The arguments made by the Plaintiff are thoroughly convincing to me. The fact that those 8 points are not enough to convict indicate the industry is fucked. Of course the defendants didn't leave a paper trail - they've already been convicted of collusion before.
It's the people and country that suffer when our government fails to ensure markets are free and fair.
If there is no agreement, then the magic term is "tacit collusion."
Why not ask the same ridiculous amount of money your competitors do? People seem to be paying for it. Their fault. If suppliers have sufficiently different products, they can make some more expensive, others cheaper; on average, everybody pays more. A high barrier to entry might help such practices.
That doesn't mean I'm saying this is what is happening. Sometimes things just suck, and somebody bought the world's supply of RAM wafers to use as frisbees.
We have to decide what part is damaging to society: the actual physical agreement, or the effects of the agreement?
If it's the actual physical agreement that's the problem - the system is working as intended.
But if we are looking to prevent the negative outcomes associated with price fixing and collusion, our system is failing us.
They are never going to find proof of conspiracy. The people involved covered their tracks, and doing so is trivial. So the best we can do is punish the appearance of collusion. And if the goal is to actually prevent harm to customers, that's a better solution anyway, since it encourages leaders of companies to behave in a manner that's the opposite of collusion.
HBM is also DRAM. I also think it's kind of a weak argument to say that them discontinuing ddr3 (which while in use still today in industrial/embedded was on the way out for consumers 10 years ago) and ddr4 which last had consumer CPUs for it 3 years ago is meaningful. What we need now is ddr5. Turning off the old fabs and moving those resources including people to ddr5 is a good thing. That's not price fixing. It's possible price fixing is in play, but discontinuing products people objectively don't use as much anymore isn't it.
Switching off DDR3 manufacturing I can understand, but DDR4 machines are still quite relevant and usable… Ryzen 5000 series boxes for example don’t feel meaningfully weaker than they did when new. My 5950X tower certainly doesn’t, and it’d really be nice to be able to upgrade its RAM should I need to because it will continue to be useful for quite some time.
AMD just re-released their 5800X3D for AM4 board users who wish to upgrade which is further evidence that shutting off DDR4 production is premature.
They're running a business not a charity. Their job is to manufacture what the market as a whole demands. If they can make more money making HBM than DDR4 then they have to make HBM. Why would a business go out of its way to make less money?
1. But it might last for at least few more years, see Nvidia 1 trillion backlog.
2. Semiconductor manufacturing is the most complex industrial process in the world. You need billions of capex and decades of experience. Even existing semi players like intel cannot switch production to memory.
China CXMT is gaing traction in DDR market. New fabs from all players wil come online in the next two years.
That isn't true, Apple uses unified memory of their own design and fabrication. They don't use Hynix, Samsung, Micron, etc. It's why they haven't been doubling the price of their machines like has happened in the PC space. They already have insane amount of reserved capacity through TSMC that was prepaid years ago.
That's incorrect. Apple purchases RAM from all of these providers to produce their unified memory. They also rely on TSMC fabs for all of the chips their memory relies on. If they haven't doubled the prices of their machines, that just goes to show how fat of a profit margin they take on everyone that buys from them...
Dropping DDR-4 is anything but meaningful. It'll easily last 10 more years, machines from this gen are still much more affordable and quite powerful. In fact for most dev and gaming workflows the difference between the DDR-4 and DDR-5 generation of hardware is more or less negligible. I am exaggerating a bit -- but really, not too much.
Of course it might be a ploy to sheep-herd consumers and companies towards the expensive DDR-5. I would not put that below the ring of RAM producers.
>Dropping DDR-4 is anything but meaningful. It'll easily last 10 more years, machines from this gen are still much more affordable and quite powerful. In fact for most dev and gaming workflows the difference between the DDR-4 and DDR-5 generation of hardware is more or less negligible. I am exaggerating a bit -- but really, not too much.
How much % of the DRAM market do you think is made from computer enthusiasts upgrading their Zen 1/2 CPUs to Zen 3? Note intel and AMD both switched to DDR5 well before the exit from DDR3/DDR4 ("2024-2025", according to the complaint).
Note though that for Intel, the first gen of DDR5 CPUs also supported DDR4, and many buyers bought the DDR4 versions of their boards because at that point DDR5 RAM was much more expensive for gains that were marginal at best, which effectively makes Intel’s following generation of DDR5 CPUs the actual transition point.
The point is that just because there's a handful of people (relatively speaking) looking to buy more RAM to upgrade their last gen systems, doesn't mean there's robust demand for DRAM manufacturers to keep DDR4 manufacturing lines going. It's like arguing Sony shouldn't have exited the CRT business because there's retro enthusiasts on youtube scouring the earth for CRT monitors.
For the reasons stated above -- that DDR5-gen hardware is expensive right now -- I'd think the DDR4-gen market will remain alive for quite a big longer. Though that's likely much more on the second-hand market side of things.
While I wouldn't necessarily agree with "a handful of people", the fact is that neither of us can prove their lean -- so no point pursuing that argument thread.
So you might be right that it's a pure numbers/statistics decision. Or I might be right that they want to herd people into the more expensive hardware while forcing them to do so by phasing out production of the cheaper hardware.
No way to truly know IMO. We are exchanging hypotheses.
I’m not sure that comparison makes much sense. By the time CRTs were phased out, demand was down to almost nothing and what little existed was confined to the extreme budget market. While I don’t have industry insights or anything I don’t think demand for DDR4 is anywhere near the bottom yet, and the remaining demand is centered on premium product (nobody running cheap DDR4 is upgrading). In a more normal market would be more than enough to justify continued production for several more years.
