Micron’s revenue quadrupled as AI memory demand pushes gross margins above 81 percent

Micron Technology posted fiscal third-quarter revenue of nearly $42bn, quadrupling from just over $9bn a year earlier and beating Wall Street estimates by a wide margin. The results, reported on Tuesday, confirm that the company riding the AI memory boom hardest is the one whose stock has already climbed roughly 700 percent over the past year.

Adjusted earnings came in above $25 a share, compared with analyst expectations of roughly $21. GAAP net income exceeded $28bn, or nearly $25 a share, up from just under $2bn in the year-ago quarter. Gross margins hit above 81 percent, up from 69 percent in the prior quarter and 27 percent a year earlier.

The headline number is revenue growth. Micron brought in nearly $42bn against a consensus estimate of roughly $36bn, driven almost entirely by surging demand for high-bandwidth memory, the stacked DRAM chips that sit next to GPUs inside AI accelerators built by Nvidia and Google. HBM has become the binding constraint on AI infrastructure expansion, and Micron is one of only three companies in the world that can make it.

CEO Sanjay Mehrotra said Micron can currently fulfil only between half and two-thirds of customer demand for HBM. The company’s entire 2026 HBM supply is sold out under multi-year contracts, and it has collected $22bn in customer cash deposits, essentially prepayments from hyperscalers desperate to lock in supply.

Micron’s next-generation HBM4 chips are ramping what the company described as twice as fast as the previous HBM3E generation. HBM4 revenue has already exceeded one billion dollars. The technology is essential for the latest accelerators from Nvidia and Google, where memory bandwidth rather than raw compute increasingly determines inference throughput.

The forward guidance was equally aggressive. Micron projected fiscal fourth-quarter revenue of approximately $50bn, plus or minus one billion, against analyst estimates of roughly $44bn and a year-ago figure of just over $11bn. The company raised its full-year capital expenditure forecast to more than $25bn, up from a previous target of $20bn, to expand production capacity for HBM and advanced DRAM.

Micron’s market capitalisation crossed one trillion dollars on 26 May, making it the latest memory chipmaker to reach that threshold as the AI-driven memory supercycle reshapes valuations across the semiconductor industry. The stock’s roughly 700 percent gain over the past year reflects a market that is pricing memory not as a cyclical commodity but as structural AI infrastructure.

The company said it expects the total addressable market for HBM to grow at a compound annual rate of roughly 40 percent through 2028, rising from approximately $35bn in 2025 to around $100bn. Micron plans to return 100 percent of excess free cash flow to shareholders, a commitment enabled by the cash deposit programme that reduces the capital risk of its expansion.

There are caveats worth noting. Micron remains the smallest of the three HBM suppliers, behind SK Hynix and Samsung, and its share of Nvidia’s HBM4 allocations is the thinnest of the trio. The broader memory market is also shifting, with Chinese manufacturers like CXMT expanding aggressively into consumer DRAM segments that the Big Three have deprioritised in favour of AI chips.

Memory pricing is cyclical by nature, and the current supercycle depends on hyperscaler capital expenditure continuing at its current pace. If AI infrastructure spending slows or HBM supply catches up with demand, the margins that Micron reported this quarter would compress rapidly. The 81 percent gross margin is historically extraordinary for a memory company and reflects shortage economics as much as product superiority.

For now, the numbers speak for themselves. Revenue that quadruples in a year, margins that triple, and a guidance print that exceeds estimates by more than $6bn are not normal results for any company, let alone one that was losing money two years ago. Micron’s earnings confirm that the AI memory shortage is intensifying, not easing, and that the companies making the chips inside AI accelerators are capturing value at a rate the market is still recalibrating to price.