highway. at this point i'm just waiting for the toll booth on the toll booth highway.
I've seen a lot of comments here and elsewhere about why is a proxy worth $8b. It is a bit more than a proxy.
It is a known brand for model routing and it has developed means to route model on their own and 3rd-party infrastructure at scale and without much issues.
Why Stripe? Who knows. I don't see much synergy but why not? If anything it is a way for Stripe to get into the market with agentic payments which may come to be worth a lot more than $7B.
Metronome has been experimenting with real-time payment on usage based billing for tokens. That technology leverages micro payments using crypto. It's cool but all it's doing is moving the accumulated balance from a provider onto your crypto wallet.
Congrats to the OpenRouter team! Been using it since the early days — the ability to try every new model without creating 20 different accounts is still the killer feature. Hope Stripe gives you the resources to keep building.
I check it's model rankings and prices daily. The charts are great. Please never change!
On another note... I'm surprised OpenAI didn't buy them.
I'm even more surprised OpenAI is giving exclusive discounts on openrouter, essentially encouraging the few API customers they had onto a much larger, cheaper and richer marketplace.
The service has been 10/10, everything has felt easy and natural since the start and continually improving. I often learn about new models from OR, and have been a happy customer for years. Kudos for getting the proper recognition, $8B is a fair price in 2026 imo.
> And as we grow, we will relentlessly aim to preserve the velocity, agility, efficiency, and talent density of the 90-person startup that we are today.
Stripe is a high-velocity environment.
OpenRouter (OR) is a startup so might move fast, but unsure if OR can handle the velocity with corporate guardrails put in place already.
Any companies that are bought go through culture crashes, and I am interested in how this plays out.
Routing seems to face a similar future as the ever-changing trends in controlling coding agents: you know, context engineering, loop engineering, etc. --Which is that OpenAI and Anthropic quickly seem to release a better way of doing it, in their own products' harnessing, and which third-party and personal solutions will only conflict with.
I guess this makes sense for the scenario when we will use machine intelligence as a currency, maybe 15-20 years from now? - like I can pay 15 minutes of inference for a dozen bananas? and the banana seller uses those 15 minutes to do banana shelling or removing weeds from their farms or whatever i guess when robots are doing everything. right??
I agree that the computing power will replace not just softwares and also as many phycial work via robots.
I am not smart enough to see further what else are disposable or replaceable
Overthinking it. "You have money and need to seem AIish for the stock market, we want real money". You notice they don't do this deal in tokens, they do it in fiat currency.
there are a few assumptions in this future- the biggest one is - we will have to attain a baseline intelligence which is useful for everyone which is not true today. even fable fumbles hard!!
second is that everyone will have to own a medium to use this intelligence (aka robots). like a currency is useless if you can't use it.
to me this is the least daunting scenario. all other scenarios lead to mass slavery.
> the biggest one is - we will have to attain a baseline intelligence which is useful for everyone which is not true today
That's not the biggest assumption.
The biggest assumption I see is that the population will have the baseline understanding of how to operate in this new paradigm, which seems impossible given how useless many people still are with computers
Incidentally, this is also why I think Bitcoin is such a non-starter for most people
So they essentially resell their enterprise tier access to tokens from each provider while also providing a nice interface/dashboard for their customers?
How in god's name is openrouter worth $7,000,000,000? Can somebody explain to me, or is everything to do with AI just astronomically overvalued these days.
Makes total sense. Stripe acquiring OpenRouter is a massive DX win—unified billing for multi-model APIs will save devs huge routing and token-handling headaches.
Stripe is positioning themselves as one of the leading adopters of AI internally. I don't know of public stats but I estimate they're spending on the order of $millions per week on tokens. I assume they've adopted Openrouter and the majority of that spend is going via Openrouter. So they understand the product value, they understand the current gaps, and there's also a threat mitigation. They've likely become very dependent on Openrouter and someone is going to buy it - so they may as well negate the risk to that. (Threat mitigation alone is not worth $7B but it justifies a certain premium.)
They also have very complementary problems and skills. The long-term roadmap for routers is auto-routing; that is, turning traffic into signals into automated decisions. Stripe has a lot of overlapping talent and experience from fraud-detection and likely other products.
Follow-up: Somehow in all that I forgot the most obvious pairing. Stripe wants to solve how agents pay for things. They’ve been quite open about this. Seems obvious that any solution to that is going to be through the model platform layer - so they just bought a major platform layer.
Their business models are actually kind of similar. Stripe is just passing money between financial institutions and taking a small cut. OpenRouter is basically the same thing. They might have shared ability on how to run an effective pass through business.
Apparently. The price does seem high, but the value of a company is based on it's finances. You could argue companies like Google are just middlemen too.
My read: The thesis is that token spend is going to be one of the larger input costs to a business in the future. They want token economics to happen on their platform so that the commission that comes out of all of this goes to them. It is not enough that openrouter uses stripe exclusively to interact with banks, since the transactions are expected to not just be tokens for dollars, but tokens for tokens or tokens for specific stablecoins tied to tokens. One of the value adds of stripe is their policy and monitoring layer. With tokens, the polices and monitoring will happen in token-land primarily, not in dollars-land. So if you want to continue demanding commission for those types of things, you have to be able to execute policies on tokens. Tokens can of course be variably priced - I am not claiming # tokens itself has the same value across providers.
The platform also has another value that is not mentioned often. People often mention the devex, which is consumer value. But one of the main things about open router is that they help model providers access a massive distribution channel with smooth demand, as with any aggregator. Similar to the value food delivery apps provide to restaurants. As such, this type of marketplace power has the potential to allow them to charge model providers commissions as well for various services in the future. You can already almost imagine an "auto mode" where they balance improvements to cx (say right sized models for tasks, saving u money) with "ad"/preferential treatment promises tomodel providers. Basically, all the typical middleman stuff.
Quoting their rumored leaked investor letter dated Aug 19 26 below:
Zooming out, we see capital and intelligence are becoming the two digital flows undergirding every business. Up until now, every developer has needed a straightforward and reliable way to manage their revenue pipeline, and serving this need gave rise to Stripe. Going forward, however, every developer will also need a straightforward and reliable way to manage their intelligence pipeline.
This observation first led us to OpenRouter. OpenRouter has built the world's largest and most trusted token routing engine, supporting all major models and providers, and beloved by its customers....We think that there are deeper reasons to pursue integration besides convenience. Our experience in working with our customers has led us to realize that intelligence is special: it
is expensive, heterogeneous, and constantly changing. As with financial capital, businesses must reason about cost and return of every unit in a deliberate and granular way. How valuable is this task? With which models can it be best handled? Who will pay, and when, and what is the time-value of that delay?
We have seen the parallels between managing intelligence and managing capital directly in our own products. Radar, for example, was initially designed to prevent financial fraud, but is proving extremely effective at guarding against token fraud at many of the world's largest A1 companies. Metronome (used by Anthropic, Nvidia, and other industry leaders) is showing that metered billing in an AI context is inseparable from token serving and consumption itself.
One of the main real values of OpenRouter is that it is a single payment relationship for users that enables access to many downstream vendors. I'm not super convinced it is a good purchase, but right now OpenRouter is the financial middleman for token spend. It is also the centralization point for tokens, allowing for value-add features that are industry-wide, for example budgets - and I think that structure has parallels to Stripe products like Checkout or Identity.
Stripe has a lot of money, but not a lot of AI-buzzword compliance. OpenRouter is highly AI-buzzword compliant, but doesn't have a lot of money. It's a match made in heaven
In the event that the AI bubble doesn't burst, Stripe needs to have a stake in whatever's coming. They have time to explore and pursue specific strategies, but they just needed to make some big and compatible buy.
In this particular case, OpenRouter represents a M x N bridging and enhancement layer much like Stripe themselves, and so must address a lot of parallel technical, dealmaking, accounting, and legal challenges.
It may not be obvious on the surface because they seem to be working in such different domain, but they have to address a lot of the same problems in comparable ways and that makes it a pretty darn good fit.
I am somewhat mixed on this, I feel though having it be part of Stripe will make it easier to sell to my leadership as "More Trustworthy" than having it by itself
Not if you’re in a different domain, then it cannot be claimed to confuse customers.
So if it’s a network router or a woodworking router, then you’re good.
I can sells of red socks, and white socks, without getting sued by Chicago or Boston.
If it’s in the same space… you might want to contact a lawyer. I believe Firefox had conflicts with earlier names like Phoenix and FireBird and they weren’t even browsers.
If you're looking for a privacy protecting version of OpenRouter, my friend Joseph Perla made trustedrouter.com.
It's a pretty slick system and if you're building on openrouter today and are worried about stripe integration messing up the product, this is a good option to consider.
Separately, congrats to the OpenRouter team. They are selling at the right time in what is undoubtably going to become a contentious market.
10% would be more typical. Perhaps the math is that OpenRouter inside Stripe makes it twice as valuable. You often see acquisitions priced on the value of the company post-acquisition.
The 7 billion figure is not just open router's current value. It's also including part of it's projected value in the future. So that 7 billion includes current value + value say 5-10 years out.
Remember when Google acquired YouTube for $1.6B in 2006? That amount seemed ridiculous at the time. Its now peanuts compared to YouTube's yearly revenue.
Congratulation to all who exit at $7b after 3 years, but I have only one question: what problem does this acquisition fix? Did OR lack infrastructure? experience? what was the main drive, other than money, for sale?
Because after PayPal, Stripe is the biggest graveyard of frozen accounts who would be otherwise perfectly fine, but some AI decided that this card from this country at this price = send the account to null.
So while before you had comfort of using OR, right now you will have Stripe emps and their AI overseeing your queries. And don't get me wrong - Stripe will love you! But once someone pulls the lever - good luck getting at least explanation why they had to boot you out. I never heard of an account being reinstated. This will be like this, only your queries instead of CC charges, will be inspected.
I don't see how this is beneficial to anyone, esp. OR users. RIP OpenRouter.
It's honestly insane how valuable so many startups end up being just from AI, wonder if therels ever going to be another breakthrough like this for future people
Just as the US deems China as a national security risk, I deem Israel to be a much bigger threat to my cybersecurity. Stripe has vocally supported Israel and I cannot trust any token coming out of OpenRouter after this. Their payment systems have been notoriously bad so this is cause for concern to me.
Stripe is a financial institution that has strict requirements for reporting and audits by the US government. It has blocked payments for numerous different services for no reason. I wouldn't trust any infrastructure built on OpenRouter.
I dont use open router so can someone explain to me how it works in that different vendors have different api parameters?
is the selling point that they just give you access to a unified access in where you get to write each request toward the particular api provider's api spec? if so why not just sign up with the vendor directly?
I would guess the real value of the position Openrouter has in the ecosystem is that they can intercept all data and sell training data from many people, training data that is about as directly relevant for training as it can get. They would not need to log all your data for this, "sampling" is enough, and might in fact be even better sampled.
Note: this would include both the request, the response, and competitive responses. This would be very high quality training data.
Now at the current point they're sporting and make this data sampling option opt-out, and indeed they do. By default, they collect your data.
Model companies will pay a lot for this, especially now, more than customers will ever give to openrouter. Think about this: this is a constant stream of data of exactly what people and companies do with AI models, for all models, updating LIVE.
Second, they can provide "competitive intelligence" as well, telling models live how well they're doing and what exactly their strong and weak points are, and in what distribution of token requests this results.
There may even be financial companies that pay for this data.
so basically anybody who uses openrouter from now on is going to have all their data exfiltrated to stripe and then to all the banks/insurance companies... when I read the article that's what I think is going on and it doesn't smell good....
> Today, we are excited to announce that we are...
Why do they always start with this. Every single one of them. You dont even want to read anything after that. Its the same "Blah blah nothing will change you will get the best of both worlds blah blah"
I don't see the use for open router, I can switch API providers without any hassle with just a couple of lines of code and a drop down. Good for them and for those who milk the AI bubble
As a final customer, I hate so much stripe that I'm disappointed that they will be able to impose their shitty rules and business logics to OpenRouter...
Sorry - that probably came across as more passive-aggressive that I mean it. Just wanted to point out that guideline. People have easy access to AI, so they can always ask it for a summary if they want to and it doesn't really add anything to the discussion.