DDR4 production is likely still quite profitable, just not drowning-in-money AI-bubble profitable. If smaller foundries existed they’d be happy to take up the business.
Maybe really what needs to happen is some busting up of the giants…
There is a minimum amount of production volume for it to fit the price equation. If the market doesn't have that demand, it is fundamentally no different to CRT.
Otherwise they could continue to make DDR4 at a higher cost and sold at a higher price to which people will complain price fixing again.
It's easy to see why your argument is wrong with a simple hypothetical: what if they were still making DDR4 today? Would people still buy it?
The answer is an obvious "fuck yeah", even if you ignore the DDR5 price gouging. People will buy it because people still have DDR4 hardware, and that hardware is still extremely relevant.
So if there's a market for it, but none of the suppliers are trying to sell to it... Wtf is happening? Basic capitalism logic says any rational supplier would sell DDR4 for easy profits, meeting an unmet demand. That it isn't happen points to some kind of collusion, IMO.
Because the market pays less for DDR4 than for HBM (or DDR5), and since HBM is heavily modified, vertically stacked DRAM, it competes for the same raw inputs and fab space than DDR4 used.
If I can produce DDR4 for modest profit or HBM for a lot more profit I will obviously produce HBM. And given physical realities producing HBM takes from existing DDR4 production capacity. Worse still, it takes roughly 3GB of ram to produce 1GB of hbm iirc.
> People will buy it because people still have DDR4 hardware
The question is whether there’s enough meaningful demand for aftermarket DDR4 upgrades to make it worthwhile to a manufacturer to keep producing DDR4 instead of switching to HBM and DDR5.
Micron claimed retail is a rounding error, a market not worth serving. So you’d presumably need to find industrial buyers who would be willing to buy DDR4.
Micron is forced to stop under-investing in plants and will increase production. This will trigger everyone else to expand production and lower prices.
The whole point of the collusion is to ensure everyone is producing the same volumes and keeping prices high. The company that expands is the company that "wins" because memory is a volume game and it's all about hanging on the longest during the glut. So once one company expands, the rest have a choice of expanding or planning their exit.
If Samsung and SK lose access to the US market, they'd be fucked long term. Micron would kill them selling at higher margins and higher volumes in the USDM, while the rest are stuck competing for the international scraps - markets Micron is also allowed to compete in, if they wanted to.
Everyone is using something from someone, you can even argue that US owes India and Europe huge compensation because pretty much everything US did in the last half century was made using technology or people funded by those people. Johny Ive is British, Almost all the AI stuff is created by Europeans, Israelis, and Canadians - thus funded by their respective taxpayers.
The thing about the US losing its grip on the world and the collapse of the global world order means that the words on the paper don't mean much. Embargoes on Russia didn't mean much so Europeans are physically taking over their ships and Ukrainians are physically sinking the rest of their ships. In Iran nothing other than physically sinking ships and blowing up places meant anything.
Europeans can ship EUV machines because they are physically building them for people who will use these to physically build the most valuable products currently there is. US wasn't able to enforce its will to Iran, what if Koreans, Europeans and the Chinese decide that its not into their interest to act according to US courts?
The collapse of global trade would greatly reduce economic efficiency, output, and investment. It has been coming for while, though greatly accelerated by the orange pdf file. It takes a lot longer to build systems of trust and belief in enforcements of global order than to disrupt them. I suppose we'll move closer to the fear side of the financial/political axis from the greed side.
If history is any guidelines that would be how World War III starts.
There is nothing that stops US from building their own Memory Fabs, or asking / funding Micron building more US Fabs. It will cost a more, but the complexity is certainly no where near replicating TSMC.
US is in a very advantageous position regarding geography and resources but its problem is that its geared towards having access to the whole worlds markets. Apple, Google etc. are all possible because they server billions of people, not just 350M. IMHO US will have serious internal trouble for years, eventually stabilizing and being a nice place again.
The market is the AI boom and the US is the host, they can sell the exact same stuff to someone else. What are the capitalists who fund the AI build up do? Invest in SaaS when they can't buy chips? I bet if something like that happens the chip manufacturers wouldn't end up with product they have no one to sell to.
So in other words, if Koreans and Europeans decide not to sell their stuff to America,quits the AI race and the capitalist do something else instead? I don't think so, in the hypothetical world where Korea and Europe don't sell to US, the American money that is invested in AI will go wherever they can actually build it, the people who are using these machines to build those models are mostly immigrants anyway.
If Koreans/Europeans decide not to sell their stuff, some of the money would go into circumventing the ban:
1) This is not good for Korean/Europen sellers, because it negatively affects sales volume, and is unlikely to be compensated by margins, because a good chunk of those will go towards circumvention instead of the original seller.
2) Some more money will go towards replacing those sellers completely. This is extra not good from the sellers perspective.
In this new world there are more important things like nationalism, the ego of the politicians and the personal ambitions of the super rich. Bot being good idea is irrelevant.
I don't understand what this hypothetical world is supposed to be. Samsung and SK Hynix are run by Korean capitalists who want to sell their stuff to America; the reasons you're describing are precisely why they would not exit the American market just to dodge an inconvenient antitrust investigation.
memory is a commodity is laughable. Then software engineering is even more a commodity, the amount of engineering going into making memory chips the vast majority of people don't understand. There are a lot of software engineers getting this field after leetcoding and copy from hellointerview. Claude can write you an app in 30 minutes. Try build a lpddr5 dram chip in 30 minutes. Manufacturing know how itself is a specialty and barrier to entry.
I mean your view isn't flawless but overall I agree. Too many people think building things amount to spending money and completely overlook the thousands of people required who are not just unskilled labor hired off the street.
There's "fat margins" and then there's the kind of margins that tech industry shareholders expect.
