I was thinking of stopping writing altogether when this caught my attention, and thought, "is it worth it"? Tech finance isn't really my thing and the AI chip hype cycle has been going on long enough that another "company hits milestone valuation" story felt like the last thing anyone needed. But some of the data wasn't being mentioned and I thought of poking the tiger so…
First: The Trillion Is Real
May 26, 2026. Micron closes at $891 a share after a 19% single-day jump. $1 trillion market cap. UBS had just upgraded them from a $535 target to $1,625, highest on Wall Street.
Before this AI cycle, Micron had never cracked $150 billion in valuation. Ever. The memory sector was always treated as kind of a low-status corner of semiconductors ,cyclical, commodity-driven, perpetually one supply glut away from a brutal earnings wipeout. "Boring but necessary" was the standard pitch.
What actually changed is that AI datacenters eat memory the way a race car eats fuel, and the specific type they need most ,High Bandwidth Memory, stacked directly alongside Nvidia's GPUs to keep them fed with data ,is something only three companies on the planet can manufacture at scale. Micron, Samsung, SK Hynix. That's the list. When Microsoft, Google, Amazon, and Meta started spending at levels that made previous datacenter buildouts look quaint, supply couldn't respond fast enough. Fabs take years and tens of billions to build. Prices went up. Margins expanded. Micron's HBM4 chips are already locked in for Nvidia's next major GPU platform. Their entire 2026 HBM supply sold out before the year started.
The rally extended a 180% gain in Micron's share price over the past 12 months. That's not hype math ,those are real customers paying real money for chips they can't get elsewhere right now.
The trillion-dollar number is a repricing of expected future earnings, not a reflection of factories that magically appeared. If a company has a billion shares and investors decide they're worth twice as much, market cap doubles instantly. No new buildings required. Whether the AI datacenter spending that drives this stays permanently elevated is genuinely unknown, but the current demand is not made up.
The Part That I Couldn't Quite Square
What I couldn't understand at first was this: if Micron is supply-constrained and demand has never been higher, why were they cutting production aggressively two years ago?
The answer is in their own earnings calls, and it's not hidden. December 2022, CEO Sanjay Mehrotra directly on record: "Micron is exercising supply discipline by making significant cuts to our capital expenditures and wafer starts." During the 2022-2023 demand slowdown the company deliberately throttled its next-generation DRAM ramp. Samsung and SK Hynix did essentially the same thing at the same time.
Memory prices stabilized. Then rose. By Q4 2025 Micron's gross margins had expanded 17 percentage points to 41%. Revenue up 196% year-over-year.
Now, a class action lawsuit in 2018 tried to argue this was illegal coordination. Ninth Circuit dismissed it in 2022. The ruling was that three companies independently cutting production in a consolidated market is "conscious parallelism," not a conspiracy, and US antitrust law requires actual proof of a preceding agreement, not just parallel behavior. The case failed on that standard.
So the behavior is legal. I'm not disputing the court. The issue here is that an industry of three players, all publicly announcing production cuts in the same window, all benefiting from the pricing outcome that follows, that's the same economic result as a cartel. Just without the back-room meetings that would make it illegal. Antitrust scholars have argued about this exact structure for decades without resolution. Courts have sided with companies. Whether that outcome reflects good law or law that hasn't caught up to concentrated markets is a separate debate that I don't have the expertise to settle.
What I do know is that Micron's own executives called it "supply discipline" in investor calls, which is a useful thing to keep in mind when the same company is talking to Congress about Chinese competitors.
The Subsidies
Micron secured $6.1 billion in direct CHIPS Act funding for their Clay, New York megafab. New York State added up to $5.5 billion in refundable Green CHIPS tax credits plus a 49-year property tax abatement estimated at $284 million and a $100 million community fund contribution. Federal Investment Tax Credits tack on another $11.3 billion.
I did check these figures, they're in the actual government agreements and state development records, not advocacy estimates.
Total public support in New York alone: over $21 billion.
Then in November 2025, Micron told officials the fab construction would be delayed to "late 2030." Simultaneously, the company filed a request with Onondaga County for an additional $1.76 billion in sales tax exemptions on construction materials. The North Syracuse School District was looking at forfeiting something like $370 million in potential revenue. County leaders postponed the vote after community pushback.
Semiconductor fab construction is genuinely complex and delays are real, not invented by bad actors. I want to be fair about that. But the sequence here — lock in $21 billion in public commitments, announce a multi-year delay, request another $1.76 billion while the delay is ongoing, puts local municipalities in a position where they've already committed infrastructure resources and made political promises, so refusing the additional request means risking the whole project. That's not fraud. It's just a negotiating position that happens to favor the company at every step. The risk sits with taxpayers, the upside accumulates to shareholders.
The MATCH Act Lobbying
This one actually took me a while to figure out how to write because the story people tell about it depends entirely on where you start.