At this point I believe that when people do not cite using AI to generate their comment it's from a position of malice... like you're intentionally trying to position yourself as capable in whatever subject you're talking on but realistically you are not given the reliance on an LLM. Not saying anything about OP but just a general observation/opinion I've been building recently. Seems more common than ever to larp using AI.
I did it once when I tried vibecoding, to a friend, nothing professional, and I watched myself stop at the explanation of how I did something. I think it comes from a place of desire, self delusion and shame (of not being able to do it alone), very similar to lying in general, people lie about what they want to be the truth.
Is that 100M of openrouter revenue? Or money that flows through openrouter? If the latter, 100M$ annualized looks like very little. It's as if a bank said that it has 100M$ of annual transfers, instead of 100M in revenue from bank fees and such.
Hard to think of any type of company that can do the topline meme as effectively as a compute aggregator proxy. Almost like a bitcoin pool mining company counting their shared miners.
Really? 100M annual? That HAS to be profit, it can't be this crazy.
EDIT: From their website, they say they have 200T monthly tokens, at 10 cents per M token ( optimistic average between free tokens and expensive tokens) that's 20M$/month, which is roughly 240M$.
Bonus track: here's chatgpt completely fumbling the above very simple calculation
We are in a bubble, but you are looking at this in the wrong way. OpenRouter basically creates lock-in through Stripe's distribution and family of products; it is much stronger within Stripe than alone. I agree valuations are absurd, but it is what it is. But if we have to go through a bubble for a16z to crash and burn, I am all for it.
So customers use OpenRouter to avoid lock-in to specific model providers, and in doing so they lock-in into openrouter's aggregation API?
Lock-in to what exactly? A string(string) function without any further restrictions? Note that aggregator users will not use more specific parameter features (because those vary by vendor), or they escape the aggregator and lock-in directly string(string, vendorOptions={"openAI.logprobs":true})
Selling vendor lock-in to vendor lock-in avoidant customers sounds like a losing proposition, like trying to beat competition on price and undercutting, which is a thing openrouter does too btw, the main reason people use these products is to get cheaper prices. I don't think 2$/M tokens is expensive, and I don't think those that try to cut costs are going to win, whether devs cutting token costs, or small business owners with 5 employees who look at AI as a way to fire 2 of them. I get that the other end of the spectrum is overspending and tokenmaxxing, but the conclusion is that extremes are bad? Nothing new here.
I love OpenRouter, long time user. Stripe will hopefully be a good custodian.
I just want to point out some features of OpenRouter that make it more than just a model selection and routing endpoint and that I find incredibly useful:
0/ Default routing is to the cheapest provider, but they're usually not the most performant. I'd guess 99% of OpenRouter integrations never tweak the default routing. Here you can setup cheapest with performance minimums:
https://openrouter.ai/docs/guides/routing/provider-selection...
You can also stack model selection in priority
1/ Using broadcast you can push all your analytics to clickhouse / s3 / snowflake and a bunch of other compatible destinations. Setup a clickhouse server ($5 VPS[0]) and send all your traces to it:
https://openrouter.ai/docs/guides/features/broadcast
customise your own observability in your dashboards from there. Superwin
2/ Model router is also a natural home for llm security - OpenRouter has the beginnings of prompt injection detection:
https://openrouter.ai/docs/guides/features/guardrails/prompt...
there is also PII detection. This will show up in observability as rejections/blocks etc.
There are so many model routing solutions (same with observability, security etc.) but they're all 80% solutions - OpenRouter really rounds out with well implemented features that you need when deploying models at any scale and I gladly pay the toll.
[0] not sure if these exist any more but clickhouse is resource efficient
> Model router is also a natural home for llm security - OpenRouter has the beginnings of prompt injection detection
Maybe I'm biased from the perspective of a "harness provider", but I think Model routers often have too little context to act as well-informed prompt injection prevention. e.g. it lacks context of where which part of the message(s) originates from and sanitization/safeguards were already performed on the application layer.
Something like OpenRouter's "flag" mode is fine, but usage of auto-redact or auto-block should really only be used if there is significant risk exposure through your harness or otherwise they are a constant source of bugs.
noob q -- does the routing hurt cache hit rate?
By default yes bit you can pin providers to avoid this.
Of course, of routing the same session to different providers.
But don't worry, many providers there charge for cache the same price like non cached ;)
I have been using openrouter only superficially. I didn’t know about the broacasting feature. Thanks for that.
They might be popular for indie developers, but nobody doing serious AI or in a corporate environment is using them.
And if they are, their compliance team is about to strike them down. The VCs forcing this acquisition do know this.
- Why would you add a penalty of 50 ms at a minimum? And that is not the p95... Just run LiteLLM in house and you dont even really need that.
- Their capacity pools are shared across the whole user base, a massive batch processing by another of their customers and think what that means for your response time...
- So instead of negotiating corporate rates with OpenAI or Anthropic, you would be using an intermediary and topping up the corporate credit card... for a 5% markdown ? Really?
- They can see all your critical corporate data on the in and out
- They present some pink SOC 2 promises but then wash their hands and defer to you and the providers. Its just the Bolt and Uber model the drivers are not our employees....
- They are a man in the middle proxy that is a massive security liability for your corporation
- They have no support for private cloud points
- No geofencing guarantees
- No intellectual property legal indemnification unlike what AWS or Microsoft or Google offers
- Its a provider roulette inconsistent with hosts providing different quantization levels causing random shifts in response quality
- Support via a Discord server...
The only reason they were not shutdown yet by Anthropic or OpenAI is because they have the same VCs, as those two. That would mean said VCs investment would go to zero. Oh and those are the same VCs that own Stripe...
Just setup a private proxy tier using something like LiteLLM, even if you really dont need it. Just code your enterprise apps to have have fallback loops on the core hyperscaler providers like AWS Bedrock or Azure Foundry...
Yep. https://news.ycombinator.com/item?id=49344067
> features of OpenRouter that make it more than just a model selection and routing endpoint
All those features show it's just a model selector and router.
Meaning you can switch in seconds. Stripe just has too much money...
This seems like something xAi could reproduce in a weekend
Could but hasn't. What exactly is xAI's USP compared to the others?
The other thing OpenRouter gets you is: Access to loads of models via a common API/sign-up/prepayment mechanism.
I want to know which vendor/model does best at my extracting-facts-from-text task? Which does best at my OCR-a-text-document task? Which can deal with a safe-for-work beach photo without a censorship system false alarm?
OpenRouter lets me run my tests against openai and anthropic and google and x and bytedance and qwen and llama, with a single sign-up and a single payment.
You can do that with AWS too.
Nobody would trust one of the labs that have their own models with this. They’d assume they’re taking all the data, and biasing the routing to favour themselves. xAI would probably be the least trusted, but the same would apply for any of them.
Everyone could reproduce this on a weekend...
xAI just not in the game, never was.
Using Open in your company name, when there's nothing about it, should be illegal.
I think it is fitting because they provide „all“ models independent from the provider. Open in this sense means to me being open to switch the model, speed and provider at any time.
And don't get me started on Sesame!
Me?
I used to demand the same but a word can have multiple meanings depending on the context, eg OpenMind means something specific. Open{$IT-term} does not necessarily mean Open Source {$IT-term} and OpenAI might mean an AI available to all people, not implying anything about cost. What could be illegal is the word Free, which a for-profit company would never use in the first place.
Great product, been using it for a while.
Turns out even a proxy can be worth $8bn with the right business model behind it.
Users get an array of providers competing behind a single API, meaning they have to compete on price and quality not vendor lock-in. This encourages users to join OpenRouter over specific model vendors.
Providers get easy access to revenue (and data) and new customers with little to no ad spending, encouraging them onto the platform too.
And that's all you need. Win win.
Well done and congratulations.
Someone noted they pass through 100 billion annually if I remember correctly, but they also take a 5.5% cut, so Stripe secures them as a “customer” while getting their money back over time. It also to me feels like an adjacent business aligned with sort of what Stripe already does to begin with, being someone elses middle man.
That feels high if they were bought at $7B? 5.5% of $100B is $5.5B in annual revenue. 1.3x multiple for AI. No way
100b sounds implausible if you consider that anthropic's revenue is only expected to reach 100b this year, 100m is more likely.
100b is too high, but 100m is way too low.
Who is that someone? And unnamed employee that told somebody else?
They make less than 140 million in revenue, what today is like 5 signing bonus for OpenAI :-) They are not profitable and see through less than 2 billion in revenue.
These are easy to check so...
I've been using OpenRouter since relatively early (I build evvl.ai - an eval platform on top of it) and think that they're selling at a good time.
An underpinning of their model is that API calls / inference are similar across providers allowing for commodity tokenization cost comparisons, but the providers are beginning to shift to non commodity features that don't easily shift.
For instance, calling Gemini with "search grounding" isn't something that OpenRouter can do (they sub their own web search in), last I checked they weren't doing real time voice models, etc.
> For instance, calling Gemini with "search grounding" isn't something that OpenRouter can do (they sub their own web search in)
It can: https://openrouter.ai/docs/guides/features/server-tools/web-...
It's apparently even the default these days (which makes sense, as it's usually better in my experience).
Nice. Happy to be mistaken on that point as that seems very useful.
Your main point is still right, the providers are beginning to shift to non-commodity features.
Another example might be all the managed agent features, where you get a VM + model, right now it's mostly Google and Anthropic that have this offering.
The service tier is another, where Google, Vertex, OAI, Anth offer it but only OAI and Google offer flex service tier and not all models from them get it.
OpenRouter for me is an abstraction over that complexity, they will have their work cut out for them.
Fortunately Stripe is in the business of providing a single abstraction over an enormous pit of complexity, so I think they'll be in good hands!
I can see Stripe adding extra services on top. However, in their main business its "easy" for them because getting into payment processing is hard intially. But whole search, harness, VMs, agent loops are not hard to find somewhere else.
I think it has to be more than just a 'proxy' what Stripe bought.
Stripe and every other big company could build this without any issues.
It either is just a really stupid business decision or it is about the name. The only model proxy i know is OpenRouter despite plenty of other model proxies existing.
It's a very straightforward business model and they started just in time to become popular as there are hundreds of alternatives out there.
I also congratulate the founders for pulling this off.
> Turns out even a proxy can be worth $8bn with the right business model behind it.
People said a similar thing for Cursor simply being a VSCode wrapper for when it crossed unicorn status.
Microsoft is really really bad at marketing and making products consumers actually want.
I really think they only know how to sell Azure.
No reason Microsoft couldn’t have shipped a better version of Cursor considering they wrote the IDE cursor is based on.
Given the costs of the providers I doubt that sharing the revenue is a win for them.
They are fighting to be the one and only
no company will reject revenue :)
But every company wants more revenue.
The smaller, less known could get a win, but the bigger lose.
Found a good european alternative: https://cortecs.ai/pricing
Probably the main EU alternative to Openrouter is actually https://requesty.ai, more fully featured and equivalent model access.
It looks like it's based in London, not EU.
Indeed. The previous one doesn't even have glm 5.3 listed.
Awesome will check this out. EUrouter is too cost prohibitive for me, and now that open router has been acquired by stripe it’s only a matter of time until it’s unusable
They seem to be about twice as expensive on a few openweight models I've checked.
Thanks for the suggestion.
Alex (the OpenRouter founder) and I went for walks to Sightglass coffee on Divisadero St during our weekly office hours @ HF0 in the Fall of '23.
On one walk I asked 'why are you doing another startup?'
For context, his last one, OpenSea, was valued well into the billions, so it wasn't for money.
His reply: "I just love solving all the puzzles."
It's incredibly hard to compete with someone who is playing the game for the love of the game.
Kudos on playing well, Alex.
OpenSea is full of scams, has not added any real value to anyone or society.
DOn't get me wrong, great for him to make money and being able to move fast and succeed, but you could play this game to if you want.
OpenRouter is great business I agree, but I'm not still convinced how the integrity of providers' models is ensured. In other words, can't the provider serve DSv4 flash advertising it as DSv4 Pro?
I'm aware that OpenRouter checks response quality onboarding, and does further checks occasionally, but I'm concerned that it's basically a cat-and-a-mouse problem between the scammers and the detectors. For example, there could be a signal that a specific pattern of requests are from OpenRouter's quality testing bots. Or, they can just route 1% of requests to an inferior model and benefit a small gain, hoping it fits into the statistically allowed margin.