OpenAI's original corporate agreements capped its returns at 100x, which is seen as too paltry for its current holders, so they scrapped those to prepare for an IPO [0]
>But god forbid if someone else captures the margins and squeeze them out of easy profits.
Yeah that's how law works. Everyone likes money, but that doesn't mean it's fine to steal money. Yes, even from maligned entities like "big tech" or "private equity"
I also feel like there should be FTC or other antitrust actions against OpenAI and other hyperscalers, with nvidia being complicit, if they are cornering the market that badly on consumer RAM, SSD and other components, especially if these volume purchases are for projected datacenter that haven't even broken ground (or have been paid for) for many months.
Furthermore, I think there should be a tax on algorithmic inefficiency, in that if a LLM, frontier or not, consumes more than a certain amount of KWH per token, it should be taxed such as to put emphasis on models than can run locally, on a normal PC
dumb peripheral question: memory as a circuit would seems easy to add directly on-chip for a company like Apple. Is that blocked by time-frame, IP, technology, cost?
Antitrust and anticompetitive behavior continue to evolve. Back when the Sherman Act was passed, we were talking about backroom deals in smoke-filled rooms. One of the more important ways of recent years is where all or most competitors in a market use the same software that outputs the same value as to what they should charge. This is effective collusion even if it isn't explicit. The goal of such software is to raise prices. You raise prices by effectively or actually colluding with other market participants.
The posterchild for this is RealPage [1]. But you're going to see this pop up in every aspect of life, such as gas prices [2] and meat processing [3].
The whole thing is kind of depressing because this is what "innovation" is now: fancy ways to collude on prices.
Maybe software people can get around this issue by not making every app a electron bloat? This is now more than doable now you got AI right? And it will save your job
Honestly, the knock on effects of this cartel behavior should concern all countries, and be remediated with expediency.
From consumer electronics to the data-center, the rising real mfg costs and lack of supply is putting huge pressure on pricing, and may just drive anyone who cant negotiate with these suppliers out of business.
Once the dominoes start, I fail to see how things recover in less than 3-5 years, not counting all the businesses wiped out in the meanwhile.
I mean there was an attempt before, and maybe I'm wrong but I'm sure they were fined in 2010 at least in the EU, this seems to be round 2 of their 'Memory Cartel' after learning some lessons, realising all they need to do is keep supply as low as they want and allow the AI companies to spend hand over fist.
Just a reminder that anyone can file a lawsuit over anything, and the initial complaint is written by lawyers and reads with tabloid levels of sensationalism and allegation. The goal being to maximize the appearance of harm as much as possible so the suit has the greatest chance of sticking.
It is not in any way, shape, or form a ruling much less even a piece of well researched work. It's "my side of the story that makes me look perfect, with lawyers turning the heat up to 11"
In reality a lawsuit needs to have some basis. The bar might be low but judges frown on having their time wasted with a truly frivolous claim or a claim with no clear damages or harm shown.
Honestly a lot of lawsuits feel like they just want the other party to pay them (settle) to make the lawsuit go away, even if there is no case; what are the repercussions for a frivolous lawsuit?
Realistically very few. The plaintiff carries the burden of proving the claim, but they are afforded the power of the law to do so via discovery. The defendant then has to spend a large number of legal resources refusing the claims and also providing the evidence required.
If the plaintiff loses the lawsuit a countersuit is pretty unlikely to succeed unless the lawyer participates in gross misconduct. Generally, countersuits are filed more to put the original plaintiff on the defense and don't result in a large judgement.
If you are operating in good faith, then you are pretty insular from kickback as a plaintiff
>what are the repercussions for a frivolous lawsuit?
None, hence the high price of liability baked into basically everything in America. And not just in nominal prices, but in terms of things like restricting access to spaces, restricting access to information, etc.
Price fixing is legal as long as your are doing it in the open. In the UK it is called "price match" and eg. if supermarket says they keep prices matched to their competitor. No regulator raises an eyebrow.
So here Samsung and SK Hynix could say they price match to Micron and they are in the clear.
> Price fixing is legal as long as your are doing it in the open.
In the U.S., competitors are allowed to act in similar ways in response to economic realities, as long as they each arrive at that decision independently. But publicly anchoring your price to a competitor’s is potentially illegal.
> Price fixing is an agreement (written, verbal, orinferred from conduct) among competitors to raise, lower, maintain, or stabilize prices or price levels.
Identical prices are often a sign of intense competition. Every gas station on a corner has the same price because it's a highly competitive market not because of collusion. The prices of the much more lucrative chocolate bars inside the gas stations are less likely to be identical.
Yes, the gas station example is directly cited in the article I linked to. It’s legal for a gas station owner, with knowledge and consideration of a competitor’s price, to reduce their price to the same or just below. What is illegal is for nominally-competing gas station owners in an area to conspire to keep their prices within a range of each other’s, even without explicit agreement.
"Every gas station on a corner has the same price because it's a highly competitive market not because of collusion."
Huh? I can go to most any gas station-occupied intersection and you will always find two that match and one (usually a Persian-owned Chevron) which is consistently a dollar or more higher per gallon across all grades of fuel.
>Price fixing is legal as long as your are doing it in the open. In the UK it is called "price match" and eg. if supermarket says they keep prices matched to their competitor. No regulator raises an eyebrow.
No, the key term is "collusion", which could be done in the open or not. If a competitor told you they were unilaterally raising prices in secret, that would still be legal. Where you get into trouble is if you are cooperating to set prices. And no, this is all determined by a judge so cute workarounds like "I'm telling my competitors that I'm raising prices then gauging his body language" won't work.
