379% Growth and an 87% Margin: Micron's Record Quarter Shows AI Now Runs on Memory

For most of the past decade, memory chips were the commodity tier of the semiconductor industry — cyclical, low-margin, and strategically boring. Micron’s fiscal fourth-quarter 2026 results, reported after the close on September 30, are the latest and loudest confirmation that this era is over. The company posted quarterly revenue of $54.23 billion, up roughly 379% year over year, with gross margin approaching 87% and GAAP net income of $37.70 billion for the quarter. And then it guided the next quarter even higher.

The numbers

The headline figures from fiscal Q4 2026, which ended in August:

• Revenue: $54.23 billion, versus $41.46 billion in the prior quarter and $11.32 billion a year earlier — up about 31% sequentially and 379% year over year
• GAAP net income: $37.70 billion, implying a quarterly net margin that most software companies would envy
• Gross margin: approximately 87% (non-GAAP), with diluted EPS of $33.42 beating analyst estimates by over 7%
• Full fiscal 2026 revenue was 3.5x last year’s record, with data center revenue up roughly fourfold, per CEO Sanjay Mehrotra on the earnings call

And the guidance: Micron told investors to expect $60–63 billion in revenue for fiscal Q1 2027 — a figure that, on its own, would have exceeded the company’s entire annual revenue in any fiscal year before 2024.

What is actually driving this

The proximate cause is no secret: AI data center buildouts have converted high-bandwidth memory (HBM) and high-capacity DRAM from commodity parts into allocation-constrained strategic resources. Three dynamics matter:

HBM is sold out through 2027. Micron said its HBM supply is effectively fully booked through calendar 2027, and that 75% of its total 2027 capacity has already been sold. That is a stunning level of forward visibility for a memory maker — an industry that historically lived quarter to quarter on spot prices. Customers are signing long-term supply agreements stretching up to five years to lock in allocation.

Memory intensity per AI workload keeps climbing. Inference and long-context reasoning models are memory-hungry in ways that training-only workloads were not. As reasoning chains, agent workflows, and multimodal context windows expand, the DRAM and HBM content per GPU rack rises faster than the GPU count itself. Micron’s DRAM revenue has become the center of gravity of the entire company.

Supply cannot respond quickly. DRAM capacity additions take years and enormous capital. Industry forecasts cited around the results see the combined DRAM and NAND market reaching $550–570 billion in 2026 and $800–850 billion by 2027 — growth that existing fab plans simply cannot serve, which is why pricing power has shifted so decisively to the memory vendors.

CEO Sanjay Mehrotra’s framing on the call leaned into this: “Micron delivered record fiscal 2026 results, and we expect an even stronger fiscal 2027. AI is becoming Super Intelligence (SI), and memory enhances this transformation.”

The Wall Street reaction — shrugging at blowout numbers

Here is the genuinely interesting part: despite beating on every line and guiding above consensus, Micron’s stock barely moved. Reports described Wall Street “barely blinking” at the blowout quarter.

The explanation is a classic cycle-top anxiety, with three strands:

1. It’s already priced in. Micron has surged hundreds of percent over the past year and cleared a $1 trillion market cap earlier in 2026. When a stock runs that far, “record everything” is the base case, not a surprise.
2. CapEx fear. Investors worry that today’s scarcity margin funds tomorrow’s overbuild — that Micron, Samsung, SK hynix, and CXMT all expanding simultaneously will recreate the familiar memory glut. On the call, management pushed back, arguing HBM demand growth will continue to outpace traditional DRAM and that disciplined long-term agreements de-risk the buildout.
3. Customer concentration. When your revenue is increasingly sold forward to a handful of hyperscalers and AI labs, the market starts asking what happens to those contracts if AI capex ever flattens — the same question now attached to every supplier in the NVIDIA orbit.

Why this matters beyond one earnings report

Micron’s quarter is a clean read on where value is migrating in the AI stack. For two years the industry’s mental model was “GPUs are the scarce resource, everything else is abundant.” Fiscal 2026 inverted that: the binding constraints are now memory bandwidth, memory capacity, power, and packaging — and the companies that control those inputs are capturing margins that look less like commodity semis and more like a toll road.

It also reframes the cost curve of AI itself. If memory is sold out through 2027 at 87% gross margins, then a meaningful share of the next two years of AI infrastructure spending is already committed to DRAM and HBM vendors — regardless of which model lab wins the capability race. The memory supercycle has quietly become one of the largest capital flows in the technology economy, and Micron — the most US-pure-play of the big three DRAM makers — is its clearest single-company expression.

The bullish scenario management sketched is straightforward: HBM market reaching ~$100 billion by 2028, growing near 40% annually, with Micron holding a structurally stronger share than it ever had in legacy DRAM. The bear scenario is equally simple and equally old: memory is still memory, cycles still cycle, and 87% margins are exactly the kind of number that attracts capacity until they’re gone.

Both can’t be right. But on the evidence of this quarter, the market’s willingness to yawn at 379% growth tells you the debate is no longer about whether the memory boom is real — only about how long it lasts.