How is that a problem that's unique to openrouter? Who's to say when I query the Anthropic API directly, wanting to use Opus, but they determine it doesn' need Opus, route it to Sonnet instead and pocket the difference? The only difference with openrouter is that it's another layer where this type of fraud could occur.
OpenRouter's list of providers is very large and includes many little known and sketchy companies.
Yes, it is awfully inconsistent today. They run some tests (accuracy table buried on model page) but they are sparse and only capture a single moment in time. I would love to see OpenRouter take this more seriously.
AI products are going to have to deal with accounting. An agent performs some work. It uses some models and perhaps some other metered services. Someone has to meter that activity. They have to attribute costs, apply the product's pricing rules, bill the customer, collect, reconcile with vendors, and maintain the ledger.
Stripe can use OpenRouter to build the financial and accounting infrastructure for every product that sells metered AI work.
I think the analogy is ADP. Payroll for all the work that's going to be done by AI agents.
Every token/llm provider already has usage metering...so not sure what exactly you're imagining stripe adds here
A single policy location for multiple models with usage limits/rate limits, spend limits all configurable using a common syntax + deep integration with account levels and high visibility into subscription/payments related to meter usage to help understand whether or not your pricing model is underwater as costs change is exceptionally valuable. AFAIK each product consuming model providers without an openrouter style proxy has to do that work themselves, and if you are already using stripe for subs/payments, the direct integration is a huge value add for customers, especially if you can alert on "upcoming price changes from your provider changes your unit economics for XYZ packages/sub tiers".
You're just describing what OpenRouter already does?
The first half yes, the second half, I don't think so, because that would require OpenRouter to know what you are charging your customers.
Similar to how Stripe is a middleman for payments across (fragmented) banks, they want to be a middleman for (fragmented) AI models as tokens are the new currency.
As AI agents/harness/human are spenders of tokens, enabling them to derisk from being locked to a specific model provider & allowing to (re)route to any model at anytime for better leverage in a single API, similar to how they are doing for payments.
paying for data access, for "deterministic computing", for letting the AI use online services. Maybe even micro payments for using tested and verified skills/mcps.
Example: access to real-time stock trading data, access to weather information, letting it make stock trades, etc.
At first I was confused why Stripe bought OpenRouter, but I think this makes sense.
Aggregation behind one interface, just like the actual token product.
That's what OpenRouter already does, so the Stripe value add here is still unclear
Not OP, but I could see the vision. Right now OpenRouter is developer-paid with no way to pass costs onto the end user. Stripe could help bill and meter at the user level for per-user billing. I would also think more SDKs/easier ways to add AI features (paid for via openrouter) into apps and websites.
Open router supports per-request usage metering, as well as per-key. Building user-level metering from this is trivial, and there are many implementations of this out there. Supporting per-user first-class is trivial considering Open router already has per-request and already supports per-request attribution. How important that is, is a big question, but definitely not something that requires Stripe to build.
Could there be use cases where you want to run inference for users in your app, but don't want to host your own backend, user authentication, and billing?
The user could log into his personal OpenRouter account, authorize your application, and optionally set a budget, all in something like a Stripe payment screen.
> The user could log into his personal OpenRouter account, authorize your application, and optionally set a budget, all in something like a Stripe payment screen.
Like standard oauth?
https://openrouter.ai/docs/guides/overview/auth/oauth
Good point, just needs an user friendly flow to sign up and set the budget in a way that is more similar to Stripe checkout than regular account creation.
OpenRouter already supports this, requests return their exact cost.
IDK, that sounds like a solution in search of a problem. What would be different about this than what OpenRouter is currently doing with pricing?
Metering as a service?
Perhaps stripe thinks metering is a lovely big market? A nice complement to their existing service. Plus a little AI buzz likely helps their valuation. OpenRouter could be a step rather than a goal.
It's not just that.
As user agents become more common it's only natural that they will be used for taking the heavy lifting out of e-commerce purchases. There will be a big need for digital payments to verify and reconcile these purchases.
With LLMs that are plugged into digital payments we will essentially have buying agents in our pocket that can find us exactly what we want for the cheapest price and the quickest delivery.
At some point AI will be able to deal with the repetitive and template based parts of accounting better than accounting.
Which is saying a lot.
I'm not sure that's true. LLMs seem too unreliable for accounting.
I'm always surprised at how many people in the tech sector don't think that things can and will change. In the next couple of years? Sure. 5+? Who knows. Necessity is the mother of invention.
I'm always surprised at how many people in the tech sector assume overcoming fundamental LLM weaknesses is always “just a matter of time,” and that time will arrive quickly enough to be relevant. If you haven’t seen any of the videos of people trying to get frontier LLMs to count consistently, you might think we’re further from LLM accountant than you realize. It’s not a data problem— I’m pretty sure they’ve encountered numeric progressions between 1 ABs l and 100 before.
And there is no necessity. We have accountants.
I'm not just talking about LLMs though. People thought the internet and email were just a fad that would never catch on. People thought that the idea of home computing was absurd. All because they only saw what was in front of them at the time.
Also, how many accountants do you know that are truly happy with their work? I'd like to think many of them would love to do something other than crunch numbers all day.
> People thought the internet and email were just a fad that would never catch on. People thought that the idea of home computing was absurd.
We're far removed from "we've sold all the computers the world can buy" (back from when computers were the size of a building).
This is a disingenuous argument. The web and email were basically instant hits and people realized it. Similar story for home computing once the computing power caught up.
> I'm always surprised at how many people in the tech sector assume overcoming fundamental LLM weaknesses is always “just a matter of time,” and that time will arrive quickly enough to be relevant.
Lots of techies are tech-optimists ("tech always improves quickly").
Lots of people also have dollar signs in their eyes (or related, such as increased visibility and scope).
Hard to tell which is which.
Things can easily change for the worse, such as a stock market crash. Who knows? Will we get AGI or will we see more homeless? History says... AGI for sure!
It's very difficult for this sort of thing to change for the worse. It's basically impossible on the software side. If next year's model is worse then keep using this year's model. Hardware technology can potentially go backwards, but things have to get really bad for that to happen.
LLMs are inherently unreliable. We do already work with unreliable technologies for many things but there's this engineering principle called "use the right tool for the job".
And it's a lot more likely that LLMs can't be changed to become unreliable, it's just how they work. So we would need more basic research, that doesn't grow on trees and for which the timelines are basically open ended. Maybe tomorrow, maybe right after cold fusion hits mass adoption.
I built modelrelay.ai that does this. Happy to give you an invite code if you want to play with it.
I'm not happy about this. Having used OpenRouter a lot and enjoying the experience, I need to say that corporate consolidation is never good for consumers.
>I need to say that corporate consolidation is never good for consumers.
Truer words, never spoken. I'm not sure how exactly this will screw me over -but I do know that it will.
Then switch. The good news is its’s fairly easy.
Great advice, until all of the alternatives are driven from the market ala youtube, ala' ebay, etc etc
EBay there are alternatives (gumtree etc.).
YouTube there’s no real alternative because hosting unlimited amounts of video for free is a money pit that’s impossible to turn into a viable business.
If some big corporation wasn’t willing to subsidise it for some ulterior motive, it simply wouldn’t exist
What other routers exist that are good in your experience?
Eurorouter
Can't find anything by that name
Maybe he meant EUrouter? But it seems to be an entirely different thing: unified API for models vs a token marketplace
Cortecs.ai
Only EU providers, seemingly huge latency, >2x pricing is just in no way an alternative.
https://cortecs.ai/detailedServerlessView/deepseek-v4-flash-...
https://openrouter.ai/deepseek/deepseek-v4-flash-0731#provid...
curious as a non-user, but what does openrouter have to do with stripe? Ignoring the corpo-speak about mission statement.. the missions don't really seem aligned in reality.
I don't think they have much to do with stripe. I'm guessing they want to round out their business more and include payments for AI since OpenRouter is basically just an AI traffic router that takes your money and provides it to the APIs you use.
I wrote a substantial comment about this a few days ago: https://news.ycombinator.com/item?id=49324405
My guess: your LLM agent makes buying decisions for you. Providers of agents can integrate witg stripe. Those using openrouter click a button for micropayment support.
That's a rule but then Stripe won the market by having very friendly interface and minimizing the pain of dealing with paperwork for merchants. When PayPal (the worst offender, just hostile company) acquired BrainTree it was clearly a terrible news. When Stripe acquires OpenRouter I wouldn't be so sure.
This must a purely aesthetic judgment apriori. Many services of value to me failed for lack of capital.
Happy for the OpenRouter team. Been using the platform since early stages and the ability to run any AI model with the same api key and same request has been great to experiment with new models and being able to switch models in prod with minimal effort.
They also support fallback by default so you don’t have to write wrappers and logic to choose models, it just works with their SDk using config.
The open router team was good people with a good product. Hopefully they can move on with their lives
Good for OpenRouter. They have a great DevEx. The $7 billion is tad high, but Stripe can afford it.
It's a mostly equity offer. Private capital is always more risky than publicly traded shares so probably less than $7bn risk adjusted, which makes sense for Stripe.
What's good about their DevEx?
I manage our company's model spend and build tools around it.
The good thing about the developer experience is that it is dead simple to create keys that are time-bound, model restricted, and have budgets against them in an easy, programmatic way (also in bulk). Costs are guaranteed to be on par or lower than any model provider, so it's an easy sell to the finance department.
To look at it from the other side, Gemini is the absolute worse when it comes to key management. The labyrinth of actions that one has to navigate to generate budgeted keys for a team is not worth it and I still haven't figured out how to manage the keys at a granularity I need.
OpenAI and Anthropic both have what I would consider to be middling DevEx. Most things work, but only on their platform, and there are some weirdnesses, like the fact that anthropic disabled its admin API key creation, so you can't programmatically create keys against a budget.
> Gemini is the absolute worse when it comes to key management.
I saw another comment thread on this recently. What I gathered was they hid it in the Cloud Platform, so for those who jumped the cloud hoops already, maybe it's alright. (However when I read what Openrouter offers, as you just laid out, I feel Google can't hold a candle to Openrouter DevEx.)
However, if there's anyone who is aware, was there a brief point where getting a Gemini API key was simple (enough)? I'm asking because, in true Google fashion, the AI product went through a lot of changes.
I work on DevEx for OpenAI's API - I'd love to hear more about what we could be improving here!
Not OP, but you can't set a per api key limit in OpenAI Platform, you also can't really easily see spending going on per key, and its also in a far off page separate from the keys, really not easy to see what is going on with spend in real time.
You're supposedly worth a trillion dollars, can't you know... pay for your own market research rather than forcing the public to provide it for you? Is this what they seriously teach devs there?
God you're miserable. Damned if you do [engage with the community directly], damned if you don't.
I am working on my own agent implementation (https://usehax.dev/), and I can praise OpenRouter for being seemingly the only provider endpoint that actually gives proper API for gathering available models and their capabilities (context length, image support, pricing, etc), reports exact routing details and total cost in the response, and provides an endpoint for obtaining current available credits (for "/usage"-like command). So for OpenRouter specifically I could deliver a really good experience out-of-the-box (always fresh models and metadata, exact pricing, etc).
You would think that these things would be standard fare, but they really are not. Other providers need metadata from a separate catalog, don't report pricing, don't really report available credits / usage either.
So yes, OpenRouter does have a good DevEx.
Speaking for myself (I'm not the GP comment): The platform 'just works', has great reporting interface to help understand excatly what usage happened when, and integrates well with lots of different harnesses.
It works
You create an account. Pay money. Get API key. Use APIs.
But if you use any decent amount of tokens, it's probably worth saving the 15% by moving to the provider they're proxying.
And if it's not worth it, then are you spending enough where it even effects Openrouter's bottom line?
With enough devs playing with hobby projects I'm not doubting it's profitable, only that $7 billion seems way too high.
Hint: It's not 15%. And there are many reasons, such as not having to keep up-to-date billing details in 70 providers, and not wasting money because most providers want you to prepay a balance that gets stuck in there if you switch to another provider.
It also has way better uptime than the underlying platforms, even for proprietary models like Claude. When Claude APIs are having issues, OpenRouter Claude still keeps working because they can route to AWS Bedrock instead of Anthropic etc. This effect is even bigger with open-weight models because they typically have 5-10 providers.
OR I can keep paying 15% markup till next month and then jump to newest / cheapest model with one-liner change instead of being locked to a model/provider.
