Judge Keeps DRAM Cartel Lawsuit Alive Against Memory Giants

11 min read
1 views
Nov 16, 2026

A federal judge just refused to toss the DRAM cartel case against Samsung, SK Hynix and Micron. The Big Three now face discovery over claims they choked conventional memory supply while prices soared nearly 700 percent. What happens next could reshape the market.

Financial market analysis from 16/11/2026. Market conditions may have changed since publication.

What if the reason your next laptop or phone costs more has less to do with pure market forces and more to do with three companies deciding, in near perfect unison, not to make enough memory chips? That question stopped being theoretical this past weekend. A federal judge in San Jose just refused to throw out a major antitrust case against the world’s dominant DRAM producers, and the decision changes the entire conversation around soaring memory prices.

Judge Rejects Dismissal Bid In High Stakes Memory Case

For more than a year the only players who seemed able to influence conventional DRAM prices were the three companies that control the vast majority of global supply. That changed with a court order dated November 14. Judge Noël Wise of the Northern District of California denied in full the joint motion by Samsung, SK Hynix and Micron to dismiss a proposed class action accusing them of conspiring to restrict the supply of everyday memory chips.

The ruling does not declare anyone guilty. It does something almost as consequential at this stage: it finds the complaint plausibly alleges an agreement that could violate Section 1 of the Sherman Act. The companies must now answer the complaint by December 7 and prepare for discovery. Emails, internal presentations, chat logs and the paper trail behind certain customer questionnaires are about to become fair game.

I’ve followed semiconductor pricing cycles for years, and this one has felt different. Prices for conventional DRAM climbed roughly 697 percent over eighteen months while new capacity for that product category barely appeared. When the same three firms move in near lockstep for that long, courts start asking harder questions.

Why This Case Looks Different From The Last One

Defendants argued the lawsuit was essentially a rerun of an earlier case covering the 2016-2018 price spike. That earlier action was dismissed, and the appeals court later affirmed the decision, finding that simultaneous output cuts by rivals in a concentrated market could simply reflect ordinary oligopoly behavior. This time the judge saw important distinctions.

At the hearing, defense counsel leaned on the classic observation that competitors in concentrated industries naturally watch one another closely. The judge’s reply cut through the legal language. What the plaintiffs describe, she said, looks “more akin to hawks flying in formation.” Hawks do not normally fly in formation. The point landed. Watching a rival is legal. Moving in coordinated steps for years while prices multiply requires a better explanation than pure coincidence.

The complaint was filed in late June by seventeen plaintiffs from five states. Fourteen are ordinary consumers. Three are small PC builders. They seek class certification, an injunction and treble damages on behalf of anyone who bought a device containing commodity DRAM. In other words, a very large group of people.

The Timeline That Raised Eyebrows

Parallel conduct alone rarely proves a conspiracy. Plaintiffs therefore focused on what antitrust lawyers call “plus factors” — actions that make little sense unless a company believes its rivals will behave the same way. The chronology laid out in the complaint is striking.

  • Late 2022 into early 2023: SK Hynix and Micron announce production cuts within roughly three weeks of each other. Samsung follows later in 2023, reversing years of prior practice in which the market leader often expanded while others pulled back.
  • After those cuts: All three shift substantial capacity toward high-bandwidth memory and lock much of that capacity under multi-year contracts with AI customers.
  • April to June 2025: All three announce the wind-down of mainstream DDR4 production inside a single quarter.
  • December 2025: Micron states it will shut down its long-running consumer brand Crucial.
  • January 2026: All three introduce customer-vetting questionnaires that are nearly identical in wording and structure.

Throughout this period, conventional DRAM prices rose almost sevenfold and meaningful new capacity for that product category did not materialize. New fabrication projects were delayed or halted even as the profit opportunity grew enormous.

The judge did not accept every allegation. She set aside claims involving DDR3 because one of the companies continued producing it. Investor call commentary and trade conference meetings received little weight. The oligopoly structure of the market itself was treated as neutral. The pure economic logic of pivoting to higher-margin high-bandwidth memory also did not, standing alone, prove coordination.

What survived is more interesting. Capacity decisions that cannot be easily reversed. The absence of long-term contracts for conventional buyers while AI customers received them. The simultaneous exits from certain product lines. And especially those nearly identical questionnaires rolled out years after the AI demand surge began. The court found the questionnaire allegations strongly support a plausible inference of conspiracy.

