What did Micron report for Q3 FY2026?

Micron Technology reported Q3 FY2026 results after market close on June 24, 2026: revenue of $41 billion (quadrupling year-over-year from ~$10B in Q3 FY2025), gross margins above 81% for the first time in company history, over $1 billion in HBM4 (high-bandwidth memory generation 4) revenue, and disclosure of 16 signed Strategic Capacity Agreements (SCAs) with four large customers and three medium customers across data center, consumer, and automotive segments. CEO Sanjay Mehrotra disclosed over $100B in signed strategic contract value. HBM is sold out through calendar 2026.

Why did Micron's revenue quadruple year-over-year?

AI memory demand. HBM (high-bandwidth memory) is the memory architecture that physically sits on every Nvidia GPU and every Google TPU — it is the most performance-critical input in AI compute infrastructure. Three companies globally make HBM (Micron, Samsung, SK Hynix), and HBM4 production capacity is structurally constrained through 2026. As a result: (1) HBM pricing has expanded margins to historic levels (81%+ gross margin is roughly double Micron's pre-AI cycle peak); (2) volume has scaled with the AI capex buildout; (3) customers are signing multi-year strategic capacity agreements rather than buying spot — Micron disclosed 16 such SCAs in this quarter alone, including major commitments from Anthropic (announced June 22) and likely Nvidia, Google, and OpenAI.

What is a Strategic Capacity Agreement (SCA)?

An SCA is a multi-year contractual commitment between a memory supplier and a large customer covering volume, pricing, and (in some cases) co-design collaboration. They differ from traditional commodity-memory purchase orders in three ways: (1) they extend 3-5 years instead of quarter-to-quarter; (2) they typically include capacity reservation — the customer pre-commits to buy a specified volume in exchange for guaranteed allocation; (3) they often include technical co-design (joint engineering on memory subsystem performance, packaging, or workload optimization). The 16 SCAs Micron disclosed across data center, consumer, and automotive segments suggest the chip-cycle has fundamentally shifted from commodity-spot to long-term partnership for the largest customers.

If Micron is doing so well, why did the stock fall 13% on June 23?

Two reasons that aren't contradictory. (1) HBM concentration cuts both ways. Micron has the highest HBM revenue concentration of the three memory suppliers, so when the market re-rates AI capex (which it did on June 23, 2026), Micron moves the most. The -13% drop was a positioning unwind ahead of earnings, not a fundamentals signal. (2) Earnings were Wednesday June 24 after close — the Tuesday sell-off priced in worst-case before the print. The actual results (revenue $41B vs ~$35B consensus, gross margin 81% vs 81% consensus) were a beat, and the stock recovered in after-hours and Thursday June 25 trading. The pattern (sharp pre-earnings selloff, in-line or better print, recovery) is characteristic of heavily-positioned AI-leverage names.

Quick Answer

Micron Q3 FY2026 Earnings: $41B Revenue, HBM4 Sold Out

Published: June 25, 2026

Micron Q3 FY2026 Earnings: $41B Revenue, HBM4 Sold Out

Micron Technology reported Q3 FY2026 results after market close on Wednesday June 24, 2026: $41 billion in revenue (quadrupling year-over-year), gross margins above 81% for the first time in company history, over $1B in HBM4 revenue, and disclosure of 16 signed Strategic Capacity Agreements representing over $100B in committed contract value. HBM is sold out through calendar 2026. The print landed the morning after a 13% stock drop on June 23 and the day after Micron’s strategic deal with Anthropic. Here’s what it means.

Last verified: June 25, 2026.

TL;DR

Reported: After market close Wednesday June 24, 2026
Revenue: $41 billion (up ~4x year-over-year from ~$10B Q3 FY2025)
Gross margin: Above 81% (first time in company history; roughly double the pre-AI-cycle peak)
HBM4 revenue: Over $1 billion (first quarter of meaningful HBM4 ramp)
Strategic contracts: 16 signed SCAs with 4 large + 3 medium customers across data center, consumer, and automotive
Total contract value disclosed: Over $100B
HBM status: Sold out through calendar 2026
Stock context: Recovered Thursday June 25 after Tuesday June 23 -13% drop ahead of print

The headline numbers

Metric	Q3 FY2026	Q2 FY2026	YoY
Revenue	$41B	$23.9B	+~310%
Gross margin	81%+	74.9%	+~600 bps
HBM4 revenue	$1B+	~$0 (pre-ramp)	n/a
Strategic Capacity Agreements (cumulative)	16 disclosed	smaller	new disclosure category
Signed contract value (cumulative)	$100B+	n/a	new disclosure category

Three of these numbers are unusual enough to deserve unpacking.

Revenue: $41B is structural, not cyclical

For context, Micron’s full-year FY2024 revenue (the year before the AI memory cycle accelerated) was approximately $25B. The company is now generating more in a single quarter than it generated in entire pre-AI years. This is not the kind of growth that comes from a normal commodity-memory upcycle — it reflects a structural change in the memory market driven by HBM-on-every-AI-accelerator demand.

Gross margin: 81%+ is unprecedented for memory

Memory has historically been a low-margin commodity business — Micron’s gross margin range in normal cycles is 30-45%. Even peak commodity-memory cycles topped out around 50%. The 81% gross margin in Q3 FY2026 reflects the fundamentally different pricing power Micron has on HBM4: it is supply-constrained, sold via multi-year contracts at pre-negotiated prices, and concentrated among a small number of strategic customers.