Which model/provider locks like this, most of them use chatgpt style apis anyway.
That requires that you create accounts and separately manage each provider you want to use. OpenRouter removes that - if a new provider gets added, you can start using them immediately.
So, you’re paying 15% tax on all AI costs to avoid creating an account?
Aren't they all the same api? Are you paying 15 % to change one line, and not, like, three?
15%? Isn't the fee 5.5% with lower fees for enterprise customers?
They charge a premium of 10%. I would imagine that's a lot of revenue.
Or rather, Stripe will find a way to make me pay for it as a Stripe and OpenRouter customer.
Filing under things I wish I knew earlier
Can someone help me understand why OpenAI, Anthropic & other proprietary models would want to make their offerings available on OpenRouter?
It seems strategically not in their interest.
They get paid the same either way, Openrouter takes their cut and passes it onto the consumer. And anyone with an Openrouter account has already escaped from their proprietary moat. So for those who have expressed a clear preference for provider-agnostic model infrastructure, do you want to just wall yourself away from them forever?
And per the rankings [0] 18.2% of the tokens sold through went to OpenAI [1]. 3.8% for Anthropic though. Google's up there at 23.4% primarily because GCP/Vertex/whatever they're offering is a billing nightmare and the only way to set a hard cap and sure you won't wake up the next day with a $10k bill is via OR.
[0] https://openrouter.ai/rankings#market-share
[1] Although gpt-oss 20b and 120b are indexed as 'openai' because they're the creator even though they don't serve gpt-oss via openrouter. But neither are in the top 20 of models this month, while Luna and Sol are.
DELETED
Read my sentence again. With your human eyes and brain:
> **Google's** up there at 23.4% primarily because GCP/Vertex/whatever they're offering is a billing nightmare and the only way to set a hard cap and sure you won't wake up the next day with a $10k bill is via OR.
OpenAI absolutely lets you set a hard or soft cap. I've set them and they've worked, I think one time I saw a $-0.01 balance. Their docs were probably written by ChatGPT. Later in that article it says:
> A spend limit can monitor spend without enforcement, or you can enforce it as a hard limit so API requests fail after spend reaches the limit.
And they still allow the most certain limit, which Google does not: have $0 on your deposited balance and don't have auto-reload enabled, you get 4XXed instead of going into debt (or at least by at most a couple cents).
Although for fun, go ask Gemini how you can set a cap on GCP spend. It'll try to write you a pub-sub client.
They already sell API access to the model on Google/Microsoft/Amazon cloud. OpenRouter is the same deal. Note that the labs are not giving out model weights to OpenRouter. They are just selling tokens.
OpenAI/Anthropic/etc don’t establish the direct customer relationship.
And OpenRouter can steer traffic away from those labs over to their own model.
The major labs don’t get personalized data about the user to train on, since it’s aggregated.
Which is just some of the reasons why strategically it doesn’t seem in their best interest to allow a middle person.
It's a land grab. Right now none of them have a dominant position. Much better to get mind share and market share now. They can always squeeze the value chain later. And whoever wins surely will.
Quite frankly, the proprietary company's products are no longer good enough for them to be able to enforce a moat. The scene is popping and nobody is satisfied staying with one provider. If they stopped selling via Openrouter, the 18.2% of Openrouter's revenue that currently goes to OpenAI or the 23.4% of their revenue that goes to Google would probably dry up. Some people would switch back, but those who are happy to juggle multiple budgets and endpoints and API key management infrastructure are already doing so to avoid Openrouter's 5-10% cut. This is really evident at any $bigco where you can wait for weeks for a purchase order, needing leadership's approval and accounting's "I'll get to it when I get to it."
> OpenAI/Anthropic/etc don’t establish the direct customer relationship.
But they also don't have to pay to support and maintain the direct customer relationship and all the billing, support, liability, etc. The credit card merchant fee (a not insubstantial % at scale) is paid by Openrouter (or Stripe now) and the bill Openrouter pays to OpenAI is transferred probably via wire or something with a lot less fees. Meanwhile if you're paying OpenAI via credit card directly, OpenAI is paying Stripe or whoever.
> And OpenRouter can steer traffic away from those labs over to their own model.
The inverse is also true. If you've deposited money on a Openrouter account, you've taken a very big step across their proprietary moat. Would the proprietary companies rather lose them as a customer forever, or be able to entice them back, which can be done with a single change of a JSON parameter?
> The major labs don’t get personalized data about the user to train on, since it’s aggregated.
(Edit: Looked it up, Openrouter does send a pseudonymous but stable hashed userid upstream, stable across different API keys in the same account [0]. That makes de-identification a lot easier.)
First, a lot of the corporate use is getting ZDRed, so there's nothing to track without going back on zero data retention.
De-identification was trivial for normal web traffic before LLMs. Think of all the crap people's harnesses and claws send without de-identifying. Many coding agents can't help themselves from scanning your home directory.
Sure, the type of person to comment on HN probably has good hygene, but the typical vibe coder doesn't even know what they're leaking.
[0] https://openrouter.ai/docs/cookbook/administration/user-trac...
When the market settles down, most of the money is going to be in bulk inference, for batch jobs. In 5 years everyone's laptop will be able to run Qwen 3.8 27B for coding tasks, but businesses will still need to run inference 24/7. Very few people will still need SOTA models once you can run an Opus 4.6 class model on your laptop.
Could someone explain why these routers are so valuable and pulling acquisitions in the billions?
It’s a marketplace with a markup on every token they sell. And I’d rather go to this shop, than sign up individually at the 70+ different providers they broker access to - even if it comes with a price.
Frontier model providers (Anthropic et al.) gate new accounts to impractically low rate limits and spending caps until clients unlock higher limits with cumulative usage, make equivalent cash deposits up front, or talk to a Sales department to work out some other arrangement. They are handicapped by the postpaid billing model.
It's an administrative burden to sign up very every model provider, and there are many independent inference providers now that serve only open source models.
OpenRouter provides a useful service by allowing easy prepaid model access with much higher rate limits, and they also aggregate different model providers to route queries by price, latency, etc.
I don’t understand it, but good for OpenRouter. Personally, I think OpenRouter’s value is somewhat perishable. In the early AI market where there are a jillion models and anew one every week, it’s great to be able to try them all without having to create accounts and keys for each one. OpenRouter, which I use, makes this easy for developers to do. Add some cost controls and other “management” knobs and it works great. But the market is not going to be this frothy forever. As things settle down and commoditize, the value of switching on a dime diminishes as people lock into their favorite models. And with the OpenAI and Anthropic APIs being defacto standards for how to talk to models, it’s easy enough to switch to another model every so often, as long as you aren’t doing it multiple times per week. The other alternative is that OpenRouter stays in the mix but its pricing gets ground down and down. If it’s (nearly) free and still adds some value, fine. But that doesn’t justify $7B. So, yea, I don’t get it from Stripe’s perspective.
> As things settle down and commoditize, the value of switching on a dime diminishes as people lock into their favorite models
I can imagine just the opposite outcome from the same scenario: as people settle into their favorite but commoditized models, competition for marginal inference cost will take over. A company like OpenRouter that promises the cheapest tokens by the minute becomes essential on the low-cost margin.
I think that OpenRouter and equivalents get pushed out of the market only if the froth calms down (as you posit) and winning models stay proprietary, perhaps with their own unique API surfaces.
> A company like OpenRouter that promises the cheapest tokens by the minute becomes essential on the low-cost margin.
Okay, I can see that, but if their value is just cost optimization, their ability to mark up the tokens becomes increasingly difficult as well. Or, people will build a router themselves to avoid paying the markup, possibly with reduced features, but someone will open source it. Heck, Claude or GPT can probably one-shot it these days. Either way, I think the whole OpenRouter model is going to struggle unless the market stays frothy.
First, cost optimization directly contributes to how much markup they can charge. Say on average they deliver savings of 20%, that is excess value OR/Stripe can mostly capture.
Also, you don't need to mark up tokens much if you're a commodity volume business. Think of Costco and their margins & membership fees. Not everything has to be high margin, not everything has to be a SaaS subscription.
Read Ben Thompson to understand aggregation theory. Many of the largest internet companies, like Google and Amazon, are simply aggregators. OpenRouter is an aggregator of AI tooling. Stripe itself was just a convenience layer on top of merchant gateways (another aggregator). OpenRouter is so big they can negotiate special contracts with OpenAI for special rates.
I don't think they are.
Just because some Higher Ups in Stripe bought this, doesn't mean its that valuable.
It can easily be, that Stripe can just afford it and think that they are part of the big boys and thats just what companies cost today.
It could be that Stripe really really like the brand name and all the positive notion of it.
It could be that Stripe doesn't want to build this from scratch in a timefrime of 1 year or 2 because Stripe might be too corporate to be able to do startup stuff.
Stripe has quite a high motivation to leverage agents thoguh because they are preparing for Agents which will buy through stripe. They already provide the SKU backend and support the agent payment stuff.
They're selling shovels during a gold rush with a 5% markup
It's a simple marketplace and they control a ton of the demand side who trust them to find the best price and providers that work. The value is huge to users - one place, multiple models, providers bid for the traffic and the user doesn't have to think about anything. The value they hold hostage against providers is gargantuan - nice model there it'd be a real shame if none of our users used it.
I don't see how stripe adds any value here (and I've had such terrible experience with stripe automatically breaking my stuff I am worried I can't trust openrouter now) but I can see stripe wanting to be in the middle of any two people giving each other money on the internet and this is squarely (lol) that.
It's Netflix before the unbundling happened
Ha that’s a really interesting comparison! I guess the difference is that OR directly sends revenue to model providers, so maybe they’re more likely to continue working together? But I can totally see it go the other way once a provider feels confident enough their users won’t switch away. At that point the OpenRouter tax, however small, will be a problem to solve.
You can save money and get higher uptime. Without it using a cheap provider for open models would be risky because they might go down a lot, but routers can detect that and instantly switch to a different provider. They basically take away all your exposure.
The other thing is convenience and centralized security from using one gateway to access everything. It's a lot better than having to deal with N accounts with separate limits and monitoring. And giving your payment details to one company instead of 20 is obviously safer.
Investors like them because the pricing is inherently usage based so there's zero risk of clients using more tokens than what they paid for. Guaranteed profit as long as they can keep a modest amount of customers.
On acquisition pricing: Why smarter AI models could drive up compute prices 10x (at least temporarily). [0] This seems mind blowing, but the big boys seem to be behaving as if it's directionally true.
If compute is constrained and expensive, OpenRouter is what you'll use to get around the constraints at individual providers.
[0] https://x.com/dwarkesh_sp/status/2084333160075055122?s=20
The art of good business is being a good middleman
OpenRouter allows you to have 1 API key to access hundreds of models across dozens of providers.
Can’t explain the valuation since everything in this space is rationally overvalued, but I don’t think OpenRouters valuation is that surprising, all things considered.
I use it because it makes A/B testing different models really easy
Contend of the LLM calls.
Distribution is a big moat and value-add.
So much so that right now GPT-5.6 Sol tokens are half price from OpenRouter compared to OpenAI directly. Which I guess means enough people are using OpenRouter, that OpenAI are more concerned about getting those users to switch (from, presumably, Anthropic) than they are about encouraging OpenRouter to exist and long term reduce margins for them by competition / lack of lock-in.
If you stick "AI" anywhere in your product pitch you get to stick a few extra zeroes on your valuation that has been pulled from your ass which you then sell to feckless VCs looking to swallow the world before the bubble pops
PE is desperately competing other PE to get into the promise of some AI thingy NOW?
Just guessing. The frenzy around popular, good already, and successful services with the corporate crap flowing from this acquisition announcement too is appalling. The "what's best for you, the user" heavy emphasis when this would be inherently evident in any honest service forecasts the opposite.
Some highlights from one of my agents asked for a no bullshit evaluation:
"By buying OpenRouter, they own the routing layer that decides which model gets called and at what price."
"Stripe wants to be the economic infrastructure for AI — meaning they want to be the toll booth and the traffic cop for the entire AI economy."
"... insider market intelligence that OpenAI, Anthropic, and Google don't fully have. Stripe would now have it."
"The $7B+ price is absurd on any financial metric, but this isn't a financial acquisition — it's a strategic positioning play to own the platform layer of the next computing paradigm."
