Micron Technology reported fiscal Q3 2026 earnings that blew past every estimate on the board, with revenue of $41.46 billion and adjusted EPS of $25.11 crushing Wall Street consensus by roughly 16% and 22% respectively, all of it driven by insatiable demand for high-bandwidth memory.
Numbers like these don't come from a company riding a wave. They come from a company that has become infrastructure. Micron's Q3 result, reported June 24, delivered $41.46 billion in revenue against a consensus sitting near $35.6 billion, and adjusted earnings of $25.11 per share against expectations of around $20.60. Net income hit $28.24 billion. Gross margins came in at roughly 81%. Year-over-year EPS growth is approaching 1,000%. Those aren't beats. They're a different conversation entirely.
The engine behind all of it is high-bandwidth memory, the stacked DRAM that sits inside Nvidia's AI accelerators and makes the difference between a data center that can train frontier models and one that can't. Micron's HBM production is sold out through 2026 under binding contracts, which means the revenue visibility here is unusual. The company isn't guessing what demand looks like next quarter. It already knows.
That sold-out status is worth understanding clearly. HBM isn't commodity memory with a new name on the package. Manufacturing it requires roughly three times the wafer capacity of standard DRAM, and only three companies in the world can produce it at scale: Micron, SK Hynix, and Samsung. When hyperscalers including Nvidia, Microsoft, and Meta are collectively spending north of $600 billion on data center buildout, and every AI accelerator they order needs HBM to function, the supply math gets very tight very fast. Pricing stays elevated. Margins expand. The Q3 results are what that looks like in practice.
If the Q3 numbers were surprising, the Q4 outlook was startling. Micron guided for $50 billion in revenue, plus or minus $1 billion, against analyst expectations near $43 billion. That guidance alone implies Micron expects to accelerate from an already record quarter. The stock rose roughly 9% in after-hours trading. More telling than the price move is what the guidance implies: that HBM demand from AI infrastructure is not plateauing. It's still climbing. Micron has outlined roughly $200 billion in planned capacity expansion in response to what it describes as a historic memory supply crunch. That's a company betting its entire forward trajectory on AI infrastructure remaining the defining capital expenditure cycle of the decade. Given that the Q3 numbers just validated exactly that bet, it's hard to argue with the logic.
The broader semiconductor investment narrative is shifting in real time around these results. For the past two years, much of the conversation around AI infrastructure has centered on GPU supply: how many Nvidia chips can be produced, how quickly, and who gets them first. What Micron's blowout quarter illustrates is that the memory layer is just as constrained, and arguably less discussed. You can't run an H100 or a Blackwell chip without HBM. Memory isn't downstream of the AI buildout. It's a co-equal bottleneck.
SK Hynix moves to capitalize on the same moment #
The timing of SK Hynix's US listing announcement, the same week as Micron's earnings, is not accidental. Bloomberg reported that SK Hynix is looking to raise $29.4 billion through a Nasdaq listing of American Depositary Receipts, with trading expected to begin July 10, making it one of the largest US listings in recent memory. SK Hynix holds roughly 62% of global HBM shipments and supplies more than half of the HBM chips inside Nvidia accelerators. Its stock has risen more than 300% in 2026 alone.
Two memory giants, both sold out, both expanding capacity aggressively, both tapping capital markets at the same moment. That's not competition undermining each other's story. It's two companies racing to build enough supply to meet demand that neither of them can currently satisfy. The AI infrastructure supercycle, it turns out, is big enough for both.
What Micron's Q3 report actually confirms is something the skeptics have been reluctant to accept: the demand driving these numbers isn't speculative. Hyperscalers are ordering AI accelerators in enormous quantities, and those accelerators don't ship without HBM. The contracts are signed. The revenue is booked. Micron just reported $41 billion to prove it. The Q4 guide suggests $50 billion is next. At some point, a trend this consistent stops being a supercycle and starts being the new baseline.
Also read: Anthropic's distillation problem reveals that export controls alone cannot hold the line in the US-China AI race • An AI law firm just beat two lawyers in court for £400 and the legal industry should be paying attention • Dubai Holding is in talks to buy into Hscale as Gulf capital moves to lock up Europe's AI infrastructure before the buildout peaks