This is closer to Nvidia gross-margin territory than traditional memory-supplier territory. It also means the downside is larger if AI capex slows — every percentage point of HBM mix that shifts back to commodity DRAM compresses margins meaningfully.

Strategic Capacity Agreements: the new contract regime

The 16 SCA disclosure is new and important. Strategic Capacity Agreements are multi-year contractual commitments — typically 3-5 years — between Micron and a large customer covering volume, pricing, and sometimes co-design.

The 16 SCAs span:

4 large customers (likely some combination of Nvidia, Google, OpenAI, Anthropic, Microsoft, Amazon, Meta)
3 medium customers (likely auto OEMs, mid-tier hyperscalers, large enterprise)
3 segments: data center, consumer, automotive

The $100B+ total committed contract value disclosed by CEO Sanjay Mehrotra is a forward-looking aggregate — not Q3 revenue, but multi-year contractual visibility. For context, that’s roughly two years of Q3-pace revenue locked in via committed contracts.

What it means for the AI capex thesis

Micron’s earnings are the most important real-time AI capex signal, because HBM demand maps almost 1:1 to AI accelerator volume. If hyperscalers are buying GPUs and TPUs, they’re buying HBM. If they’re slowing GPU and TPU purchases, HBM ramps slow.

The June 24 print sends three signals:

1. Demand is structural, not cyclical, through 2026

HBM4 sold out through calendar 2026 means every chunk of HBM4 capacity Micron can produce in the next ~18 months is already contractually committed. There is no spot supply to buy. If Nvidia, Google, or OpenAI wanted to suddenly accelerate GPU/TPU/Jalapeño production, they couldn’t — HBM is the constraint.

2. Pricing power is durable through the SCAs

The 16 multi-year SCAs lock in pricing at current premium levels. Even if HBM4 demand softens in 2027, the committed contracts insulate Micron’s revenue and margin through the contracted period. This is the structural difference from prior memory cycles, where pricing collapsed within quarters of demand softening.

3. The Anthropic deal context

Two days before this print, on June 22, Micron and Anthropic announced their strategic Series H + multi-year supply agreement. The 16 SCA disclosure almost certainly includes the Anthropic agreement. The Anthropic deal validates the SCA pricing and structure — and signals that more AI-lab equity-aligned deals are likely in the pipeline.

How it reconciles with the June 23 sell-off

Micron fell 13.18% on Tuesday June 23, 2026 — its worst single-day move in years. Then it printed $41B revenue and 81% margins on Wednesday June 24 and recovered.

This pattern is characteristic of heavily-positioned AI-leverage names heading into earnings:

Pre-print positioning unwind: investors sell into earnings to lock in gains; momentum funds sell on technical breaks
In-line or better print: fundamentals beat consensus
Recovery: stock recovers most of the pre-print drop within 1-3 sessions

The June 23 sell-off was correctly interpreted as a re-rating of AI-capex leverage, not a Micron-specific concern. But it cut Micron disproportionately because Micron has the most concentrated HBM exposure of the three memory suppliers — so any “AI capex might slow” narrative hits Micron hardest.

What changes from here

For Micron’s stock thesis

The Q3 print confirms the structural AI-memory cycle is intact through 2026. The $100B+ in committed SCA value gives multi-year revenue visibility. The Anthropic Series H investment adds equity-aligned customer alignment. The 81% gross margin shows pricing power is real.

Risks: (1) HBM capacity ramps from Samsung, SK Hynix, and Chinese makers (CXMT, YMTC) could eventually pressure pricing; (2) AI capex sentiment is fragile, as June 23 showed; (3) HBM4 → HBM5 transition risk in 2027-2028.

For AI capex watchers

Micron’s earnings are the cleanest read on whether the AI accelerator buildout is continuing at the pace hyperscalers have guided. Q3 FY2026 says: yes, through 2026, with no signs of slowing. The next data point will be Samsung and SK Hynix earnings in late July — if both confirm HBM sold-out status and multi-year contract commitments, the structural AI capex thesis is intact.

For Anthropic, OpenAI, Google, Meta

HBM allocation is now the binding constraint on AI compute scale-up. Without an SCA-equivalent commitment, you can’t grow your inference and training footprint faster than your existing HBM allocation allows. Expect every major AI lab to have signed or to sign equivalent multi-year HBM commitments with one or more of the three memory suppliers over 2026-2027.

Bottom line

Micron’s Q3 FY2026 print is the largest single quarterly revenue and margin print in the company’s history, driven by AI memory demand that is now structural rather than cyclical. The 16 SCAs and $100B+ in committed contract value give multi-year revenue visibility. HBM is sold out through 2026.

The June 23 sell-off was a positioning unwind, not a fundamentals signal. The recovery on June 24-25 confirms it. The AI capex thesis, measured through the most concentrated bottleneck (HBM), is intact through the end of 2026 — and probably well beyond.

For investors: the risk-reward shifts from “is the demand real?” (answered: yes) to “how durable is the pricing?” (answered: contractually locked for 3-5 years through SCAs). For AI infrastructure planners: HBM is your binding constraint, sign a strategic capacity agreement or fall behind on compute capacity.