"140x revenue multiple: At ~$50M revenue, this is not a financial valuation. It's a strategic land grab — buying the chokepoint before someone else does."
I do not feel a particularly strong smell of 'best for the user' here for some reason... More like the usual 'how do we squeeze out more for our PE folks from this' kind of scent.
Are you seriously posting an LLM generated “analysis” of financial event.
They get all your prompts, the LLMs companies only some
Tokens are just a type of currency. Stripe is a middleman for currency.
Using something like OpenRouter (or any of the AI Gateways) is better than tying yourself to one LLM provider that can rug pull on pricing or change models in a way you don't like.
The value is in the network effect I think. OpenRouter is popular and has a good head start over anything Stripe could build internally.
Tokens are not a currency, tokens aren’t fungible and cannot be traded
to a certain extent, openrouter did manage to turn tokens into a currency. The fungibility is there, as much as you don't care so much if you're getting your token from novita or some other player, these become fungible.
They aren’t trading tokens, they are a market for compute. The difference matters quite a lot. What you have in openrouter is the ability to exchange money for compute at the vendor of your choice.
We are basically back to 2020 trying to get people to understand that an NFT is NOT the underlying asset it abstractly represents…
> The difference matters quite a lot.
Please go on as I do not get your point.
You spend $100 to get tokens from Claude. Then go to your friend and tell them you are willing to sell the tokens you got for $50, it’s a bargain! They look at you, tell you that makes absolutely no sense, and why the fuck would they want to buy your tokens, ie something that only makes sense in your very specific prompting context.
Tokens don’t have an inherent value and have no other properties that would make they tradeable.
Another way to think about it: a company that spends $1M on tokens is burning that money hoping they will be able to recoup by generating a better product. If it would be tradeable, they could spend $1M to generate tokens, produce absolutely no product, and just resell all those tokens to get their money back. That’s obviously nonsense, that’s not at all what tokens are. And they cannot transform the tokens they got into compute, you need to consume hardware and energy to mint tokens, you cannot convert tokens into hardware
Isn’t this what Venice.AI is essentially trying to do?
A dollar bill and a euro coin isn’t fungible either.
? In general, it is, unless it has some collector value, for example.
Yet
Tokens are minted then consumed. You cannot trade them, by definition. What would it even mean to trade a token? You can maybe trade a voucher that allows you to then redeem it for some token generation (aka compute), but the tokens themselves don’t make sense to be tradeable. And because they aren’t fungible you cannot swap a token for another one, that’s just not what LLMs are
I think you are reading too much into the "fungible" term.
No one is talking about trading tokens.
The idea being that getting your tokens from Provider A is no different than Provider B, especially if they both offer the same model. You can change one value in a request to openrouter and suddenly be hitting a different provider but offering the same tokens, because they offer the same model and the same settings.
> The idea being that getting your tokens from Provider A is no different than Provider B, especially if they both offer the same model.
You're describing a commodity. However, the commodity isn't the tokens, but the compute capacity, i.e., serving a model. And compute capacity isn't a currency — at least not, until you can acquire compute capacity from one party and exchange it with another party.
dgellow is absolutely right: tokens aren't a currency, nor can you trade them, nor are they fungible. The original claim that "tokens are just a type of currency" [1] makes no sense.
[1] https://news.ycombinator.com/item?id=49364984
Yeah, I mean, it looks like a network effect, but anybody can aggregate the models; rather, it is the switching costs in the logs and work, and the cost savings, etc.
This has to be all about cashflow, right?
Surely stripe if anyone have learned to harness the cash flowing through their system. Hell, they could be emitting bonds on expected token consumption bills!
Alex's first post about OpenRouter got 6 upvotes and 0 comments on HN https://news.ycombinator.com/item?id=35481760
same with cursor - which was submitted multiple times:
https://news.ycombinator.com/item?id=34423387
are there more examples of this? a bit of an HN anti-portfolio. good reminder to others.
There's the infamous post dismissing DropBox ( https://news.ycombinator.com/item?id=9224 )
This is, somewhat ironically, the most well-known HN comment outside of HN.
It's also very representative of the HN comment section vibes overall.
But such is normal with normies providing startup critiques, if you just look at every announcement and say "this will fail", you'd be right 99% of the time.
I mean, the AI skepticism is still so prevalent today in HN, despite it solving decades old math problems, hacking into companies, making software engineer no longer code, etc. (all this just in the last year).
So nothing really has changed in this community's vibes.
If you think HN is anti-AI, you are in deep. There isn't many places on the internet (sans X/Twit which i don't check) that is as AI-positive as HN. Sure, holdouts remain but the frontpage has a 60% constant vibecoded project submission.
You are mixing supporting AI (which HN is mostly against, but more supportive than other places), and thinking AI is dumb, which HN has been pretty persistent on for years.
I mean - just looked at this downvoted comment essentially saying the same thing - https://news.ycombinator.com/item?id=49119464. And also the intensity of the replies.
Thanks for your input, GPT-5.
> It does not seem very "viral"
love that Dropbox has became textbook example of viral marketing
Poor BrandonM... He has not been invited to cocktail parties at all since then.
What does window.ai have to do with OpenRouter? I don't use either, so your link confuses me.
It's the predecessor. See https://news.ycombinator.com/item?id=38415092
I kinda expected it's filled with comments telling him how stupid the idea is..
That’s the difference between a priori and a posteriori knowledge.
that subject line is pretty bad though.
(i know much of it is a crap shoot and easy to quarterback on Monday morning.)
What a good find! Makes me feel better about my projects that get zero traction here.
getting zero traction here is clearly the path to success :)
I've not used Open Router but I'm wondering what are the reasons to route LLM queries at large scale?
Isn't the trend increasingly to let an orshestrator agent decide which model to use in terms of problem difficulty vs. size and thinking budget?
Once you fix those parameters there's little to optimize over, no? I mean you want to find the cheapest one across providers but that's not a recurring decision.
I suspect only a very small subset users are sophisticated/interested enough to do that for themselves on a regular basis. More will likely hop between agents based on cost and the perceived nature of the task, without hard data (or an orchestrator model) to support that decision.
In the second case OpenRouter is very useful; in the first case, maybe it’s an area they move into in time - providing and owning/optimising the orchestrator?
> reasons to route LLM queries at large scale?
Convenience. Paying only one provider, but get access to different models hosted around the world. Need a model for a project where latency matters? EU hosted and ZDR for another customer? Cheapest for personal use?
You get everything under one roof.
One thing that always worries me with these acquisitions: how are they gonna handle the impending 5.6M+ Stripe Connect users who rely on OpenRouter's APIs? Will they integrate it into their core infrastructure or leave it as a separate product?
I rather have protocols be built and less middlemen PaaS.
One great example is Open Banking. I would support an Open Router that mirrors Open Banking, unfortunately the Open here doesn’t carry the same connotation.
Kudos to the team for succeeding in executing around a clear DX issue during the chaos of providers, though long term I am not really a fan of it staying as is even if Stripe gets to be the one carrying it forward.
What open banking? I can't even get a read-only token for my personal use to see the balance on my account and I sit two team calls away from people who cooked the API
Where it is already operational and available elsewhere worldwide. Unfortunately complex lobbying and Wall Street not being a fan of it as they’re the biggest loser makes it slow to get it done in the US.
In the EU you also cannot just get a token to open banking.
Deleted. I don’t like starting rage baits… ;)
With vibe coding, you can make a bot that pretends to be human accessing the banks website, and be able to hit the bot with an API, and have the bot interact with the bank for you.
However, you’d be liable for the losses due to fraud in this scenario
Every single Open Banking implementation I know also requires everyone to go through a useless certified middleman (eg. Plaid, GoCardless, ...) instead of letting people use the APIs directly
Brazil did PIX which in that case the middleman is the government and everyone got to benefit from
Isn't PIX a payment scheme?
Open banking is slightly different, as it's more about account access (although with read+write access, you can kind of build it into a payment scheme as well, although it's usually clunky).
You're right, they are actually one step ahead and into what is now called Open Finance and aims at extending Open Banking beyond the original vision.
The brazilian central bank has a page about it[1]
[1] https://www.bcb.gov.br/estabilidadefinanceira/openfinance
would encourage you to read up about how the rest of the world has solved this
When you say rest of the world, what countries exactly do you mean? I'm from the Czech Republic and the process is also insurmountable for a regular company. You either go through a middleman or you can apply for a license (only 4 companies in the entire country ever got it and it costs upwards of $50k in legal fees just to file the application).
Why don't you just tell us, rather than give us homework?
Can you just link us to the relevant pieces? I rather not search on my own and read wall street propaganda on why it's bad (which is what I'm finding and not interested in).
Here's a McKinsey report on Pix, Open Banking, and Open Finance in Brazil[1]
[1] https://www.mckinsey.com/br/~/media/mckinsey/locations/south...
One problem here with the protocols is the field isn't yet so stable. Most people are copying OpenAI's endpoint, and so you have "OpenAI compatible", but there are always little hairs on it. Case in point, OpenAI itself no longer offers multiple basic inference options on their new models. Can't pass temp, get back logits, etc. on anything after GPT 4.1. It's not a total zoo and Open Router is pretty easy but everything is just enough a special snowflake that models aren't a commodity. Maybe later though.
Agreed, though that's not a reason to discourage building it or putting it forward. That's more the reason why a protocol would be beneficial, we are ending up with price wars and middlemens everywhere because we're racing.
Open banking protocols exist and are widely used but they’re called crypto, which everyone on HN hates for some reason
Feels the same - businesses from pizza shops to eCommerce sites will be arguing with not their credit card provider about transaction fees, but their AI provider (who supports all 3 major networks, including Discovai, Clankers Club Aierican Express
Stripe acquiring the biggest model gateway makes sense when you think of inference as the next payments. Both are developer API infrastructure where routing and provider abstraction matter.
"to power the next wave of GDP growth" is a wonderfully shallow but impressive clause.
It is amazing how we watch from the sidelines these aquisitions and think .... why such big bill ? But it takes a good team to sell and everytime someone is successful it is because probably they have great sellers who believe in their value and demand a price. It is certainly not an easy task to get to the finish line for a startup and pay back their investors.
Curious on what OpenRouter's true moat is? It's just an LLM API routing framework, right?
Yeah, and Stripe is "just" a payments platform.
Openrouter is easy, reliable and performant (like Stripe).
> Stripe is "just" a payments platform
Counterpoint: you can't just go to Visa/Mastercard or a merchant acquirer out there and set up an account on the same terms that Stripe can.
On the other hand, you can sign up to any LLM provider and get API access on terms that are the same or better (since I'm sure they don't appreciate having a middleman and would benefit from incentivizing direct usage) than OpenRouter gets.
Openrouter is fragile, one good competitor and they are at risk.
Or if one of the provider suddenly decide to forbid openrouter from using their api. Why would they do that tho ?? Well it's not like execs never take dumb decisions.
In my humble opinion, it's extremely overpriced. But then, it's just the standard with AI currently. Divide every ai company valuation by 1000 and you might get it's real value.
Yes, you can get better pricing if you do it yourself. But the true advantage of OpenRouter is that, in a space where there's a new model being released every week, you can easily switch to whatever model is best at any given time without having to set up accounts with multiple providers. Or you can just experiment with the latest release. Their product is the convenience. Of course if you decide to only use a specific provider or two, then you don't need OpenRouter.
But it's not really that difficult to make a clone of OpenRouter's service. What they do isn't really that original.
Their only value comes from the fact that the currently have lots of traffic. And I dkn't think that their cumstomers are really bound to theur servuce. They could switch to a competitor without too much hassle.
Yes, but as l9ng as there's no clone, they're good. On the potential clone side, I guess many will be put off by the fact that "there's already OpenRouter". Also, a clone would need to find a way to make existing OpenRouter customers to switch, which isn't easy.
Will that necessarily be true in the future? Would we not expect that openrouter will develop the ability to negotiate special pricing?
To add to this; payment infrastructure requires a lot of heavy lifting. There's a lot of regulations you need to adhere to, different payment systems in different countries, settlement, chargebacks, etc.
There is a reason why not doing your own payment processing is a thing.
OpenRouter may have some interesting things in streamlining the process of switching LLM providers, but it is indeed something easy to replicate in comparison to payment processing.
Also it's name has "open" in it. Which means you can do what you want with it, and it's all open source!