Micron CEO Sanjay Mehrotra held a closed-door roundtable with members of the House Foreign Affairs Committee roughly a month before the vote, and a similar meeting with Senate Banking Committee Republicans. The bill, the MATCH Act, is designed to cut Chinese memory makers CXMT and YMTC off from the deep ultraviolet lithography equipment they need to keep advancing. It cleared committee in late April 2026.
Micron's argument: these companies are acquiring restricted equipment through loopholes, the technology has direct military applications, this is a national security issue.
The counterargument, from the R Street Institute: CXMT is gaining market share not because of unfair trade practices or technology theft, but because Chinese firms are producing competitive products at competitive prices and moving up the value chain. The instinct to address competitive pressure through export restrictions rather than through investment and market performance is regulatory protectionism.
I genuinely don't know which framing is more correct. The dual-use concern around semiconductor manufacturing equipment is real. The Chinese military-civilian fusion policy is documented. At the same time, the company writing the lobbying checks is the company whose entire market position depends on slowing Chinese competition. Closed-door meetings with the exact committees writing the legislation. That's worth sitting with even if the national security rationale is legitimate.
What I do know is that when Micron lobbies Washington under "national security," the economic outcome is that their competitors get cut off from the tools they'd use to undercut Micron on price. Whether you think that's acceptable depends on where you come down on the question of whether US industrial policy should protect specific companies or let them compete. I don't have a clean answer to that.
The Historical Record That Doesn't Come Up in the Milestone Coverage
Between 1998 and 2002, Micron participated in a documented DRAM price-fixing cartel. DOJ investigated. In December 2003, Micron regional sales manager Alfred Censullo pled guilty to obstructing the grand jury investigation, specifically, altering and concealing handwritten notes about phone calls with competitor sales managers where price recommendations for customers including IBM's server division were discussed. The broader industry paid $731 million in criminal fines. Micron escaped prosecution because they cooperated as the first whistleblower under the DOJ's Corporate Leniency Program, then settled civil antitrust claims in January 2007.
This is in DOJ press releases. It happened.
I'm not arguing that a 20-year-old conviction defines the current company. Executives turn over, legal functions mature, the regulatory environment tightened after those fines. Holding 2026 Micron permanently accountable for 2002 conduct would be unfair and I'm not doing that.
But when people in tech forums speculate about whether current "supply discipline" is really just legal coordination, the answer isn't purely theoretical. There's a documented baseline for what this industry was willing to do when it thought nobody was watching. That context exists. It's public record. It just doesn't appear in the trillion-dollar celebration pieces.
Boise
The PFAS situation in Boise is where the story gets genuinely murky and where I'd push back on some of the more aggressive framing I've seen from advocacy groups.
What's actually confirmed: Micron's operations discharge pretreated wastewater into Boise's municipal system. Neither of the city's two main treatment plants has PFAS removal capability, no reverse osmosis, no activated carbon filtration. PFAS compounds from semiconductor manufacturing pass through and discharge into the Boise River. Water testing in 2023 found PFAS downstream of the treatment plants.
What nobody has established: that contamination levels are causing specific public health harm, or that Micron's specifically at fault in a way that triggers regulatory liability versus being one industrial contributor to a widespread class of chemicals that are everywhere.
Issue is this is too complex for me to explain more, I just leave it here and readers can search it on Google.
The Consumer Exit
December 2025: Micron formally exits its Crucial consumer brand. No more retail DRAM, no more consumer SSDs. Their executive framing was about serving "larger, strategic customers in faster-growing segments." Translation: enterprise AI pays more per gigabyte than you do building a PC, so you're not a priority anymore.
Let's get to the subsidy context. Over $21 billion in public support, substantially justified by job creation and domestic chip supply arguments. Then the company exits the consumer market to maximize enterprise margins. Nobody attached conditions about serving retail customers to those agreements, so there's no breach. But if you were a policymaker who signed off on those subsidies partly because "domestic memory supply" matters for everyday consumers, you might have a thought or two about how that worked out.
The Sum of It
Micron is a legitimate trillion-dollar company riding a structural shift in how memory infrastructure gets valued in an AI-driven world. The HBM story is real. The supply constraints are real. The long-term contracts with hyperscalers are real. Three companies in the world can do what Micron does, and that scarcity has pricing power attached to it.
The other story , $21B in subsidies with a construction delay and additional benefit requests stacked on top, an industry history of documented price-fixing, legal-but-eyebrow-raising supply discipline, lobbying to kneecap competitors through export controls, exiting consumer markets while holding billions in public support ,hat's also real and documented.
The trillion-dollar coverage mentions the first story. I wanted to write the second one down somewhere.
Going forward I won't be adding source lists at the bottom of articles. Tracking down every link, formatting them, adding them after the piece is already written — it's a lot of work for a site this small. Everything I reference is named and dated in the text itself. That'll have to be enough.


