The Great RAM Heist of 2026 (Maybe)

So RAM prices went up like 700% over four years and now everyone’s suing Samsung, SK Hynix, and Micron for supposedly colluding to make it happen. I’ve been reading through this thing for a few days now and honestly, I have thoughts. Not “expert” thoughts, just, you know, a guy who watched RAM prices and thought “wait what.”

Quick background, because it matters and I don’t want this to just be vibes: there’s an actual case now, filed June 25, 2026, in the Northern District of California — Garciaguirre v. Samsung Electronics, case number 3:26-cv-06345, assigned to Judge Noel Wise. 17 plaintiffs, mostly regular consumers plus a few small PC-building businesses, suing under Section 1 of the Sherman Act (the “you’re not allowed to conspire to restrain trade” part of antitrust law). So paper got filed. Real paper. Whether it holds up in front of a judge is a completely different question, and honestly that’s kind of the whole point of me writing this.

The core claim, once you strip the legal padding off it: Samsung, SK Hynix, and Micron control something like 90% of the world’s DRAM (roughly 38%, 29%, and 22% respectively as of Q1 2026, if you want the actual breakdown), and instead of competing normally when AI demand exploded, they all just… quietly did the same thing. DDR3, DDR4, the boring stuff that goes in your laptop and gaming rig, got scarce. Meanwhile HBM — the memory AI data centers are desperate for — got prioritized, because it’s way more profitable per wafer than regular consumer memory. Prices for normal DRAM apparently jumped 700% in four years. That’s not a typo. And weirdly, nobody undercut anybody, which if you know how markets are supposed to behave, is the weird part.

Why did the internet immediately jump to “they’re guilty”? Simple: these companies have actually colluded before. Samsung and SK Hynix (well, Hynix at the time) both pled guilty to price-fixing back in the early 2000s. Real fines, real executives doing real prison time, over $700 million total between the settlements once you add it up. So when prices spiked again with the same three names attached, people’s reaction wasn’t “huh, interesting,” it was “oh, here we go again.” I get it. If your neighbor got caught stealing your newspaper once and now your newspaper’s missing again, you’re allowed to be suspicious. Doesn’t mean he did it this time.

Here’s what I think is the strongest stuff on the “yeah, something fishy is going on” side. Three companies, that’s it — you don’t need some sprawling conspiracy with fifty players, you need three companies watching each other and nobody blinking first. Nobody “cheated,” which economists find genuinely weird — normally in a shortage some company goes “screw it, I’ll produce more and eat everyone’s lunch,” and that didn’t happen here, at least not that anyone’s pointed to yet. The messaging is suspiciously synced too: all three companies keep using nearly identical phrases, “capacity discipline,” “value over volume,” stuff like that. Not illegal to say the same buzzwords. Still makes you go hmm. And there’s the history, obviously, which I already covered but it bears repeating (yes, I did that on purpose).

There’s also this detail I keep coming back to: the complaint points to Apple raising the price of its cheapest MacBook Pro by $400, citing memory costs. That’s not proof of collusion at all, to be clear. But it does show the squeeze is real and it’s hitting regular people, not just PC builders complaining about DDR5 kits on Reddit.

Okay but flip side. HBM is legitimately, wildly more profitable than regular DRAM. If DDR5 nets you ten bucks a chip and HBM nets you a hundred, you don’t need a secret meeting to figure out which one you’re gonna make more of, that’s just math. Every company on earth would independently make that switch. Also — and people underestimate this constantly — you cannot just flip a switch and produce more RAM. Fabs take years to build and cost billions. Wafer capacity is finite and reallocating it is slow and expensive, not a “just make more” button.

There’s a genuinely boring economic explanation for the “nobody cheated” thing too. It’s called tacit coordination and it’s not illegal, usually. If Samsung isn’t expanding, and SK Hynix notices and thinks “well I’m not gonna be the one flooding the market either,” that’s not a phone call, that’s reading the room. Three airlines all raising ticket prices when fuel spikes isn’t a cartel, it’s just shared incentives producing similar outcomes on their own. Same logic could apply here. And post-2000s, these companies know regulators are watching them like hawks — an overt cartel today is way riskier than it was twenty years ago, and getting caught again would probably wreck all three of them. That’s a real disincentive, not a small one.

So did they do it? I genuinely don’t know, and nobody outside the actual discovery process knows either.

My opinion, for what it’s worth: there’s enough here to justify the lawsuit existing and being taken seriously. Market concentration plus prior convictions equals reasonable scrutiny. But “reasonable to investigate” and “probably guilty” are very different bars, and I think a lot of people online are treating this like a done deal when it very much isn’t. What’ll actually decide this is discovery — emails, internal memos, meeting notes. If there’s a paper trail showing coordination, this case gets serious fast. If discovery turns up nothing beyond “we all independently noticed AI money was better than PC money,” the companies’ defense gets a lot stronger. Worth mentioning: one analysis I read pointed out a similar prior case actually failed because it lacked direct evidence like emails or an actual agreement, parallel conduct alone wasn’t enough. Which is a useful reminder that “looks suspicious” and “is illegal” aren’t the same legal standard, even if they feel identical when you’re standing in a store looking at a $300 RAM kit that used to cost forty bucks.

If I had to bet, and I want to be clear I’m not a lawyer, this is just my gut: the truth is probably somewhere annoyingly in the middle. AI demand is real, it genuinely exploded, that part isn’t fake. But 700% is a hard number to explain away with “oh we just really like HBM now.” Maybe the AI boom gave these three companies cover to do something they’d have wanted to do anyway. Maybe not. That’s exactly what the case is supposed to figure out, and I’d rather wait for actual evidence than pretend I already know the verdict, which is more than I can say for about 80% of the comments I’ve read on this.

Micron’s already denied everything publicly, saying they “compete vigorously, fairly and in compliance with all applicable laws.” Samsung and SK Hynix haven’t said much yet, as of when I’m writing this. Fair enough, maybe their lawyers told them to shut up, or maybe they’ve got nothing to hide. Both possible. Annoyingly.

Three companies, one very suspicious price chart, one legitimately huge AI boom, and a genuinely uncomfortable history of doing exactly this before. Both stories, cartel or three firms independently chasing the same money, fit the public facts we’ve got right now. Not a cop-out answer. Just the honest one. This is gonna take years to resolve unless it’s dismissed or settled early, and until then anyone telling you they know for sure what happened is guessing same as the rest of us.

I’ll probably write an update once discovery starts turning things up. Or I won’t, who knows, depends how much RAM costs by then.