Micron Q3 FY2026 earnings: 84.9% margins and $22B in contracts

A plain-English decode of Micron's Q3 FY2026 print, the $22 billion in customer contracts nobody flagged first, and what it means for NVIDIA, ASML, and the equipment trio.

As of June 24, 2026, Micron Technology (MU) reported Q3 FY2026 revenue of $41.5 billion (up 370% year over year), a company-record gross margin of 84.9%, and disclosed $22 billion in customer commitments across 16 Strategic Customer Agreements. Q4 FY2026 guidance came in at $50 billion in revenue and $31 in EPS. CEO Sanjay Mehrotra said memory tightness persists "beyond calendar 2027." Source: BargoAI research.

What did Micron report for Q3 FY2026?

Micron (MU) reported Q3 FY2026 revenue of $41.5 billion, non-GAAP EPS of $25.11, and a company-record gross margin of 84.9% on June 24, 2026 after market close.

All figures are historical, sourced from Micron's investor relations press releases on June 24, 2026 and management's forward guidance. Live prices, options positioning and order flow update every session on the BargoAI Micron dashboard.

For twenty years memory investors have argued about one thing: when the cycle turns. Every rally ended when supply caught demand. Every crash ended when demand caught supply. It was the most reliable pattern in semiconductors, and the most reliably disappointing.

On June 24, 2026, Micron printed the quarter that broke that mental model. Not because the numbers were big. Because the story around them changed. Revenue of $41.5 billion, up 370% year over year. EPS of $25.11 against a $20.49 consensus. Gross margin of 84.9%, a company record, up from 39% one year earlier. And a Q4 guide of $50 billion in revenue with an 86% gross margin and $31 in EPS, against a Street looking for $44 billion and $26. The tape had been prepared for a beat. This was not a beat. This was a reset.

The single most important disclosure did not make the headline numbers. It was buried in the prepared remarks and picked up by a Reuters wire in Ireland the next morning: $22 billion in customer commitments, locked in through 16 multi-year Strategic Customer Agreements signed in the quarter alone. The prior quarter Micron had one such contract. This quarter it signed sixteen. That is what changed the memory story. Not the EPS. The contract book.

What did Micron's CEO actually say on the call?

Micron CEO Sanjay Mehrotra said memory demand exceeds supply across all segments and that tight conditions persist "beyond calendar 2027," pushing the AI capex peak debate out by 12 to 18 months.

Mehrotra has been running Micron since 2017. His public style is uniformly cautious. On this call he delivered the single most bullish sentence a memory CEO can put on the tape.

We are disappointed to be unable to meet demand from our customers across all market segments.

Sanjay Mehrotra, CEO, Micron Q3 FY2026 earnings call, June 24, 2026

Read that twice. A memory company said out loud that it is turning customers away. In a cyclical commodity business, that is the top of every cycle. Except this time the customers were not walking away. They were pre-paying. Mehrotra continued:

Sustained and strong industry demand, along with supply constraints, are contributing to tight market conditions. And we expect these conditions to persist beyond calendar 2027.

Sanjay Mehrotra, CEO, same call

That phrase, "beyond calendar 2027," was the sentence that repriced the memory group overnight. The consensus assumption at the start of 2026 was that memory tightness would peak in the year and start rolling over in 2027. Mehrotra pushed the tightness runway out by twelve to eighteen months. Every "2026 is the top" model now has to be rebuilt.

Why did the $22 billion contract disclosure matter?

The $22 billion in signed Strategic Customer Agreements matters because it converts Micron's revenue from spot-priced cyclical to multi-year contracted, which changes the earnings multiple the market should apply.

Memory has always been a spot-priced business. Chips get made, buyers negotiate contracts a quarter at a time, and the spot price swings 40% to 80% between cycle peaks and troughs. That is why memory trades at a low multiple. The earnings are famously unpredictable.

Strategic Customer Agreements change that math. An SCA is a multi-year supply contract at agreed prices, with agreed volumes, over three to five years. Once signed, the buyer cannot cancel and the seller cannot repurpose the capacity. It is Netflix pricing for silicon.