Just the other day, I tried to have openrouter route to my local ollama, and it refused. Apparently I have to pay for this privilege. How wonderful.
Open open (op)en. Open open (op)en. Clo-o-o-ooooo-sed.. (sing with Wagner flight of the V)
Now that's the value. OpenRouter has the power to step on the air hose of any provider they don't like.
That's Google's real power. Works for them. Between search, ads, and the "app store", they can crush most companies. OpenRouter's power is only in one area. For now.
Payments is quite hard, actually.
Payment platforms existed for years before Stripe came along. They solved the problems the others couldn't.
Having worked on web apps that processed online payments before and after Stripe I totally agree, this was an area of real pain that became suddenly extremely simple because of Stripe. The alternatives were terrible–100 page Word docs of SOAP API docs for Authorize.net, massive PCI compliance requirement specifications, horrible legacy merchant services businesses.
On the other hand, I operate an app that talks to (and logs prompts/meters costs) to many different LLM API providers, and I do not consider it painful at all. I have an AI agent to deal with any integration quirks, if needed. Mostly they provide OpenAI-compatible APIs anyhow. It's basically a no-brainer to go direct with the providers and save 5%, the great majority of the cost of an AI-powered app is no longer dev time implementing the integration, it's the tokens themselves.
There are already other providers that can say the same, some with very small teams and better rates.
Yeah and the kids down the street are “just” a lemonade stand.
Stripe operates in a very regulated industry where new entrance is difficult
OpenRouter is commodity stuff, I've never used it personally and picked alternatives. The sense I got from people who said they used it was that they are on average penny pinchers. That does not seem like an ideal user base
OpenRouter is already the entrenched go-to router tbh. There's competitors (Vercel) but they're already well known. Stripe can also make them more competitive because they can waive / profit off the transaction cost. Considering stripe takes like a 3% fee, and the openrouter markup is 5%, that's a pretty big boost.
Now is that enough revenue to earn back 7B.... I'm not too sure. I guess they're betting realy big that these sorts of model routers will gain explosive token usage in the future
I don't get the 'transaction fee reduction' arguments. Yeah they can reduce the fee, which in turn reduces profit margin... any other product/service in a similar position can do the same... it's not some sort of free win
This was the same thinking behind Vanguard funds.
Not the same at all.
And trustedrouter which is a bit cheaper and more private
Doesn't TrustedRouter cost more than OpenRouter? (5.5% markup vs 5%)
Also, TrustedRouter's website is full of slop, which does not inspire much confidence.
I presume they are sitting on a lot of usage data and will be able to roll out efficient auto-routing based on it. There is a huge demand for this as most users are overwhelmed by provider options and just want to route to the best/cheapest model based on the task.
I think it's kind of like what GitHub has despite getting rid of the quota system for Copilot: existing customers that are already setup to deal with them.
In this case I imagine that OpenRouter's moat is going to be that businesses that already use Stripe will have a lower bar to choosing OpenRouter to provide the AI access, and clean integration with Stripe to pay for stuff like AI support bots.
Isn't copilot a cheaper harness than claude code and offers more flexibility?
I don't know about cheaper.
The flexibility is kind of what I'm pointing to, if I understand you right. Instead of dealing with Anthropic, OpenAI, XAi etc directly and having systems to handle user assignment, budgeting etc for each, companies just deal with GitHub, which they've probably been dealing with anyway.
They have enough customers to have leverage against providers. As an example, if they decide they can be "punished" if they claim ZDR but lie about it.
As with Cursor it's the data - all the prompts and responses that can be used to train AI models. Getting a head start on that at the pace of current development costs.
I dont know exactly but dont you think 90% of Vertical AI Companies are doing same?
They are the leading LLM API routing tool. Stripe is the leading payments API. They are betting on OpenRouter being as important to developers as payments. Both let them take a small percentage of a shitload of transactions. Being the goto service for APIs and executing well is the moat. Someone else would have to beat the volume leader on price or beat them on execution. Good luck. The only realistic competitors would be Amazon, Google or Microsoft and none have shown any interest so far.
I don't understand why one would let a router handle their LLM request (other than for routing it to the cheapest inference provider for open models).
When I build an application that uses LLMs, it's tuned and tested for a specific model only. LLMs are not really drop-in replacements of each other, even different versions of the same series.
The pertinent question isn't why OpenRouter but why a router company is worth 7B.
You can use one key for multiple models, you can have many keys with a budget on each, you can manage all your keys in one piece of glass, you don't have to deal with the upstream vendors directly just to name a few. These are all things that may be important to developers. I can't even get an OpenAI API key because those morons can't get their phone verification to work properly. OpenRouter is the only way I can use their models.
OpenRouter’s “moat” is currently “why would I use anything else?”.
Stripe has presumably only purchased them because they think there’s consumer-surplus to monetize here, which presumably will soon be giving me lots of reasons to use something else.
It isn't clear to me what advantages OpenRouter has over running a similar router-aggregator locally for ~free. I guess they have stronger bargaining power than a solo engineer or small startup, and can therefore negotiate volume discounts with model providers. The question is if those discounts let them offer cheaper rates than a local router-aggregator even after the markup. And if they can do that, then OpenRouter is undercutting the providers themselves on price, which I would imagine the providers would not want to see.
Because the others are cheaper
The others have to charge a payment processing fee and take some profits.
Which other aggregators that have anywhere near the model selection are cheaper?
OpenCode Go is too good to be true, and I suspect it won't last, like GitHub Copilot.
Minor nitpick: don’t confuse sales with negotiations.
Negotiations start when sales is done.
These were highly capable negotiators.
But most likely also good salesmen.
If you meant that generally: because productivity is increasing; the world is getting richer. Not all of it flows down, but some does; top SV salaries are up too. By a lot.
edit: Inflation does not account for the observed growth.
We are certainly seeing higher monetary velocity in certain sectors. That usually follows from real economic expansion but could also be due in part to inflation.
Corollary to this is that whenever companies acquire smaller companies and lay off the small company's sales teams, they deserve what they get. Looking at you, big company that turned my last one from nearly a billion valuation into a $20M fire sale in 3-4 years.
It's just the same VCs moving money from one pocket to the other.
I've seen three or four others building routers. How hard is that really? The folk I spoke to needed only a few millions to build it (they claim). I've seen prototype of some too that looked pretty simple. Maybe the whole thing is driven by enterprise wanting the external vendor support?
The software is "easy", it's the business and operations that are hard: scaling the infrastructure reliably; finding revenue sources with secure margins; deflecting regulatory and risks responsibility; billing and collecting on massive low-margin high-volume transactions; enterprise sales; etc
It's a business line more suited to finance, law, sales, and accounting people than tech people and a pretty laborious one. That's often the case when the tech looks "easy" but the sector only seems to have a few big winners.
It's hard.
- They support 400 different models, 80 different providers, and an unlimited number of new custom agents. Every single model, provider, and agent, has its own weirdness that has to be accounted for. Tool calls change by model. Effort changes by model. Backend APIs (messages, responses, etc) change by provider. There are thousands of specific tweaks, fixes, hacks, that need to be implemented to make this thing "just work". And you have to keep updating it all, weekly.
- On top of that, they support providers running in multiple countries, which increases the legal, logistical, financial, and networking complexity.
- On top of that, they need to figure out pricing and deals with providers, and ensure the providers allocate the necessary networking and compute capacity. They need to work with providers directly to troubleshoot and fix all kinds of issues, from networking to application.
- On top of that, they implement multiple kinds of request routers. An auto router to route your request to the best general model, a fusion model to attempt the request on multiple models and pick the best response, a pareto router to route requests to the best coding model for your request.
- On top of that, they build custom features that businesses and users want, like Data Loss Prevention (which I'm blown away they actually provide for free). Lots of very useful business functionality for managing not only what model and provider to use, but also limits on usage, filters, etc. They also implement SSO, prompt injection guardrails, logging/auditing, workspaces, etc.
- On top of that, you have to be very good at just implementing HTTP APIs. Most people aren't good at it. API design is hard, HTTP is way more complicated than it seems, network traffic shaping is a black art.
- On top of that, running applications and infrastructure, and scaling it 10x every year, is a subtle yet critical skill. No matter how good your code is, it's pointless without working hardware/running apps.
- On top of that, they provide a generous free tier, which has to be subsidized, and is probably only partially subsidized by providers.
Why pay for this rather than build one? Because there's no sense in building one. There's a reason restaurants don't build their own factories to manufacture their own pots and pans. Their business isn't selling pots and pans, it's selling people cooked steak. It's a bad idea to waste time, effort, and money trying to make things you aren't going to sell.
I did almost all of that alone, for TrustedRouter.com
Vibe-coded content & design aside, you haven't implemented all their features, you've got some basic security holes and missing networking, and your pricing is the same as OpenRouter despite your company being 5 months old. It might not be as easy as it seems.
My surface-level assessment is that building a router is easy, but routing efficiently is difficult. For the latter problem, traffic provides signals which you can use to improve routing rules, creating a feedback loop. More traffic => more feedback => better routing. So a lot of the value of Openrouter would be that they are the most adopted router (I think?) and so likely have the most data to (potentially) produce the best router. I'm only speculating.
It's a fairly high scale, but yeah I don't think it's terribly hard. Negotiating rates with the providers seems like it'd be the hardest part to me.
Probably they want to use the userbase, not the technology? There was a heavy emphasis - too heavy for my taste so it bounces me back - about that they will do what is the best for the user. I think that was already there, Stripe was not needed for that. This is more for the benefit of others.
On acquisition pricing: Why smarter AI models could drive up compute prices 10x (at least temporarily).
This seems mind blowing, but the big boys seem to be behaving as if it's directionally true.
If compute is constrained and expensive, OpenRouter is what you'll use to get around the constraints at individual providers.
https://x.com/dwarkesh_sp/status/2084333160075055122?s=20
It seems like a misunderstanding of basic economics to assume that the price for this would track FTE salaries.
> Today, we are excited to announce that we are joining forces with Stripe, to power the next wave of GDP growth globally.
This is some high level meaningless corpo speak.
I get where it's coming from, I like Stripe's long-stated mission ("To increase the GDP of the Internet").
This version is dumb/empty, and reads like brown-nosing Stripe.
> I like Stripe's long-stated mission ("To increase the GDP of the Internet").
Also corporate speak. Their mission is take their cut of the Internet's GDP. Nothing against them, they provide real value for that cut. But that's what the actual mission is.
That feels unfair to read it that way. The mission as stated is that they will not only take a cut of the pie but grow the pie. If the total value they add is greater than the value they capture, everyone wins.
You might not believe them, but that’s what the words are supposed to mean.
> You might not believe them, but that’s what the words are supposed to mean.
It's corporate bullshit not because I don't believe they can grow the pie, but because that's not their motivation.
In fact, a nice formula for coming up with your corporate mission statement would be:
1. Ignore your company's ground-truth financial incentives
2. Find a positive 2nd order effect E (or a plausible one) that you don't really care about, but maybe you genuinely think is nice
3. Claim your "passion for E" is what drove you to start your company, and keeps you working, because, of course, all great companies must have a "deeper purpose"
It's not mutually exclusive. I remember the Bad Old Days of trying to make the PayPal button work; if Stripe makes it easier for people to take payment online then it increases the GDP of the Internet.
Well given the commissions of Stripe's, is mostly about its own GDP.
I'm curious about the GDP vs. GNP of the internet :)
thats what billions does
One unexpected "benefit" of LLMs is that it is training us all to be even better at recognizing "you don't even know what you mean by that, do you?"
https://www.axios.com/2026/08/19/stripe-payments-openrouter-...
https://stripe.com/annual-updates/2025
> Stripe last said businesses on its platform generated $1.9 trillion in payment volume in 2025, up 34% year over year.
More transaction volume = more enterprise value potential. More GDP, total rake goes up of said economic activity they facilitate. Middleman Moat, accumulating volume (present via Paypal, future via OpenRouter).
Big line go up!
It's particularly stupid given that if you look at where the tokens are being spent, the top application overwhelmingly is Hermes Agent. It's more than the next six clients combined.
Hermes Agent is AI slop "ai assistant" software that is being entirely developed by a bunch of AI agents. They are stuck in a constant whack-a-mole bug fix session because the agents keep breaking things.
