Three Companies Control 90% of DRAM. Now a Lawsuit Wants to Know Why RAM Costs 700% More
A new federal antitrust complaint accuses Samsung, SK Hynix, and Micron of using the AI memory boom as a business excuse to squeeze commodity DRAM supply. Seventeen plaintiffs, including individual buyers and two small computer-repair businesses, filed the case on June 25, 2026, in the U.S. District Court for the Northern District of California. The outcome matters well beyond gaming PCs: it will test whether three companies controlling nearly 90% of the world’s DRAM supply crossed from smart capacity planning into an illegal agreement.
Global memory prices have not moved so quickly in over a decade, and the demand driver behind the current spike differs from past cycles: hyperscale AI training and inference, not consumer device refreshes. Datafloq readers running IT budgets, cloud infrastructure, or hardware refresh cycles are already feeling the effect regardless of how the lawsuit turns out.
The Complaint: An AI Alibi for Commodity Scarcity
Garciaguirre et al. v. Samsung Electronics Co., Ltd. et al., docketed in the Northern District of California as case number 5:2026cv06345, names Samsung Electronics, Samsung Semiconductor, SK Hynix, SK Hynix America, and Micron Technology as defendants. The case was initially assigned to Magistrate Judge Nathanael M. Cousins, then reassigned on June 26, 2026, to District Judge Noel Wise, with discovery referred to Magistrate Judge Virginia K. DeMarchi. The plaintiff group includes individual PC buyers alongside small businesses such as Troy’s Computers LLC and a Florida repair shop operating as My Florida PC, who say they overpaid for DRAM-based products.
The complaint invokes Section 1 of the Sherman Act and alleges the three suppliers coordinated cuts to conventional DRAM, including DDR3 and DDR4 modules, a shift plaintiffs date to 2022. Plaintiffs argue the companies shifted capacity toward high-bandwidth memory, the specialized DRAM feeding AI accelerators, and used growing AI demand as public cover for tightening commodity supply. The filing puts the resulting increase in conventional, or commodity, DRAM prices at roughly 700% over four years, a figure circulating widely across financial and trade press coverage of the complaint but not yet tied to a specific, confirmed price benchmark. The filing states plainly, “Since 2022, these firms have fixed supply and prices for DRAM.” Plaintiffs reportedly seek class certification, injunctive relief, and treble damages, based on press summaries of the filing rather than a public complaint text.
Micron told Investors Business Daily, “We compete vigorously, fairly and in compliance with all applicable laws.” The company has denied the allegations and said it will defend itself against the suit. Samsung and SK Hynix have not issued public statements addressing the case, based on open-source checks.
DRAM Has Been Here Before
The lawsuit lands against a well-documented history of DRAM cartel behavior. The U.S. Department of Justice ran a criminal DRAM price-fixing investigation covering conduct from July 1998 through June 2002, and the fallout produced real prison time. Samsung agreed to a $300 million criminal fine in 2005, and Hynix paid $185 million the same year. Four Hynix executives pleaded guilty and served five to eight months in prison, and a Samsung sales manager received an eight-month sentence. The DOJ’s broader DRAM investigation eventually produced more than $731 million in criminal fines, a total covering Infineon and Elpida, which separately paid an $84 million fine, alongside Samsung and Hynix. Micron cooperated with investigators at the time and avoided a corporate criminal fine. A parallel 2000s civil case, In re Dynamic Random Access Memory Antitrust Litigation, settled for $406 million and won class certification in the same Northern District of California courthouse now hosting the new complaint.
The industry has a criminal record. Plaintiffs know it, and so will the judge.
A more recent precedent cuts the other way. In March 2022, the Ninth Circuit affirmed dismissal of a separate DRAM indirect-purchaser case against the same three companies, a ruling which shapes how the new complaint will likely be tested at the pleading stage.