When completed, we expect approximately half or more of our company revenue to be under these.

Sanjay Mehrotra on Strategic Customer Agreements

Half or more of company revenue under multi-year contracts is what a utility looks like. It is what a SaaS company looks like. It is not what a cyclical memory business looks like. The market had always applied a low multiple to Micron because the earnings series was spiky. If half the revenue is contracted at locked prices, the earnings series is no longer spiky. The multiple has to move.

The best independent commentary on the SCA disclosure came the same night from Gene Munster:

The most underappreciated data point on Micron's earnings call was that the number of five-year SCAs increased from one last quarter to seven this quarter. That's significant because it suggests customers expect this buildout to last for at least five years and want greater pricing visibility over that time frame.

Gene Munster, Deepwater, June 24, 2026

Seven five-year contracts is not hyperscalers hedging a soft demand quarter. It is hyperscalers pre-committing to five more years of AI infrastructure spend. That is the "is this a bubble" question, answered by the customers themselves in the form of signatures on multi-year purchase orders.

Why are 84.9% gross margins monopoly math, not memory math?

Micron posted a company-record 84.9% gross margin because HBM demand vastly exceeds supply, forcing Micron to pick which customer to serve and at what price. Memory companies normally run 30% to 40% gross margins in good times.

For context, Apple runs roughly 46%. Microsoft runs roughly 70%. Visa runs roughly 80%. Micron just posted higher margins than all three.

Source: Micron company filings and Q4 FY26 midpoint guidance. BargoAI research.

The best framing of this margin story came the night of the print from a market commentator on X:

Micron didn't just beat. It repriced the entire memory complex. 86% gross margin is allocation economics. You only print that when supply is rationed and customers are signing strategic deals to lock capacity.

That is exactly right. 84.9% is not a good quarter for a commodity business. It is a monopoly rent. It happens only when the customer has no other option and knows it. Every other high-margin business in tech (Microsoft, ASML, Visa) prints those numbers because their product is genuinely differentiated and there is no viable second source. Micron just printed those numbers because HBM (high-bandwidth memory) demand exceeds supply by so much that management can pick which customer to serve, and at what price.

The follow-up question every investor should ask: how long can 84.9% last? Management explicitly refused to guide costs for the rest of the year. That is the honest asterisk. But the Q4 guide of 86% implies at least one more quarter of margin expansion, and the "beyond 2027" language implies the shortage runway is measured in years, not quarters.

How did Micron's Q4 guide compare to Wall Street consensus?

Micron's Q4 FY2026 guide of $50 billion in revenue and $31 in EPS came in 14% above the consensus revenue estimate of $44 billion and 19% above the consensus EPS estimate of $26. That gap is what broke the sell-side model.

Source: Micron Q4 FY26 management guidance and LSEG consensus at time of print. BargoAI research.

At $31 in quarterly EPS, Micron is annualizing north of $120 per year in earnings. On the stock price around the print, that translated to a forward P/E of about 10 for a company growing 370% year over year with 86% gross margins. There is no name in technology that offers that combination today.

What questions did Sanjay Mehrotra refuse to answer?

Mehrotra refused to answer four specific analyst questions during the Q3 FY2026 Q&A: HBM as a percentage of DRAM revenue, HBM share versus Samsung at HBM3E and HBM4, HBM 2028 market share target, and cost guidance for the rest of the year. Each refusal is a signal.

Earnings calls are as much about what management does not say as what they do. The Q3 FY26 Q&A had four hard questions from senior analysts that got non-answers.

The pattern is consistent. Management refused to say anything that could telegraph how they would allocate capacity or price it. That is not defensive. That is a monopolist protecting its leverage. When a memory CEO chooses opacity over transparency about market share, it is because the answer is favorable and management does not want customers to know it.

What does the Micron print mean for NVIDIA, ASML, and equipment stocks?