Openrouter is being entirely propped up by garbage-tier software that is only popular because it's being endlessly hyped by idiot youtuber AI Agent Bros who are so uncreative and incapable of critical thought, they're using AI agents to tell them what content to make.
When people finally realize how bad Hermes is, OpenRouter is pretty fucked. And then when all the weebs writing furry porn get bored and move on, they'll be fucked even more.
What's really bad about this: Stripe is so tightly tied to the credit card industry that said industry's obsession with morality policing will force Stripe to start censoring openrouter.
Probably a mixed bag at OpenRouter today. Some employees probably just became very wealthy. Others will see what it's like being owned by a different company with different goals. Layoffs tend to accompany most acquisitions within 6-12 months these days. A $7B price tag probably comes with some hefty expectations.
I think OpenRouter has a pretty lean staff
the toll booth on the AI highway just got acquired by the toll booth on the payments highway. at this point i'm just waiting for the toll booth on the toll booth highway.
Hopefully it will be good. I have started using openrouter.
I've seen a lot of comments here and elsewhere about why is a proxy worth $8b. It is a bit more than a proxy.
It is a known brand for model routing and it has developed means to route model on their own and 3rd-party infrastructure at scale and without much issues.
Why Stripe? Who knows. I don't see much synergy but why not? If anything it is a way for Stripe to get into the market with agentic payments which may come to be worth a lot more than $7B.
Raw Cookie header test - OpenRouter + Stripe API limits?
"inference is quickly becoming the largest line item for every company." — not really?
"that will remain essential in a post-AGI economy." — ooof. The magical future economy
This feels defensive and reactionary after Ramp started getting deeper into AI.
and pretty sure they're gonna drop crypto currencies as payment on the platform too
I doubt it. One of the founders was on a podcast recently and they were singing stablecoin praises.
Why? How do these relate?
Metronome has been experimenting with real-time payment on usage based billing for tokens. That technology leverages micro payments using crypto. It's cool but all it's doing is moving the accumulated balance from a provider onto your crypto wallet.
Routing the AI traffic turned out to be a better business when you also own the toll booth.
Congrats to the OpenRouter team! Been using it since the early days — the ability to try every new model without creating 20 different accounts is still the killer feature. Hope Stripe gives you the resources to keep building.
Congrats, OpenRouter is a great product!
I check it's model rankings and prices daily. The charts are great. Please never change!
On another note... I'm surprised OpenAI didn't buy them.
I'm even more surprised OpenAI is giving exclusive discounts on openrouter, essentially encouraging the few API customers they had onto a much larger, cheaper and richer marketplace.
Cynically, I suspect oAI is subsidizing OR tokens to manipulate perceived traction, since OR usage stats get widely publicized
What exactly does this mean? You can verify the pricing differences from their official API vs. OpenRouter.
If you have some grand conspiracy, you should be able to validate it. Yes, on occasion, OpenAI has offered Discounts on their API.
> please never change I've got bad news for you...
The service has been 10/10, everything has felt easy and natural since the start and continually improving. I often learn about new models from OR, and have been a happy customer for years. Kudos for getting the proper recognition, $8B is a fair price in 2026 imo.
> And as we grow, we will relentlessly aim to preserve the velocity, agility, efficiency, and talent density of the 90-person startup that we are today.
Echoes of WhatsApp. Huge congrats to the team!
Stripe is a high-velocity environment. OpenRouter (OR) is a startup so might move fast, but unsure if OR can handle the velocity with corporate guardrails put in place already.
Any companies that are bought go through culture crashes, and I am interested in how this plays out.
In what way is Stripe a 'high-velocity environment' ?
I was saying, they grill you hard in a nicest possible way
They grill their engineers for the mistakes of their incompetent leadership, that's true.
man, whatsapp was sure something in early 2017 though.
Routing seems to face a similar future as the ever-changing trends in controlling coding agents: you know, context engineering, loop engineering, etc. --Which is that OpenAI and Anthropic quickly seem to release a better way of doing it, in their own products' harnessing, and which third-party and personal solutions will only conflict with.
I guess this makes sense for the scenario when we will use machine intelligence as a currency, maybe 15-20 years from now? - like I can pay 15 minutes of inference for a dozen bananas? and the banana seller uses those 15 minutes to do banana shelling or removing weeds from their farms or whatever i guess when robots are doing everything. right??
This won't work because you consider humans as a logical being (Homosapiens is a misnormer).
People will revolt first should one have to trade physical goods (such as food that we need for daily lives) with computing power.
It only takes a few people to incite.
i like what you did with misnormer. it's like pobody's nerfect.
you are right, there will be revolt definitely. i think computing power will replace a lot of things though.
I agree that the computing power will replace not just softwares and also as many phycial work via robots. I am not smart enough to see further what else are disposable or replaceable
I think this is exactly the bet they're making.
I was totally confused by what connected the two companies until I realized "tokens" are basically a nascent, and rapidly growing, "currency".
Overthinking it. "You have money and need to seem AIish for the stock market, we want real money". You notice they don't do this deal in tokens, they do it in fiat currency.
Except that tokens aren’t completely fungible yet. But, I think we’re approaching the point where it is close enough to be a “proof of work” analog.
I hope we aren't on track for this future and if we are I hope a substantial number of people rebel.
This is among the more hellish futures I can imagine. Our worth as people reduced to how much access we have to compute. Gross.
I hope not;
there are a few assumptions in this future- the biggest one is - we will have to attain a baseline intelligence which is useful for everyone which is not true today. even fable fumbles hard!!
second is that everyone will have to own a medium to use this intelligence (aka robots). like a currency is useless if you can't use it.
to me this is the least daunting scenario. all other scenarios lead to mass slavery.
> the biggest one is - we will have to attain a baseline intelligence which is useful for everyone which is not true today
That's not the biggest assumption.
The biggest assumption I see is that the population will have the baseline understanding of how to operate in this new paradigm, which seems impossible given how useless many people still are with computers
Incidentally, this is also why I think Bitcoin is such a non-starter for most people
It doesn't make much sense as compute isn't all that scarce (relative to other commodities) and will be less so in 15 years.
Does it have to be scarce for it to become a currency? It just needs to be regulated which I think it will be.
So they essentially resell their enterprise tier access to tokens from each provider while also providing a nice interface/dashboard for their customers?
I like OpenRouter. Excited to see where their incredible journey takes them.
Did you meant this for LinkedIn?
I was actually thinking of [1] but hopefully it doesn't turn out like that.
[1] https://www.tumblr.com/ourincrediblejourney/89180616013/what...
How in god's name is openrouter worth $7,000,000,000? Can somebody explain to me, or is everything to do with AI just astronomically overvalued these days.
Makes total sense. Stripe acquiring OpenRouter is a massive DX win—unified billing for multi-model APIs will save devs huge routing and token-handling headaches.
I am glad it’s so easy to code. All these tools keep getting bought, cursor then this, and I have been able to build my own replacements.
Hm. This seems random. Other than the fact that they are both technology companies, is there anything else that they have in common?
Stripe is positioning themselves as one of the leading adopters of AI internally. I don't know of public stats but I estimate they're spending on the order of $millions per week on tokens. I assume they've adopted Openrouter and the majority of that spend is going via Openrouter. So they understand the product value, they understand the current gaps, and there's also a threat mitigation. They've likely become very dependent on Openrouter and someone is going to buy it - so they may as well negate the risk to that. (Threat mitigation alone is not worth $7B but it justifies a certain premium.)
They also have very complementary problems and skills. The long-term roadmap for routers is auto-routing; that is, turning traffic into signals into automated decisions. Stripe has a lot of overlapping talent and experience from fraud-detection and likely other products.
Follow-up: Somehow in all that I forgot the most obvious pairing. Stripe wants to solve how agents pay for things. They’ve been quite open about this. Seems obvious that any solution to that is going to be through the model platform layer - so they just bought a major platform layer.
Stripe execs said "we should own some AI"
Founders found out that they could finally afford to pay their medical, tax, housing bills if they said yes
One-to-many implementation. Stripe is one api to every bank. Openrouter is one api to every language model.
And... How do they combine? I really curious.
Their business models are actually kind of similar. Stripe is just passing money between financial institutions and taking a small cut. OpenRouter is basically the same thing. They might have shared ability on how to run an effective pass through business.
So… they’re both middlemen. And that’s enough to justify a $7B acquisition???
Apparently. The price does seem high, but the value of a company is based on it's finances. You could argue companies like Google are just middlemen too.
My read: The thesis is that token spend is going to be one of the larger input costs to a business in the future. They want token economics to happen on their platform so that the commission that comes out of all of this goes to them. It is not enough that openrouter uses stripe exclusively to interact with banks, since the transactions are expected to not just be tokens for dollars, but tokens for tokens or tokens for specific stablecoins tied to tokens. One of the value adds of stripe is their policy and monitoring layer. With tokens, the polices and monitoring will happen in token-land primarily, not in dollars-land. So if you want to continue demanding commission for those types of things, you have to be able to execute policies on tokens. Tokens can of course be variably priced - I am not claiming # tokens itself has the same value across providers.
The platform also has another value that is not mentioned often. People often mention the devex, which is consumer value. But one of the main things about open router is that they help model providers access a massive distribution channel with smooth demand, as with any aggregator. Similar to the value food delivery apps provide to restaurants. As such, this type of marketplace power has the potential to allow them to charge model providers commissions as well for various services in the future. You can already almost imagine an "auto mode" where they balance improvements to cx (say right sized models for tasks, saving u money) with "ad"/preferential treatment promises tomodel providers. Basically, all the typical middleman stuff.
One of the other companies in this space, ramp, has also started their own router at router.com. See https://x.com/tryramp/status/2090146780512227825
Quoting their rumored leaked investor letter dated Aug 19 26 below:
Zooming out, we see capital and intelligence are becoming the two digital flows undergirding every business. Up until now, every developer has needed a straightforward and reliable way to manage their revenue pipeline, and serving this need gave rise to Stripe. Going forward, however, every developer will also need a straightforward and reliable way to manage their intelligence pipeline. This observation first led us to OpenRouter. OpenRouter has built the world's largest and most trusted token routing engine, supporting all major models and providers, and beloved by its customers....We think that there are deeper reasons to pursue integration besides convenience. Our experience in working with our customers has led us to realize that intelligence is special: it is expensive, heterogeneous, and constantly changing. As with financial capital, businesses must reason about cost and return of every unit in a deliberate and granular way. How valuable is this task? With which models can it be best handled? Who will pay, and when, and what is the time-value of that delay? We have seen the parallels between managing intelligence and managing capital directly in our own products. Radar, for example, was initially designed to prevent financial fraud, but is proving extremely effective at guarding against token fraud at many of the world's largest A1 companies. Metronome (used by Anthropic, Nvidia, and other industry leaders) is showing that metered billing in an AI context is inseparable from token serving and consumption itself.
Both are the most popular developer api in their space.
One of the main real values of OpenRouter is that it is a single payment relationship for users that enables access to many downstream vendors. I'm not super convinced it is a good purchase, but right now OpenRouter is the financial middleman for token spend. It is also the centralization point for tokens, allowing for value-add features that are industry-wide, for example budgets - and I think that structure has parallels to Stripe products like Checkout or Identity.
Stripe has a lot of money, but not a lot of AI-buzzword compliance. OpenRouter is highly AI-buzzword compliant, but doesn't have a lot of money. It's a match made in heaven
In the event that the AI bubble doesn't burst, Stripe needs to have a stake in whatever's coming. They have time to explore and pursue specific strategies, but they just needed to make some big and compatible buy.
In this particular case, OpenRouter represents a M x N bridging and enhancement layer much like Stripe themselves, and so must address a lot of parallel technical, dealmaking, accounting, and legal challenges.
It may not be obvious on the surface because they seem to be working in such different domain, but they have to address a lot of the same problems in comparable ways and that makes it a pretty darn good fit.
Taking a % cut on top of every operation
greed
Oh joy! Stripe is getting to be an even larger beast! Happy day!
That's either sarcasm as you are not a fan of Stripe, or actual joy if you are an employee of Stripe holding stock.
Pardon my ignorance but...
why is OpenRouter worth 7 billion dollars?
Well Stripe is owned by Thiel, I'm sure there is some nefarious reason he wants to see everyone's chat history.
I am somewhat mixed on this, I feel though having it be part of Stripe will make it easier to sell to my leadership as "More Trustworthy" than having it by itself
What's with naming proprietary stuff "open"?
