Micron Revenue Quadrupled to $41 Billion. Wall Street Is Calling It the Next Nvidia.
A memory chip maker from Boise just briefly outranked Meta by market cap. Here is what is driving it - and what could end it.
Micron HBM supply has become a chokepoint in the AI data center buildout - and Wall Street priced that in after Q3 earnings. Revenue hit $41.45 billion last quarter, four times the same period a year ago. Profits climbed from $1.88 billion to $28.2 billion. Shares closed Friday at $1,132, up 236% in a single month from a stock that spent years below $100 a share. Micron briefly overtook Meta's $1.39 trillion market cap on Thursday before settling back to $1.27 trillion by Friday's close. Q4 guidance: $49 to $51 billion. Investors who have spent three years hunting for another public AI infrastructure play after Nvidia appear to have found their candidate.
AI Servers Need 10 Times the Memory a Consumer Laptop Ships With
High-Bandwidth Memory is DRAM stacked in layers and bonded directly to a processor die, built to move data fast enough to feed modern GPU clusters. A single AI training server requires 80 to 160 GB of HBM. A high-end consumer laptop ships with 16 GB, sometimes 32. Nvidia's H100 GPUs ship with 80 GB of HBM2e; the H200 and Blackwell chips step up to HBM3 and HBM3e stacks. Every major cluster buyer - Nvidia, Microsoft, Amazon AWS, Google, Meta, Oracle - is buying at a rate that has drained available global supply.
Standard DRAM and NAND flash are feeling it too. PC makers like Dell and HP have started hoarding commodity memory to hedge against future shortfall. Analysts named the crunch "RAMageddon." IDC projects it continues into 2027, and the price pressure has already reached consumer electronics - Apple products and Xbox consoles cost more because data centers consumed the memory supply first.
Micron makes HBM, DRAM, and NAND. Samsung and SK Hynix compete in the same segment, but Micron is the only US-based producer of all three. That distinction carries real weight for customers managing supply chain risk after years of watching geopolitics fracture chip supply across Asia.
16 Long-Term Supply Agreements Are Micron's Answer to the Boom-Bust Cycle
Past Micron surges followed a predictable arc. Demand outpaced manufacturing capacity, prices spiked, competitors built more fabs, supply flooded back, and the stock collapsed. That cycle played out roughly every four to six years, and investors who remember 2018 and 2022 apply it reflexively to every Micron earnings beat.
CEO Sanjay Mehrotra's team signed 16 strategic customer agreements across data center, consumer, and automotive markets - including named contracts with Nvidia and Anthropic. Micron expects these deals to shift its business model from spot-price commodity supplier toward contracted infrastructure vendor. William Blair analyst Sebastien Naji cited "improving revenue visibility" from the agreements and reiterated an Outperform rating, noting that demand growth continues to outpace the rate new cleanroom space can come online.
Anthropic's deal with Micron fits a pattern worth tracking alongside its other procurement activity: Anthropic is simultaneously in talks to run Claude inference on Microsoft's Maia 200 chip, locking in both memory and compute supply at the same time. For labs operating at scale, Micron HBM is no longer a line item they price at spot - it is a constraint they plan around quarters in advance. That kind of customer behaviour is what makes the 16 SCAs plausible as a structural shift rather than a marketing claim.
Gross Margin Topped Nvidia's - and That Is Where the Risk Lives
Micron's gross margin hit 84% last quarter. Nvidia's came in at 74%. A commodity memory supplier printing margin above the world's dominant GPU maker is not a situation that holds without structural support. Memory prices are highly sensitive to supply, and supply is sensitive to manufacturing capacity arriving on a two-to-three-year lag from groundbreaking to production wafer.
Samsung and SK Hynix are both scaling HBM capacity. Custom silicon efforts like OpenAI's Jalapeño chip could reduce some inference-side memory demand over time, though training clusters show no sign of slowing. If major cloud providers reach Micron's contracted volumes and find spot alternatives improving, margin compression follows.
Micron's Q4 forecast of $49 to $51 billion implies sequential revenue growth above 20%. Analysts will set their 2027 expectations against that number. How much of Micron's revenue mix is contracted versus exposed to spot pricing will become visible in margin variance over the next two to three quarters - and that spread is probably more informative than the market cap comparison to Nvidia.