Part of the ongoing Big Tech's War on Users series.
A class action lawsuit was filed in California on June 25th against Samsung, SK Hynix, and Micron, the three companies that dominate global DRAM production. Counterpoint Research puts their Q1 2026 revenue share at Samsung 38%, SK Hynix 29%, and Micron 22% — roughly 90% combined. The core allegation: that they coordinated to slash conventional DDR3 and DDR4 supply while pivoting production toward High Bandwidth Memory, the stacked memory that AI datacenters are consuming at a scale that would have seemed fictional three years ago. As a result, consumer memory prices have risen something like 700% over four years. Consoles cost more. Laptops cost more. Phones cost more. Apple has already raised the price of its cheapest MacBook Pro by $400, to $1,999, citing memory and storage costs it could no longer absorb. Xbox raised prices. Valve announced that their Steam Machine pricing target was no longer viable. Corsair started putting tamper-evident seals on RAM packaging because scammers are now apparently worth worrying about for memory sticks.
The frustration driving this lawsuit is completely understandable. I feel it too.
The lawsuit itself, though, I think is going to have a hard time. Micron has already denied the allegations and said it intends to defend itself.
This Isn't a "Screw you to Customers" Situation
Here's the thing that gets lost when prices go this sideways: sometimes markets do genuinely terrible things to consumers without anyone in a boardroom actually deciding to screw you. This is one of those situations, and it matters for understanding why the lawsuit is an uphill climb.
Samsung, SK Hynix, and Micron are publicly traded companies. They have shareholders, boards, and quarterly earnings calls. They've collectively shifted roughly 80% of their manufacturing capacity toward HBM — the memory type that AI accelerators and datacenter GPUs require. Micron went far enough that they shut down Crucial, their consumer DRAM brand, at what the lawsuit itself describes as "the most profitable price point in its history."
The complaint calls this "contrary to all economic and business logic." I'd argue the opposite. HBM margins relative to commodity DDR4 are apparently so substantial that Crucial at peak profitability still couldn't compete with the opportunity cost of producing it. That's not a conspiracy — that's a capital allocation decision any CFO with a pulse would make when the AI buildout is buying up memory at a scale Sam Altman once described with a $1.4 trillion letter of intent. (That number got quietly revised down, but the direction of it tells you everything about the demand pressure these companies are sitting inside.) The boards of these companies would be facing shareholder revolts — possibly derivative suits of their own — if they hadn't chased those margins.
The problem for consumers is real. The mechanism isn't malice. It's that the AI buildout is simply a larger and more immediate customer than we are, and in an oligopoly with three players and no meaningful path for new competitors to enter, there's nothing to check that.
What Actual DRAM Price-Fixing Looks Like
The lawsuit leans heavily on history, and the history is real — so it's worth being precise about what actually happened, because the 2005 case looks very different from the current situation on the mechanics that matter in court.
The DRAM price-fixing conspiracy that the DOJ prosecuted ran from 1998 to 2002 and was a genuine international cartel. Five companies — Samsung, Hynix, Micron, Infineon, and Elpida — pleaded guilty to explicit coordination. Executives from competing companies communicating across company lines to fix prices. Documented agreements. An international conspiracy the DOJ could point to with actual evidence. When the grand jury subpoenas hit, a Micron regional sales manager was charged with obstruction of justice for destroying and altering documents — which tells you something about what those documents contained. Samsung paid $300 million. Hynix paid $185 million. Infineon paid $160 million, which was the third-largest antitrust fine in US history at the time. European regulators separately went after nine manufacturers for another €331 million in 2010.
Micron paid nothing and faced no criminal charges — because they blew the whistle on the cartel and cooperated with prosecutors. They were participants who turned government witness. The Samsung executive who pleaded guilty and did eight months in federal prison was later promoted to President of Samsung Germany, then President of Samsung Europe, which is a whole separate conversation about how seriously these companies internalize lessons learned, but that's a tangent.
The point is: that case had evidence. Actual cross-company coordination. Document destruction. Executives talking to each other about fixing prices. Someone deciding cooperation with the DOJ was a better deal than going down with the ship. That's what Section 1 of the Sherman Act is built to prosecute — an actual agreement, not three companies reading the same demand signals and making the same obvious call.
Some of you are probably thinking this sounds familiar. That's probably because your employer makes you sit through yearly compliance training on anticompetitive practices — the "do not do this, do not discuss this with competitors, report it immediately if you see it" slide deck that everyone clicks through before the holiday break. Price fixing, bid rigging, market allocation. The training exists because the conduct is real, it happens, and companies have learned through very expensive experience that it needs to be explicitly prohibited in writing. The 2005 case is essentially a textbook example of exactly what that training is describing. Executives from competing companies, coordinating. That's the thing.
What's happening now looks like price-fixing from the outside because three dominant players all did the same thing at the same time and prices went vertical. But "looks like" and "is" are doing very different legal work here, and the distinction is exactly what that compliance training is trying to help you recognize too — parallel conduct isn't coordination, even when the outcome is indistinguishable from the consumer end.
The 2018 Preview
Here's the part of the history the current lawsuit doesn't highlight: we've been here before, more recently.
In 2018, Hagens Berman filed a near-identical class action against Samsung, Hynix, and Micron, alleging that the same trio engaged in price-fixing that caused DRAM prices to nearly triple between 2016 and 2018. Same defendants. Same market structure. Same theory.
The district court dismissed it.
The legal challenge in both 2018 and now is the same: proving explicit coordination versus parallel conduct. In an oligopoly with three dominant players, the fact that they all made the same decision at the same time is not evidence of a conspiracy — it's what you'd expect from three rational actors watching the same market. For this lawsuit to go differently, plaintiffs would need something the 2018 case didn't have — internal communications, coordinated messaging, some artifact of actual agreement rather than parallel market behavior. Investment bank Jefferies is currently projecting DRAM prices up another 40-50% in Q3 and 30-40% in Q4 without meaningful relief before 2028. If there's a smoking gun buried in discovery, that timeline gets complicated fast. If there isn't, this is a sympathetic case with bad odds.
The Actual Problem
The frustration is real and the prices are genuinely brutal. But what's happening isn't three companies colluding to extract money from consumers — it's an oligopoly with no meaningful competition responding rationally to a massive external demand shock, in a market where the barriers to entry are so high that nobody can step in to undercut them when they do.
That's not a Sherman Act problem. That's a "we let three companies corner 90% of a critical input for modern computing and then AI happened" problem. And no class action lawsuit is going to fix it, because the fix would require either antitrust structural remedies that don't exist yet, domestic fab capacity that doesn't exist yet, or the AI buildout to cool off enough that HBM margins stop being the most attractive place to put manufacturing resources.
None of those happen in discovery.
Find me on Mastodon at @ppb1701@ppb.social
Part of the ongoing Big Tech's War on Users series.