I worked in a small town gas station as a kid. We always phoned our competitor to let them know when we were changing our prices, and they extended us the same courtesy. Usually we followed them, but sometimes we didn't.
Gas prices are posted on massive highly visible signs and are public information. This wasn't collusion, it was a sign of intense but friendly competition.
That is not what price fixing actually is. In the UK "price matching" is a one-to-many relationship meaning that the price of goods is set to the "lowest available"
Price fixing is a many-to-one all the manufactures agree to the highest prices they all agree on and set it there.
That is your assumption without evidence. Once the targeted market knows others are tracking the price, it can up the price without fear competing market will undercut.
cxmt is also selling their memory similar to the big three. No one hardware company in their right mind will sell their products not as high as possible after you learn how much harder hardware engineer and fab people work for.
There is no reason for there not to be a price fixing. But the OpenAI's announcement from 2025-10-01 about buying 40% of supply, removes any need for collusion. It was a public signal for everyone to rise prices, one that each company could figure on their own. And it will be very hard to prove otherwise.
Two things can be true; they could theoretically make agreements to limit supply, like the OPEC does to exert control over international oil prices. But the OPEC is all above board and there's plenty of international competition on the oil market.
Prices skyrocketing create a perfect opportunity to slip in a little artificial bump and hope everyone blames the market. See also: egg prices in 2024.
if demand is too high, the market will adjust by adding supply. If Apple can't take the price of memory, why don't they make memory themselves. Oh no, the technical and manufacturing know-how is a barrier to entry and there is no talent available to make it. That is memory or semi companies' competitive advantages and their pricing power. It also takes Phd degrees to work at hardware companies. You can't have people doing leetcode for 2 months and todo apps and getting into the field and make 300k a year and hardware engineers wih Phds making 100k before this boom. It is long overdue for a rebalance of pricing power between hardware and software companies
,,The plaintiffs claimed the three companies reduced D-RAM supply under the pretext of transitioning to high-bandwidth memory (HBM). "The D-RAM oligopoly companies systematically coordinated the shift to HBM and the discontinuation of DDR3 and DDR4," they said. They added that Apple's recent sweeping product price increases were the trigger for the lawsuit.''
How can they do price fixing and discontinuing a product at the same time?
It just looks like some companies are angry that AI / VC industry is outpricing them.
> An agreement to restrict production, sales, or output is just as illegal as direct price fixing, because reducing the supply of a product or service drives up its price. For example, the FTC challenged an agreement among competing oil importers to restrict the supply of lubricants by refusing to import or sell those products in Puerto Rico. The competitors were seeking to pressure the legislature to repeal an environmental deposit fee on lubricants, and warned of lubricant shortages and higher prices.
Looking forward to getting my $10 settlement 15 years from now like what happened with the DVD drive price fixing lawsuit (paid as a digital Visa gift card code so it can't actually be spent anywhere).
This was attempted before in 2022 but fell apart because plaintiff couldn't show an agreement took place https://cdn.ca9.uscourts.gov/datastore/opinions/2022/03/07/2...
The arguments made by the Plaintiff are thoroughly convincing to me. The fact that those 8 points are not enough to convict indicate the industry is fucked. Of course the defendants didn't leave a paper trail - they've already been convicted of collusion before.
It's the people and country that suffer when our government fails to ensure markets are free and fair.
If there is no agreement, then the magic term is "tacit collusion."
Why not ask the same ridiculous amount of money your competitors do? People seem to be paying for it. Their fault. If suppliers have sufficiently different products, they can make some more expensive, others cheaper; on average, everybody pays more. A high barrier to entry might help such practices.
That doesn't mean I'm saying this is what is happening. Sometimes things just suck, and somebody bought the world's supply of RAM wafers to use as frisbees.
https://en.wikipedia.org/wiki/Tacit_collusion
We have to decide what part is damaging to society: the actual physical agreement, or the effects of the agreement?
If it's the actual physical agreement that's the problem - the system is working as intended.
But if we are looking to prevent the negative outcomes associated with price fixing and collusion, our system is failing us.
They are never going to find proof of conspiracy. The people involved covered their tracks, and doing so is trivial. So the best we can do is punish the appearance of collusion. And if the goal is to actually prevent harm to customers, that's a better solution anyway, since it encourages leaders of companies to behave in a manner that's the opposite of collusion.
It wouldn't be a first: https://en.wikipedia.org/wiki/DRAM_price_fixing_scandal
It's been happening once every decade or so. I'm not surprised that they are being sued again.
I thought the title missed the year (in parenthesis)
Total fines of about half a billion. Thats not nothing but it also does not seem like a real disincentive.
Samsung is about to hand out ~$26B in bonuses. SK Hynix something similar-ish.
Yes, but for context Samsung is a massive chaebol. You need to compare it to the business unit in question.
It’s different to, say, Google’s vertical monopoly in advertising where that’s most of their revenue.
HBM is also DRAM. I also think it's kind of a weak argument to say that them discontinuing ddr3 (which while in use still today in industrial/embedded was on the way out for consumers 10 years ago) and ddr4 which last had consumer CPUs for it 3 years ago is meaningful. What we need now is ddr5. Turning off the old fabs and moving those resources including people to ddr5 is a good thing. That's not price fixing. It's possible price fixing is in play, but discontinuing products people objectively don't use as much anymore isn't it.
Switching off DDR3 manufacturing I can understand, but DDR4 machines are still quite relevant and usable… Ryzen 5000 series boxes for example don’t feel meaningfully weaker than they did when new. My 5950X tower certainly doesn’t, and it’d really be nice to be able to upgrade its RAM should I need to because it will continue to be useful for quite some time.
AMD just re-released their 5800X3D for AM4 board users who wish to upgrade which is further evidence that shutting off DDR4 production is premature.