The Numbers Support Two Different Stories
Every fact plaintiffs cite as evidence of coordination also has an innocent explanation, and the market data shows why. TrendForce puts Samsung, SK Hynix, and Micron at a combined 89.7% of Q1 2026 DRAM revenue, split between Samsung’s 38.5%, SK Hynix’s 28.8%, and Micron’s 22.4%, a concentration level which makes any shared capacity decision look coordinated even when it isn’t. TrendForce also forecast conventional DRAM contract prices would rise 90 to 95% quarter over quarter in Q1 2026, with PC DRAM prices expected to at least double. IDC’s research points to the same underlying driver: AI data centers and hyperscalers, including Microsoft, Google, Meta, and Amazon, are pulling a disproportionate share of global memory capacity. The shift is pushing manufacturers to reallocate wafer production toward HBM and high-capacity DDR5. IDC frames the change as a structural reallocation rather than a normal cyclical shortage, projecting 2026 DRAM supply growth of just 16% year over year and NAND supply growth of 17%, each below historical norms. China’s CXMT holds roughly 8% share, per Counterpoint, nowhere near enough to offer buyers a real alternative anytime soon.
The Wall Street Journal has reported data centers are on pace to consume more than 70% of 2026 production of high-end memory chips, the premium HBM and high-capacity DDR5 products AI infrastructure depends on. IDC’s public research backs the broader reallocation narrative, AI buyers pulling a disproportionate share of premium memory capacity, but IDC does not publicly confirm the 70% figure itself, and the number applies to high-end chips rather than memory output overall. Micron’s numbers support the demand story regardless: the company disclosed $22 billion in customer supply commitments and projected tight memory supply through at least 2027.
Defendants also have a physics problem working in their favor. DRAM fabs require years of lead time, enormous capital, and lengthy customer-qualification cycles, so even a supplier eager to expand commodity output cannot flip a switch and add capacity within a single product cycle. Plaintiffs will need to explain why three independent companies made the same capacity bet at the same time for reasons other than shared incentive economics.
Why Parallel Conduct Alone Won’t Win the Case
Section 1 of the Sherman Act targets agreements, not results, and the distinction will decide the case. The Ninth Circuit’s 2022 opinion held plaintiffs relying on parallel conduct must plead what the court called “some further factual enhancement” showing an actual agreement, and the panel found the earlier DRAM plaintiffs came up short on the same standard. Plaintiffs in the new case face the same bar: shared market concentration, simultaneous capacity shifts toward HBM, and a history of prior cartel conduct describe a pattern, not proof of a deal. To survive a motion to dismiss, plaintiffs likely need something closer to internal communications, customer-allocation discussions, or synchronized supply commitments not explained by independent business logic. Plaintiffs able to point to trade-association meetings, internal memos discussing rivals’ output, or customer-allocation coordination stand a far better chance than plaintiffs pointing only to matching production curves. DRAM’s cartel history gives the new complaint narrative weight, but courts generally treat past misconduct as context rather than evidence of a current conspiracy.
What Remains Unconfirmed
A few important facts remain unconfirmed. The full complaint has not surfaced through open court-record searches, so the exact pricing benchmark behind the 700% figure and the full scope of relief requested trace back to press summaries rather than the filing itself. No U.S., European, South Korean, Chinese, or Japanese regulator has announced a parallel investigation into current DRAM pricing, based on publicly available records so far. Datafloq will update coverage as the docket develops and as Samsung and SK Hynix respond.
What Enterprise Buyers Should Watch Next
The lawsuit will not resolve the deeper question facing procurement teams: has memory permanently become a strategic input rather than a background commodity? My take: it already has, regardless of how the case turns out. Three suppliers controlling nearly 90% of DRAM revenue means every capacity decision they make, lawful or not, ripples through hardware refresh budgets, AI PC rollouts, server upgrades, and edge deployments. If the case survives a motion to dismiss, discovery could expose internal communications, customer-allocation records, and pricing-strategy documents capable of reshaping how buyers negotiate supply contracts going forward. Even without a liability finding, prolonged litigation tends to change supplier behavior and contract language. A trial remains years away, and large antitrust class actions usually settle rather than reach verdict.
Procurement teams reading the docket should treat memory the way they already treat GPUs: as a constrained input worth locking into multi-year contracts rather than a line item to shop on price each quarter. Buyers who wait for the case to resolve before adjusting their sourcing strategy will spend the next two years negotiating from a weaker position, since Micron’s $22 billion in customer commitments shows large buyers are already securing allocation ahead of everyone else.
The case will not settle whether AI caused the memory shortage. It will settle whether Samsung, SK Hynix, and Micron used the shortage as cover for something courts consider illegal. Either way, enterprise buyers who still treat RAM as an afterthought in procurement planning are already behind.
