Let's assume there are only three bakeries supplying bread to a town. Coincidentally, as more young people in this town got married, orders for wedding cakes poured in, and all three bakeries switched their ovens to wedding cake production to meet demand. Then everyday loaves quickly became scarce, and prices jumped as much as tenfold. Legal experts explain the logic of the DRAM price-fixing lawsuit filed against Samsung Electronics, SK hynix, and Micron with this analogy, referring to the complaint submitted to the U.S. District Court for the Northern District of California on the 25th of last month. If the marriage boom is the artificial intelligence (AI) boom, the wedding cake corresponds to high-bandwidth memory (HBM) used in AI servers.
However, the prevailing view among legal circles at home and abroad is that it is hard to see this situation as "collusion" right away. If the three companies each judged there was no reason to insist on baking loaves when wedding cake orders were high and made similar choices, there is no legal basis to challenge that. In 2007, through the Twombly ruling (Bell Atlantic v. Twombly), the U.S. Supreme Court set a standard that for a plaintiff to win a collusion suit, it is not enough to show merely that "prices rose similarly," and that from the filing stage the "plausibility" of an agreement must be specifically supported. If this standard is not met, the case ends at the motion to dismiss stage before trial even begins.
Therefore, in actual litigation, it is difficult to raise suspicions of collusion based solely on the fact that all companies raised prices together. More direct evidence of collusion is needed. For example, indications that competitors exchanged sensitive information such as prices or production volumes, or that there were actual opportunities to coordinate prices through industry associations or gatherings, are typical. Without such clear indications, if each party judged "wedding cakes are more profitable, so I'll run my oven too," it is not subject to punishment.
◇ "If there is no proof of 'coordination among firms,' collusion is hard to prove"
U.S. courts call this phenomenon "conscious parallelism." Literally translated into Korean it may feel difficult, but put simply, it means "without evidence of an agreement, you don't punish based only on the outcome."
There have been similar lawsuits in the past. When DRAM prices surged in 2016–2017, the law firm Hagens Berman brought a similar suit against Samsung Electronics, SK hynix, and Micron, but the case was dismissed—first at the trial court in 2020 and then affirmed by the 9th U.S. Circuit Court of Appeals in Mar. 2022—stating that "the defendants' conduct is better explained by lawful parallel conduct than by collusion." The circumstances presented by the plaintiffs did not even clear the courthouse threshold.
The recent suit was filed by 14 U.S. consumers and three small PC assembly and distribution companies on the 25th of last month in the U.S. District Court for the Northern District of California. The plaintiffs claim that Samsung Electronics, SK hynix, and Micron, which together hold about 90% of the global DRAM market, used the expansion of HBM production as a pretext to simultaneously and widely cut production of legacy DRAM such as DDR3 and DDR4. As a result, the core of the complaint says, general-purpose DRAM prices have surged about 700% over the past four years. The complaint dubbed this situation "RAMpocalypse." The plaintiffs also said Apple's recent price hikes for iPads and Macs are part of this fallout.
Of course, a past victory by memory semiconductor corporations does not guarantee an unconditional win here. In 2005, Samsung Electronics and Hynix Semiconductor, the predecessor of SK hynix, were caught actually fixing DRAM prices from 1999 to 2002 and were fined $300 million and $185 million, respectively, by the Ministry of Justice, and some executives at the time served prison terms. However, that case differs from this one in that there was direct evidence—such as emails—proving an actual agreement.
◇ Will the plaintiffs' "HBM-as-excuse" claim clear the courthouse threshold?
Industry watchers see the skeleton of this suit as almost identical to the 2018 case. As noted earlier, it is not enough that "prices rose together"; more explicit and direct evidence that the three companies actually coordinated is needed to pass the court's threshold. The new card the plaintiffs have played is the theme of switching to HBM, and whether that can be recognized as such direct evidence is effectively the only issue in this case. A semiconductor attorney said, "The 2018 case was dismissed because the plaintiffs failed to present evidence beyond the circumstance that 'everyone acted similarly,' and this time as well, if all they show is that the timing of the HBM switch happened to overlap, they may hit the same wall."
The corporations' rebuttals are also strong. Because there is a clear market factor—an actual surge in demand from AI servers—the argument that "prices rose not due to collusion but because real demand exploded" could be even more persuasive than in 2018, observers say. There are also signs that bolster this. Recently, Micron announced a large-scale expansion plan in Hiroshima, Japan; SK hynix did so in Yongin; and Samsung Electronics is also continuing to invest. If the three companies had conspired to tighten supply, there would be no reason to spend so much money now to increase capacity.
If this case clears the motion-to-dismiss threshold and proceeds to the discovery stage, communications inside the three companies about production plans could be disclosed in court. However, considering that the 2018 case failed to clear both the trial and appellate thresholds on precisely the same logic and ended at an early stage, this suit also appears more likely to conclude early without reaching a full evidentiary fight. Ultimately, the key is whether the plaintiffs' "HBM card" is special enough to produce a different outcome than before, but judging from the trajectory of precedents so far, legal circles broadly see that possibility as low.