They're running a business not a charity. Their job is to manufacture what the market as a whole demands. If they can make more money making HBM than DDR4 then they have to make HBM. Why would a business go out of its way to make less money?
1. The AI bubble is an insane distortion and the gravy train isn’t going to last forever. Betting the farm on everlasting datacenter demand is myopic.
2. In a healthy, competitive market there would be smaller manufacturers that’d be happy to take up the big guys’ discarded business.
1. But it might last for at least few more years, see Nvidia 1 trillion backlog.
2. Semiconductor manufacturing is the most complex industrial process in the world. You need billions of capex and decades of experience. Even existing semi players like intel cannot switch production to memory.
China CXMT is gaing traction in DDR market. New fabs from all players wil come online in the next two years.
People will remember these shenanigans and will buy Apple after getting ripped off by PC manufacturers.
Not sure I understand this statement. Apple is affected by DRAM price fixing the same as any other PC manufacturer.
That isn't true, Apple uses unified memory of their own design and fabrication. They don't use Hynix, Samsung, Micron, etc. It's why they haven't been doubling the price of their machines like has happened in the PC space. They already have insane amount of reserved capacity through TSMC that was prepaid years ago.
Apple just increased their prices a few days ago, citing high memory costs.
That's incorrect. Apple purchases RAM from all of these providers to produce their unified memory. They also rely on TSMC fabs for all of the chips their memory relies on. If they haven't doubled the prices of their machines, that just goes to show how fat of a profit margin they take on everyone that buys from them...
Dropping DDR-4 is anything but meaningful. It'll easily last 10 more years, machines from this gen are still much more affordable and quite powerful. In fact for most dev and gaming workflows the difference between the DDR-4 and DDR-5 generation of hardware is more or less negligible. I am exaggerating a bit -- but really, not too much.
Of course it might be a ploy to sheep-herd consumers and companies towards the expensive DDR-5. I would not put that below the ring of RAM producers.
>Dropping DDR-4 is anything but meaningful. It'll easily last 10 more years, machines from this gen are still much more affordable and quite powerful. In fact for most dev and gaming workflows the difference between the DDR-4 and DDR-5 generation of hardware is more or less negligible. I am exaggerating a bit -- but really, not too much.
How much % of the DRAM market do you think is made from computer enthusiasts upgrading their Zen 1/2 CPUs to Zen 3? Note intel and AMD both switched to DDR5 well before the exit from DDR3/DDR4 ("2024-2025", according to the complaint).
Note though that for Intel, the first gen of DDR5 CPUs also supported DDR4, and many buyers bought the DDR4 versions of their boards because at that point DDR5 RAM was much more expensive for gains that were marginal at best, which effectively makes Intel’s following generation of DDR5 CPUs the actual transition point.
The mere "enthusiast" word in your question suggests the percent is not too big. But I am not sure I get your point -- elaborate, please?
The point is that just because there's a handful of people (relatively speaking) looking to buy more RAM to upgrade their last gen systems, doesn't mean there's robust demand for DRAM manufacturers to keep DDR4 manufacturing lines going. It's like arguing Sony shouldn't have exited the CRT business because there's retro enthusiasts on youtube scouring the earth for CRT monitors.
For the reasons stated above -- that DDR5-gen hardware is expensive right now -- I'd think the DDR4-gen market will remain alive for quite a big longer. Though that's likely much more on the second-hand market side of things.
While I wouldn't necessarily agree with "a handful of people", the fact is that neither of us can prove their lean -- so no point pursuing that argument thread.
So you might be right that it's a pure numbers/statistics decision. Or I might be right that they want to herd people into the more expensive hardware while forcing them to do so by phasing out production of the cheaper hardware.
No way to truly know IMO. We are exchanging hypotheses.
I’m not sure that comparison makes much sense. By the time CRTs were phased out, demand was down to almost nothing and what little existed was confined to the extreme budget market. While I don’t have industry insights or anything I don’t think demand for DDR4 is anywhere near the bottom yet, and the remaining demand is centered on premium product (nobody running cheap DDR4 is upgrading). In a more normal market would be more than enough to justify continued production for several more years.
DDR4 production is likely still quite profitable, just not drowning-in-money AI-bubble profitable. If smaller foundries existed they’d be happy to take up the business.
Maybe really what needs to happen is some busting up of the giants…
There is a minimum amount of production volume for it to fit the price equation. If the market doesn't have that demand, it is fundamentally no different to CRT.
Otherwise they could continue to make DDR4 at a higher cost and sold at a higher price to which people will complain price fixing again.
It's easy to see why your argument is wrong with a simple hypothetical: what if they were still making DDR4 today? Would people still buy it?
The answer is an obvious "fuck yeah", even if you ignore the DDR5 price gouging. People will buy it because people still have DDR4 hardware, and that hardware is still extremely relevant.
So if there's a market for it, but none of the suppliers are trying to sell to it... Wtf is happening? Basic capitalism logic says any rational supplier would sell DDR4 for easy profits, meeting an unmet demand. That it isn't happen points to some kind of collusion, IMO.
Because the market pays less for DDR4 than for HBM (or DDR5), and since HBM is heavily modified, vertically stacked DRAM, it competes for the same raw inputs and fab space than DDR4 used.
If I can produce DDR4 for modest profit or HBM for a lot more profit I will obviously produce HBM. And given physical realities producing HBM takes from existing DDR4 production capacity. Worse still, it takes roughly 3GB of ram to produce 1GB of hbm iirc.
> People will buy it because people still have DDR4 hardware
The question is whether there’s enough meaningful demand for aftermarket DDR4 upgrades to make it worthwhile to a manufacturer to keep producing DDR4 instead of switching to HBM and DDR5.