Capacity Decisions That Cannot Be Easily Unwound

Building a modern memory fab is a multi-year, multi-billion-dollar commitment. Once you decide not to expand conventional DRAM capacity, reversing that choice takes time. Plaintiffs argue that when prices offered an unprecedented incentive to add supply, none of the three companies meaningfully expanded conventional output. Two of them allegedly delayed or stopped new fab projects. Courts have long treated irreversible capacity decisions as the kind of move companies hesitate to make without some degree of comfort about rivals’ plans.

In my view this is one of the strongest points in the complaint. Markets usually respond to sustained high prices with new supply. When that response fails to appear for years in a concentrated industry, the burden of explanation shifts. The companies will have their chance to explain. Discovery will test those explanations against internal documents.

Long Term Contracts For Some, None For Others

All three producers signed multi-year deals for high-bandwidth memory. At the same time, reports indicate Samsung and SK Hynix declined long-term agreements with large conventional DRAM buyers. Micron later stated it had signed a number of long-term supply agreements spanning data center, consumer and automotive customers, which complicates the picture. Still, the contrast between the treatment of AI customers and everyone else remains part of the factual landscape the court must consider.

Nothing illustrates the desperation of buyers better than reports that major Chinese technology firms approached the Korean producers seeking ten-year server DRAM contracts. When sophisticated customers beg to lock in supply for a decade, the market is not functioning like a normal competitive arena.

The Questionnaire That Stood Out

Perhaps the single most awkward allegation involves the customer questionnaires. In January 2026, well after the AI boom was underway, all three companies rolled out vetting questions that were strikingly similar. The court found these allegations particularly supportive of a conspiracy inference. Parallel product decisions can sometimes be explained by shared market conditions. Parallel administrative forms issued years into a demand surge are harder to dismiss as coincidence.

This detail is the kind of concrete, low-level evidence that often proves more damaging in discovery than high-level strategy documents. Lawyers on both sides know it. That is why the defense fought so hard to keep the case from reaching this stage.

Why Memory Prices Became A Macro Issue

If this dispute only affected gamers buying RAM sticks, it would remain a niche technology story. It became a broader economic concern when analysts began linking memory, software and electricity costs to core inflation measures. One major bank estimated that AI-related increases in these categories were already adding more than two-tenths of a percent to core PCE and could reach half a percentage point by year-end. The memory-exposed portion of the basket was projected to contribute meaningfully in the current quarter.

Central bankers can raise interest rates. They cannot print DRAM wafers. When a concentrated group of suppliers accounts for more than ninety percent of global DRAM revenue and appears to have decided not to expand conventional capacity, the practical constraint on certain components of inflation sits in factories in South Korea and Idaho rather than in Washington.

Recent trade data only sharpened the picture. Korean DRAM exports hit new records while year-over-year growth rates remained extraordinary for months on end. Contract prices rose sharply through successive quarters. The rate of increase has slowed, yet the absolute level remains dramatically higher than two years ago. On simple math the product has roughly tripled in less than a year even after the deceleration.

The Human And Corporate Cost

Downstream effects are visible across the technology sector. PC unit forecasts have been revised lower. Smartphone shipments face similar pressure. Certain gaming hardware sales have plunged. Small system builders, the kind of companies represented among the plaintiffs, face margin compression or worse when memory costs explode while end-customer willingness to pay has limits.

I’ve spoken with people who assemble custom machines for a living. The sudden inability to source conventional modules at predictable prices forced some to pause operations or raise quotes so high that orders dried up. Those stories rarely make national headlines, but they illustrate why class actions of this type attract plaintiffs.

What Analysts Are Saying About Future Supply

Independent research houses and sell-side analysts largely agree that meaningful new conventional DRAM capacity is unlikely before 2028. After hosting industry specialists, one major firm reiterated a constructive pricing outlook through 2027 and into the first half of 2028 while maintaining positive ratings on the two Korean producers. The companies will interpret that view as confirmation that fabs take years to build and clean-room space is finite. Plaintiffs will ask why ground was not broken in 2024 when prices first turned sharply higher and each firm supposedly had every incentive to capture share from the others.

Both readings contain elements of truth. Fabrication plants are extraordinarily complex. At the same time, the decision not to expand a product category generating record margins is not forced by physics. It is a choice. Discovery will explore how those choices were made and whether they were made independently.

Stock Market Reaction And The Over-Earning Debate

Equity markets have already rendered a partial verdict. Memory stocks trade at low single-digit forward earnings multiples while the products they sell have multiplied in price. Valuations near tangible book value coexist with still-rising spot prices and guidance for favorable supply-demand dynamics through 2027 and beyond. The market’s shorthand for this tension is “over-earning.” A plaintiff’s lawyer uses a different word: damages.