Clearly these products are being named by topologists, who have no objection to things being simultaneously open and closed.
Schrödinger's Product
It's following the open closed principle.
Its sad that you gotta be a software engineer, to immediately get this joke, because its pretty good
It could be open in the literally sense of you aren't locked in, you have free access to any model you want.
Just because a name has open in it doesn't mean it has to be open source.
It’s an AI era buzzword. It’s not a new process, e.g. when blockchain was trendy, “crypto-“ and “bit-“ became similarly popular and meaningless.
"Open" is a load bearing reference we all get :p
blame openai they started it
Would "OpenSourceRouter" be a trademark violation?
I've built one in Rust and want to market it.
Not if you’re in a different domain, then it cannot be claimed to confuse customers.
So if it’s a network router or a woodworking router, then you’re good.
I can sells of red socks, and white socks, without getting sued by Chicago or Boston.
If it’s in the same space… you might want to contact a lawyer. I believe Firefox had conflicts with earlier names like Phoenix and FireBird and they weren’t even browsers.
If you're looking for a privacy protecting version of OpenRouter, my friend Joseph Perla made trustedrouter.com.
It's a pretty slick system and if you're building on openrouter today and are worried about stripe integration messing up the product, this is a good option to consider.
Separately, congrats to the OpenRouter team. They are selling at the right time in what is undoubtably going to become a contentious market.
7 million would sound fair to me. 7 billion? Explain yourself.
OpenRouter already takes in around $140M in yearly revenue. How would paying 5% of yearly revenue make any sense for an acquisition??
> How would paying 5% of yearly revenue make any sense for an acquisition??
50x revenue is also a crazy number. I wonder what happens more, companies selling for ≤5% revenue or companies selling for ≥50x revenue.
10% would be more typical. Perhaps the math is that OpenRouter inside Stripe makes it twice as valuable. You often see acquisitions priced on the value of the company post-acquisition.
The 7 billion figure is not just open router's current value. It's also including part of it's projected value in the future. So that 7 billion includes current value + value say 5-10 years out.
Remember when Google acquired YouTube for $1.6B in 2006? That amount seemed ridiculous at the time. Its now peanuts compared to YouTube's yearly revenue.
Let's just hope the pitch wasn't selling everybody's data for training
2006 called and wants their valuations back.
just imagine that the money isn't real and you're there.
I love that they won as a remote team and no constant hype about revenue.
Middleman company buying middleman company. Makes sense.
Love openrouter, excited for future feature development!
Here's Stripe's letter to investors explaining its acquisition of OpenRouter (LEAKED)
https://x.com/EricNewcomer/status/2090133049291788434
Any non X source for this so people who don't use X can have a look?
https://xcancel.com/EricNewcomer/status/2090133049291788434
Edit: ninja'd?
for you and me: https://xcancel.com/EricNewcomer/status/2090133049291788434
Just a reminder that cortecs.ai exists. Pretty much the same as OpenRouter but build in Europe.
So what's the endgame? Typical late acquisitions are meant to destroy things. What will they destroy with this merger? Openrouter has no moat.
Congratulations to the OpenRouter team!
Reminds me of eBay acquiring Skype
eBay actually ended up making a net gain of approximately $1.4 billion on its total investment in Skype.
Are you perhaps thinking of Microsoft acquiring Skype for $8.5 billion? (a similar amount as the OpenRouter deal)
Shut up and take my money (via Stripe)!
Congratulation to all who exit at $7b after 3 years, but I have only one question: what problem does this acquisition fix? Did OR lack infrastructure? experience? what was the main drive, other than money, for sale?
Because after PayPal, Stripe is the biggest graveyard of frozen accounts who would be otherwise perfectly fine, but some AI decided that this card from this country at this price = send the account to null.
So while before you had comfort of using OR, right now you will have Stripe emps and their AI overseeing your queries. And don't get me wrong - Stripe will love you! But once someone pulls the lever - good luck getting at least explanation why they had to boot you out. I never heard of an account being reinstated. This will be like this, only your queries instead of CC charges, will be inspected.
I don't see how this is beneficial to anyone, esp. OR users. RIP OpenRouter.
It's honestly insane how valuable so many startups end up being just from AI, wonder if therels ever going to be another breakthrough like this for future people
Fees will magically appear between agents where they don't need to be
What could go wrong
good move by stripe
Why Stripe? 7 billion reasons to build a squidproxy
Just as the US deems China as a national security risk, I deem Israel to be a much bigger threat to my cybersecurity. Stripe has vocally supported Israel and I cannot trust any token coming out of OpenRouter after this. Their payment systems have been notoriously bad so this is cause for concern to me.
Stripe is a financial institution that has strict requirements for reporting and audits by the US government. It has blocked payments for numerous different services for no reason. I wouldn't trust any infrastructure built on OpenRouter.
What exactly is "Open" about OpenRouter anyway?
Congratulations to the open router team. Also: Praying for stripes imminent demise.
With this I am looking to moving my operations to Eurorouter. I am so done with these people (stripe)
Kiss of death for OpenRouter.
In five years they will be struggling.
In ten they will be dead.
I dont use open router so can someone explain to me how it works in that different vendors have different api parameters?
is the selling point that they just give you access to a unified access in where you get to write each request toward the particular api provider's api spec? if so why not just sign up with the vendor directly?
Absolutely nuts to me that a front end aggregator of ai vendors — basically an proxy — plus telemetry is worth 7B
What a time to be alive. I’m going to take a 1% stake in a bridge and rename it bridge.ai and sell it for billions too
Amazing. Huge congrats to the Open Router team
I would guess the real value of the position Openrouter has in the ecosystem is that they can intercept all data and sell training data from many people, training data that is about as directly relevant for training as it can get. They would not need to log all your data for this, "sampling" is enough, and might in fact be even better sampled.
Note: this would include both the request, the response, and competitive responses. This would be very high quality training data.
Now at the current point they're sporting and make this data sampling option opt-out, and indeed they do. By default, they collect your data.
Model companies will pay a lot for this, especially now, more than customers will ever give to openrouter. Think about this: this is a constant stream of data of exactly what people and companies do with AI models, for all models, updating LIVE.
Second, they can provide "competitive intelligence" as well, telling models live how well they're doing and what exactly their strong and weak points are, and in what distribution of token requests this results.
There may even be financial companies that pay for this data.
so basically anybody who uses openrouter from now on is going to have all their data exfiltrated to stripe and then to all the banks/insurance companies... when I read the article that's what I think is going on and it doesn't smell good....
Previously: https://news.ycombinator.com/item?id=49323381
> Today, we are excited to announce that we are...
Why do they always start with this. Every single one of them. You dont even want to read anything after that. Its the same "Blah blah nothing will change you will get the best of both worlds blah blah"
This is before Stripe completing their acquisition of PayPal (which will happen). [0]
Maybe there is some plan to pump Stripe's valuation to over $1TN before an IPO.
Who knows.
[0] https://www.reuters.com/business/finance/stripe-advent-offer...
I think the PayPal acquistion is to buy something that throws of real cash, to prevent a down round for Stripe when valuations reset.
glad I never used them now
Woah now also a 75perc discount on OpenRouter for flash 3.7. Is it really the same product (speed? and up time)? Why would Google do that?
Gemini Flash 3.7 is already "technically" 50% off everywhere until January 2027.
Why stripe page? The money... LOL
> For years, OpenRouter has been called “Stripe for LLMs.”
holy crap, save some butter for the bread omg
aaand they took crypto payments offline. Meaning all accounts will need some form of KYC since you go through stripe.
I don't see the use for open router, I can switch API providers without any hassle with just a couple of lines of code and a drop down. Good for them and for those who milk the AI bubble
good
nice!
As a final customer, I hate so much stripe that I'm disappointed that they will be able to impose their shitty rules and business logics to OpenRouter...
> We are the best way to discover and use any AI model
He says it like it’s a huge accomplishment that he achieved when it’s an opinion about something trivial
hahahhaha
There should be a ban on using Open* for for-profit VC-backed companies.
That prefix is dead and has lost its meaning. Now you gotta completely avoid it to stand out.
That’s great for the team. Though as a user of openrouter, I am worried that the much larger company will soon find ways to enshittify it :S
Switching costs are zero and alternatives are plentiful.
OpenRouter is a name.
What are some good alternatives?
Some other comments suggested https://cortecs.ai/. I never tested it, but from a brief look, it does seem a viable alternative.
You're right in that OpenRouter alone does not have embedded switching costs, but bundled with Stripe's other offerings, it creates lock-in.
Congrats!
"Don't post generated text or AI-edited text. HN is for conversation between humans."
https://news.ycombinator.com/newsguidelines.html
sorry, I thought it might be interesting to see the backstory of the company. I also clearly marked that part of my comment is from ChatGPT.
I can delete my comment, though.
Sorry - that probably came across as more passive-aggressive that I mean it. Just wanted to point out that guideline. People have easy access to AI, so they can always ask it for a summary if they want to and it doesn't really add anything to the discussion.
upd: removed ai-edited text from my comment
Worth noting that quoting and citing an LLM is miles better than just pasting AI generated text and passing it as your own.
AND, if you are too harsh on the former, you incentivize the latter.
At this point I believe that when people do not cite using AI to generate their comment it's from a position of malice... like you're intentionally trying to position yourself as capable in whatever subject you're talking on but realistically you are not given the reliance on an LLM. Not saying anything about OP but just a general observation/opinion I've been building recently. Seems more common than ever to larp using AI.
I did it once when I tried vibecoding, to a friend, nothing professional, and I watched myself stop at the explanation of how I did something. I think it comes from a place of desire, self delusion and shame (of not being able to do it alone), very similar to lying in general, people lie about what they want to be the truth.
> quoting and citing an LLM is miles better
Not really. If someone is interested in an LLM's opinion, they can ask it themselves, all LLM providers are one browser tab away.
I'm as antiAI as many. And this isn't a great example, but there's can be value added in both the context and the prompt. E.g:
"I recall topic X being related to fact Y.
Chatgpt says this when asked about how Y relates to X
"blablabla"
"
Is that 100M of openrouter revenue? Or money that flows through openrouter? If the latter, 100M$ annualized looks like very little. It's as if a bank said that it has 100M$ of annual transfers, instead of 100M in revenue from bank fees and such.
Hard to think of any type of company that can do the topline meme as effectively as a compute aggregator proxy. Almost like a bitcoin pool mining company counting their shared miners.
Really? 100M annual? That HAS to be profit, it can't be this crazy.
EDIT: From their website, they say they have 200T monthly tokens, at 10 cents per M token ( optimistic average between free tokens and expensive tokens) that's 20M$/month, which is roughly 240M$.
Bonus track: here's chatgpt completely fumbling the above very simple calculation
https://chatgpt.com/share/6a85f831-9118-83e9-8a06-d32195c557...
I don't doubt the chinese bootleg product that openrouter is peddling is even worse than that.
We are in a bubble confirmed.
Sorry for the negativity, but this is highly bearish news.
We are in a bubble, but you are looking at this in the wrong way. OpenRouter basically creates lock-in through Stripe's distribution and family of products; it is much stronger within Stripe than alone. I agree valuations are absurd, but it is what it is. But if we have to go through a bubble for a16z to crash and burn, I am all for it.
So customers use OpenRouter to avoid lock-in to specific model providers, and in doing so they lock-in into openrouter's aggregation API?
Lock-in to what exactly? A string(string) function without any further restrictions? Note that aggregator users will not use more specific parameter features (because those vary by vendor), or they escape the aggregator and lock-in directly string(string, vendorOptions={"openAI.logprobs":true})
Selling vendor lock-in to vendor lock-in avoidant customers sounds like a losing proposition, like trying to beat competition on price and undercutting, which is a thing openrouter does too btw, the main reason people use these products is to get cheaper prices. I don't think 2$/M tokens is expensive, and I don't think those that try to cut costs are going to win, whether devs cutting token costs, or small business owners with 5 employees who look at AI as a way to fire 2 of them. I get that the other end of the spectrum is overspending and tokenmaxxing, but the conclusion is that extremes are bad? Nothing new here.
Wow, that's a win-win! OpenRouter makes things a breeze.
Going through the openrouter team - turns out I know the CEO from an exchange trip we did several years ago (Oxford/Stanford)! Congrats Alex and team!