Micron claimed retail is a rounding error, a market not worth serving. So you’d presumably need to find industrial buyers who would be willing to buy DDR4.
Basic capitalism logic is that if you think it's stupid, you put your money where your mouth is, set up a DRAM fab, and get rich.
I'll do it. Will you give me the seed money?
No, basic capitalism logic is that you already have enough money because barriers to competition are low.
What happens if Samsung and SK Hynix simply stop selling to US at all? Micron is in US but are the rest still in the US jurisdiction?
They are selling the hottest commodity of the day. It’s made outside of the US using non-American tooling.
Micron is forced to stop under-investing in plants and will increase production. This will trigger everyone else to expand production and lower prices.
The whole point of the collusion is to ensure everyone is producing the same volumes and keeping prices high. The company that expands is the company that "wins" because memory is a volume game and it's all about hanging on the longest during the glut. So once one company expands, the rest have a choice of expanding or planning their exit.
If Samsung and SK lose access to the US market, they'd be fucked long term. Micron would kill them selling at higher margins and higher volumes in the USDM, while the rest are stuck competing for the international scraps - markets Micron is also allowed to compete in, if they wanted to.
>It’s made outside of the US using non-American tooling.
Depends if US can demand ASML which uses plenty of US tech inside. In reality even the DRAM and NAND supply chain has plenty of US technologies.
And you say Micron are US but they have lots of Fabs in Japan as well since they acquired Elpida.
Everyone is using something from someone, you can even argue that US owes India and Europe huge compensation because pretty much everything US did in the last half century was made using technology or people funded by those people. Johny Ive is British, Almost all the AI stuff is created by Europeans, Israelis, and Canadians - thus funded by their respective taxpayers.
The thing about the US losing its grip on the world and the collapse of the global world order means that the words on the paper don't mean much. Embargoes on Russia didn't mean much so Europeans are physically taking over their ships and Ukrainians are physically sinking the rest of their ships. In Iran nothing other than physically sinking ships and blowing up places meant anything.
Europeans can ship EUV machines because they are physically building them for people who will use these to physically build the most valuable products currently there is. US wasn't able to enforce its will to Iran, what if Koreans, Europeans and the Chinese decide that its not into their interest to act according to US courts?
The collapse of global trade would greatly reduce economic efficiency, output, and investment. It has been coming for while, though greatly accelerated by the orange pdf file. It takes a lot longer to build systems of trust and belief in enforcements of global order than to disrupt them. I suppose we'll move closer to the fear side of the financial/political axis from the greed side.
If history is any guidelines that would be how World War III starts.
There is nothing that stops US from building their own Memory Fabs, or asking / funding Micron building more US Fabs. It will cost a more, but the complexity is certainly no where near replicating TSMC.
US is in a very advantageous position regarding geography and resources but its problem is that its geared towards having access to the whole worlds markets. Apple, Google etc. are all possible because they server billions of people, not just 350M. IMHO US will have serious internal trouble for years, eventually stabilizing and being a nice place again.
> Steve Jobs is Syrian
lol he is not. at no point was he a syrian. his mom was from ohio or something.
You are going to have to cite more for those claims. Steve Jobs was born in San Francisco. He is as Syrian as Trump is.
This is as biased an Anti-US take as any. Will not grace the rest of the claims with a response.
> What happens if Samsung and SK Hynix simply stop selling to US at all? Micron is in US but are the rest still in the US jurisdiction?
They would lose access to their largest market, I'm sure shareholders would havesomething to say about that ?
The market is the AI boom and the US is the host, they can sell the exact same stuff to someone else. What are the capitalists who fund the AI build up do? Invest in SaaS when they can't buy chips? I bet if something like that happens the chip manufacturers wouldn't end up with product they have no one to sell to.
US is the vast majority of their market - Apple, hyper-scalers, AI labs
Why can't they just buy the exact same product and install it in Kazakhstan or somewhere else?
The RAM buyers have no interest in entertaining something like that because their revenue comes mainly from the US, too.
Corporations avoid picking fights with large nations where lots of revenue comes from for very obvious reasons.
So in other words, if Koreans and Europeans decide not to sell their stuff to America,quits the AI race and the capitalist do something else instead? I don't think so, in the hypothetical world where Korea and Europe don't sell to US, the American money that is invested in AI will go wherever they can actually build it, the people who are using these machines to build those models are mostly immigrants anyway.
If Koreans/Europeans decide not to sell their stuff, some of the money would go into circumventing the ban:
1) This is not good for Korean/Europen sellers, because it negatively affects sales volume, and is unlikely to be compensated by margins, because a good chunk of those will go towards circumvention instead of the original seller.
2) Some more money will go towards replacing those sellers completely. This is extra not good from the sellers perspective.
In this new world there are more important things like nationalism, the ego of the politicians and the personal ambitions of the super rich. Bot being good idea is irrelevant.
I don't understand what this hypothetical world is supposed to be. Samsung and SK Hynix are run by Korean capitalists who want to sell their stuff to America; the reasons you're describing are precisely why they would not exit the American market just to dodge an inconvenient antitrust investigation.
And hyper-scalers, Apple, AI labs all will die if memory makers can't sell to them?
regime change in South Korea. President Lee Jae Myung isn't exactly popular among Washington circles
memory is a commodity is laughable. Then software engineering is even more a commodity, the amount of engineering going into making memory chips the vast majority of people don't understand. There are a lot of software engineers getting this field after leetcoding and copy from hellointerview. Claude can write you an app in 30 minutes. Try build a lpddr5 dram chip in 30 minutes. Manufacturing know how itself is a specialty and barrier to entry.