The debate is genuine. High prices may simply reflect a temporary mismatch between explosive AI demand for advanced memory and the multi-year lead times required to reallocate or expand capacity. They may also reflect coordinated restraint. The court has now decided that the latter theory is plausible enough to proceed. That decision alone introduces a new variable into valuation models that previously treated the profit surge as purely cyclical.

What Comes Next In The Legal Process

Answers to the complaint are due in early December. A case management conference is scheduled for mid-January. After that the parties enter the phase defendants most wanted to avoid. Document production and depositions will test the narrative that every decision was independently driven by AI demand and ordinary profit-seeking.

History offers a cautionary note. The earlier DRAM investigation that covered the late 1990s and early 2000s produced guilty pleas, prison sentences for executives, and hundreds of millions of dollars in fines. One company received immunity after cooperating. The documents, not the public announcements, ultimately proved decisive. That pattern could repeat.

Expectations should still remain measured. Surviving a motion to dismiss is a far cry from winning at trial or on appeal. The same appellate court that previously rejected a similar theory will eventually review any final judgment. Even a complete plaintiff victory would not lower prices in time for this year’s holiday shopping season or the next several quarters of inflation data. Antitrust litigation moves on a multi-year calendar.

The Real Supply Response May Come From Outside The Courtroom

The most immediate relief for buyers is likely to arrive from a source no U.S. court ordered. Chinese producers are expanding output. One domestic player is expected by some analysts to reach roughly half the volume of a major Korean competitor by 2028. Other industry participants are exploring memory-related investments. In commodity markets, sustained extreme prices eventually attract new supply even when incumbents prefer otherwise.

That dynamic does not resolve the legal questions. It does suggest that the period of extreme scarcity for conventional DRAM may eventually ease for reasons unrelated to this lawsuit. Markets have a way of solving shortages that courts cannot match in speed.

The Broader Lesson About Concentrated Markets

This case sits at the intersection of technology, antitrust enforcement and macroeconomic policy. When three firms control more than ninety percent of a critical input, ordinary competitive pressure can weaken. Parallel behavior becomes easier to sustain and harder to distinguish from collusion. Courts then face the difficult task of drawing lines between lawful oligopoly and unlawful agreement.

Judge Wise’s order does not resolve that tension. It simply refuses to resolve it at the pleading stage. The companies will have a full opportunity to demonstrate that every capacity decision, product exit and customer form was independently rational. Plaintiffs will try to show the opposite. Documents that never saw the light of day will decide which story holds up.

For investors the development adds uncertainty to an already volatile sector. Stocks that look cheap on current earnings may face different scrutiny if discovery uncovers evidence of coordination. For consumers and device makers the ruling offers no immediate price relief, only the possibility that future capacity decisions will be made with greater awareness of legal risk.

The next twelve months of discovery will reveal far more than public earnings calls ever could. Whether the three largest memory producers were simply responding to the same powerful demand signals or something closer to formation flying remains an open question. The court has decided the question is worth asking under oath. That decision alone marks a meaningful shift in how this price surge will be remembered.


In the end the DRAM market may still be driven primarily by the extraordinary demands of artificial intelligence infrastructure. High-bandwidth memory absorbs wafer capacity that once served conventional products. Lead times for new plants stretch measured in years rather than quarters. Price-insensitive hyperscale buyers willing to sign long-term contracts change the incentive structure for producers. All of those factors can generate elevated prices without any formal agreement.

Yet the combination of simultaneous production cuts, synchronized product exits, the absence of conventional capacity expansion despite record incentives, and nearly identical customer questionnaires creates a factual pattern that a federal judge found sufficiently suggestive of coordination to allow the case to proceed. That pattern now moves from the realm of market commentary into the more rigorous environment of discovery and sworn testimony.

Anyone who follows semiconductor cycles knows that periods of extreme pricing power eventually end. New entrants appear. Existing players eventually expand. Technology shifts. The open question is whether this particular episode will also produce lasting changes in how the industry’s largest participants make capacity decisions and how regulators and courts examine those decisions. The answer will emerge only after the documents come out.

For now the practical reality remains straightforward. Conventional DRAM stays expensive. Device makers continue to absorb or pass on higher costs. Consumers feel the impact in higher prices for computers, phones and other electronics. And three companies that together dominate global supply face the prospect of explaining their parallel decisions in a courtroom rather than solely on earnings calls. That is a new development in a story that has already reshaped parts of the technology supply chain and touched broader inflation measures. The formation, if there was one, is about to face closer inspection than ever before.

❝
If you cannot control your emotions, you cannot control your money.
— Warren Buffett
Author

Steven Soarez passionately shares his financial expertise to help everyone better understand and master investing. Contact us for collaboration opportunities or sponsored article inquiries.

Related Articles

?>