I mean your view isn't flawless but overall I agree. Too many people think building things amount to spending money and completely overlook the thousands of people required who are not just unskilled labor hired off the street.
Everybody in our industry loves fat margins. But god forbid if someone else captures the margins and squeeze them out of easy profits.
There's "fat margins" and then there's the kind of margins that tech industry shareholders expect.
OpenAI's original corporate agreements capped its returns at 100x, which is seen as too paltry for its current holders, so they scrapped those to prepare for an IPO [0]
That is, in a word, insane.
[0] https://abhs.in/blog/openai-for-profit-conversion-ipo-develo...
>But god forbid if someone else captures the margins and squeeze them out of easy profits.
Yeah that's how law works. Everyone likes money, but that doesn't mean it's fine to steal money. Yes, even from maligned entities like "big tech" or "private equity"
I also feel like there should be FTC or other antitrust actions against OpenAI and other hyperscalers, with nvidia being complicit, if they are cornering the market that badly on consumer RAM, SSD and other components, especially if these volume purchases are for projected datacenter that haven't even broken ground (or have been paid for) for many months.
Furthermore, I think there should be a tax on algorithmic inefficiency, in that if a LLM, frontier or not, consumes more than a certain amount of KWH per token, it should be taxed such as to put emphasis on models than can run locally, on a normal PC
dumb peripheral question: memory as a circuit would seems easy to add directly on-chip for a company like Apple. Is that blocked by time-frame, IP, technology, cost?
Antitrust and anticompetitive behavior continue to evolve. Back when the Sherman Act was passed, we were talking about backroom deals in smoke-filled rooms. One of the more important ways of recent years is where all or most competitors in a market use the same software that outputs the same value as to what they should charge. This is effective collusion even if it isn't explicit. The goal of such software is to raise prices. You raise prices by effectively or actually colluding with other market participants.
The posterchild for this is RealPage [1]. But you're going to see this pop up in every aspect of life, such as gas prices [2] and meat processing [3].
The whole thing is kind of depressing because this is what "innovation" is now: fancy ways to collude on prices.
[1]: https://www.justice.gov/archives/opa/pr/justice-department-s...
[2]: https://abcnews.com/US/wireStory/ai-helping-gas-stations-col...
[3]: https://www.justice.gov/opa/pr/justice-department-requires-a...
I don't see how this can really go anywhere, but it'd be nice if it does.
Not like it'd be the first time someone shook a stick at them.
At this point the only hope for change is if China finally decides to get in the game rather than just threatening to.
And our illustrious leader doesn't tariff Chinese RAM and allows its sale in the US.
Maybe software people can get around this issue by not making every app a electron bloat? This is now more than doable now you got AI right? And it will save your job
And replace all your apps with some horrible Tcl/Tk and .NET interface?
Nobody ever made good native widgets.
Honestly, the knock on effects of this cartel behavior should concern all countries, and be remediated with expediency.
From consumer electronics to the data-center, the rising real mfg costs and lack of supply is putting huge pressure on pricing, and may just drive anyone who cant negotiate with these suppliers out of business.
Once the dominoes start, I fail to see how things recover in less than 3-5 years, not counting all the businesses wiped out in the meanwhile.
How do we know it's cartel behavior, not just free market competition?
I mean there was an attempt before, and maybe I'm wrong but I'm sure they were fined in 2010 at least in the EU, this seems to be round 2 of their 'Memory Cartel' after learning some lessons, realising all they need to do is keep supply as low as they want and allow the AI companies to spend hand over fist.
https://ec.europa.eu/commission/presscorner/detail/en/ip_10_...
>keep supply as low as they want and allow the AI companies to spend hand over fist
That's indistinguishable from not wanting to get burned by another semiconductor boom/bust cycle.
Anyone know what will lawfirms cut if they win from the lawsuit?
How many millions are we talking.
Just a reminder that anyone can file a lawsuit over anything, and the initial complaint is written by lawyers and reads with tabloid levels of sensationalism and allegation. The goal being to maximize the appearance of harm as much as possible so the suit has the greatest chance of sticking.
It is not in any way, shape, or form a ruling much less even a piece of well researched work. It's "my side of the story that makes me look perfect, with lawyers turning the heat up to 11"
In reality a lawsuit needs to have some basis. The bar might be low but judges frown on having their time wasted with a truly frivolous claim or a claim with no clear damages or harm shown.
That might have been true at some point in the past and the judiciary might even feel that way still.
However the practical penalty for filling absurd lawsuits is zero unless you do it repeatedly to random people for a decade.
Absolutely nothing bad will happen to the plaintiff or the lawyers representing them in this case.
Western legal systems are broken.
Honestly a lot of lawsuits feel like they just want the other party to pay them (settle) to make the lawsuit go away, even if there is no case; what are the repercussions for a frivolous lawsuit?
Realistically very few. The plaintiff carries the burden of proving the claim, but they are afforded the power of the law to do so via discovery. The defendant then has to spend a large number of legal resources refusing the claims and also providing the evidence required.
If the plaintiff loses the lawsuit a countersuit is pretty unlikely to succeed unless the lawyer participates in gross misconduct. Generally, countersuits are filed more to put the original plaintiff on the defense and don't result in a large judgement.
If you are operating in good faith, then you are pretty insular from kickback as a plaintiff
>what are the repercussions for a frivolous lawsuit?
None, hence the high price of liability baked into basically everything in America. And not just in nominal prices, but in terms of things like restricting access to spaces, restricting access to information, etc.
everything is political
Price fixing is legal as long as your are doing it in the open. In the UK it is called "price match" and eg. if supermarket says they keep prices matched to their competitor. No regulator raises an eyebrow.
So here Samsung and SK Hynix could say they price match to Micron and they are in the clear.
> Price fixing is legal as long as your are doing it in the open.
In the U.S., competitors are allowed to act in similar ways in response to economic realities, as long as they each arrive at that decision independently. But publicly anchoring your price to a competitor’s is potentially illegal.
> Price fixing is an agreement (written, verbal, orinferred from conduct) among competitors to raise, lower, maintain, or stabilize prices or price levels.
[Emphasis added]
https://www.ftc.gov/advice-guidance/competition-guidance/gui...
Identical prices are often a sign of intense competition. Every gas station on a corner has the same price because it's a highly competitive market not because of collusion. The prices of the much more lucrative chocolate bars inside the gas stations are less likely to be identical.
Yes, the gas station example is directly cited in the article I linked to. It’s legal for a gas station owner, with knowledge and consideration of a competitor’s price, to reduce their price to the same or just below. What is illegal is for nominally-competing gas station owners in an area to conspire to keep their prices within a range of each other’s, even without explicit agreement.
"Every gas station on a corner has the same price because it's a highly competitive market not because of collusion."
Huh? I can go to most any gas station-occupied intersection and you will always find two that match and one (usually a Persian-owned Chevron) which is consistently a dollar or more higher per gallon across all grades of fuel.
>Price fixing is legal as long as your are doing it in the open. In the UK it is called "price match" and eg. if supermarket says they keep prices matched to their competitor. No regulator raises an eyebrow.
No, the key term is "collusion", which could be done in the open or not. If a competitor told you they were unilaterally raising prices in secret, that would still be legal. Where you get into trouble is if you are cooperating to set prices. And no, this is all determined by a judge so cute workarounds like "I'm telling my competitors that I'm raising prices then gauging his body language" won't work.
I worked in a small town gas station as a kid. We always phoned our competitor to let them know when we were changing our prices, and they extended us the same courtesy. Usually we followed them, but sometimes we didn't.
Gas prices are posted on massive highly visible signs and are public information. This wasn't collusion, it was a sign of intense but friendly competition.
That is not what price fixing actually is. In the UK "price matching" is a one-to-many relationship meaning that the price of goods is set to the "lowest available"
Price fixing is a many-to-one all the manufactures agree to the highest prices they all agree on and set it there.
You are arguing semantics. Effect is the same.
Price match creates downward pressure, nobody is gonna sue over lower prices.
This is like if you showed a supermarket that their competitor's oranges were more expensive, and they "matched" by raising their prices for everyone.
That is your assumption without evidence. Once the targeted market knows others are tracking the price, it can up the price without fear competing market will undercut.
I think the only solution to this issue is China. If CXMT can supply, it will put all these monopolies in check.
That’s a nice short-term solution. How do you envision adding an infinitely-deep-pocketed state-sponsored supplier to the mix over the longer term?
What would the issue be, in the long term? I'm struggling to see a downside to more manufacturing capacity
cxmt is also selling their memory similar to the big three. No one hardware company in their right mind will sell their products not as high as possible after you learn how much harder hardware engineer and fab people work for.
I think it's incredibly unlikely to be deliberate price fixing this time. Demand is too high.
There is no reason for there not to be a price fixing. But the OpenAI's announcement from 2025-10-01 about buying 40% of supply, removes any need for collusion. It was a public signal for everyone to rise prices, one that each company could figure on their own. And it will be very hard to prove otherwise.
Two things can be true; they could theoretically make agreements to limit supply, like the OPEC does to exert control over international oil prices. But the OPEC is all above board and there's plenty of international competition on the oil market.
But the challenge is in proving it.
It isn't like OPEC and oil where you can turn on and off bumping out from the ground.
The most expensive foundry / fab isn't the leading edge, it is the "empty" ones. You can't have a fab sitting idle.
Prices skyrocketing create a perfect opportunity to slip in a little artificial bump and hope everyone blames the market. See also: egg prices in 2024.
if demand is too high, the market will adjust by adding supply. If Apple can't take the price of memory, why don't they make memory themselves. Oh no, the technical and manufacturing know-how is a barrier to entry and there is no talent available to make it. That is memory or semi companies' competitive advantages and their pricing power. It also takes Phd degrees to work at hardware companies. You can't have people doing leetcode for 2 months and todo apps and getting into the field and make 300k a year and hardware engineers wih Phds making 100k before this boom. It is long overdue for a rebalance of pricing power between hardware and software companies
,,The plaintiffs claimed the three companies reduced D-RAM supply under the pretext of transitioning to high-bandwidth memory (HBM). "The D-RAM oligopoly companies systematically coordinated the shift to HBM and the discontinuation of DDR3 and DDR4," they said. They added that Apple's recent sweeping product price increases were the trigger for the lawsuit.''
How can they do price fixing and discontinuing a product at the same time? It just looks like some companies are angry that AI / VC industry is outpricing them.
> An agreement to restrict production, sales, or output is just as illegal as direct price fixing, because reducing the supply of a product or service drives up its price. For example, the FTC challenged an agreement among competing oil importers to restrict the supply of lubricants by refusing to import or sell those products in Puerto Rico. The competitors were seeking to pressure the legislature to repeal an environmental deposit fee on lubricants, and warned of lubricant shortages and higher prices.
https://www.ftc.gov/advice-guidance/competition-guidance/gui...
But Micron didn't restrict the output: it stopped producing DDR2-3-4 completely, so it's not profiting from DDR customers at all.
hmm maybe the plaintiff should sue nvidia?
Looking forward to getting my $10 settlement 15 years from now like what happened with the DVD drive price fixing lawsuit (paid as a digital Visa gift card code so it can't actually be spent anywhere).