Micron's Q3 FY2026 print de-risked NVIDIA's HBM4 supply for the 2027 Rubin GPU generation, gave ASML a fresh multi-year EUV commitment, added roughly $2 billion in incremental capex orders to Applied Materials, Lam Research, and KLA, and added base-die wafer volume for TSMC.

The Micron print did not happen in isolation. Every major AI-semiconductor position moved on it, and each moved for a specific reason.

NVIDIA (NVDA). Mehrotra implicitly confirmed HBM4 ramps into NVIDIA. Rubin and Vera Rubin GPUs need HBM4 12-high stacks. Micron said that ramp is real and over $1 billion in HBM4 revenue already shipped in Q3 FY26. NVIDIA's 2027 HBM4 supply risk was one of the last-standing bear points on the name. It got de-risked.

ASML. CFO Mark Murphy called out a multi-year ASML EUV agreement as part of Micron's capex build-out. ASML got a fresh anchor customer commitment on the tape at exactly the moment its own next earnings print was approaching. A supply-side positive of that magnitude shifts sell-side models.

Applied Materials (AMAT), Lam Research (LRCX), KLA (KLAC). Micron raised FY26 capex from $18 billion to $20 billion, specifically earmarked for HBM and 1-gamma DRAM capacity (the newest leading-edge process). Tool deliveries were pulled in. The equipment trio all traded meaningfully higher in the sessions following the Micron print. This is a real order-book event, not a sentiment trade.

TSMC (TSM). Micron's HBM4 uses TSMC-made base logic dies. Every HBM stack Micron ships is base-die wafer volume for TSMC. Another quiet leg of the TSMC story confirmed on someone else's earnings call.

SK Hynix and Samsung Electronics. The two other HBM makers. If Micron's HBM4 is sold out and ramping 2x ahead of HBM3E, SK Hynix (the leader) is even more sold out. The entire memory complex re-rates higher together.

Is the China DRAM (CXMT) worry overblown?

Yes, the CXMT worry is mis-sized. The report driving the panic actually projects DRAM stays undersupplied even after CXMT ramps, and CXMT competes at the low-margin commodity end where Micron makes the least money, not in HBM.

The counterweight to everything above is the CXMT story. ChangXin Memory Technologies is China's state-backed DRAM champion. In June 2026, SemiAnalysis published a deep-dive report titled "China's CXMT Is Set to Challenge DRAM Incumbents." The report timing lined up with a sharp sell-off across the memory group in the following sessions. The narrative went viral.

Two things about that narrative do not survive contact with the actual report.

First, the report's own model has DRAM staying undersupplied even after CXMT's ramp, high-single-digit percentage in the near term widening to low-to-mid teens the year after. The panic assumes Chinese supply floods the market. The report says the opposite.

Second, CXMT competes at the low-margin end of the market. Its 1Q26 gross margin trails Micron's badly. "Nearly all" of CXMT's bits are conventional DDR and LPDDR, per SemiAnalysis. "HBM is still a very minimal contributor." The AI memory profit pool (HBM, server RDIMM) is structurally protected through 2027 and likely beyond.

The real CXMT story is not share-stealing. It is Samsung, SK Hynix, and Micron VACATING the low-end commodity market to chase HBM, and China backfilling the hole. A margin overhang on 15% to 20% of Micron's revenue mix, not a threat to the AI memory profit pool.

⚠️ Honest risks

• Samsung HBM3E qualifies at NVIDIA. The one thing Mehrotra refused to talk about. If Samsung breaks through Micron's allocation power, the pricing dynamic weakens fast. Single biggest live risk on the thesis.
• The cost-down curve breaks. CFO refused to guide costs. 84.9% gross margin holds only if the cost trajectory continues. Any inventory write-down or yield issue and the "monopoly margin" thesis cracks in a single quarter.
• AI capex cools materially. The whole thesis is downstream of hyperscaler capex sustaining. If Microsoft, Meta, Google, Amazon, or Oracle cut AI infrastructure spending guidance, the SCA contracts become less bulletproof.

How did Micron stock react after earnings?

Micron stock ripped 16.75% on June 25, 2026, the first full trading session after the print, cleared the $1,200 call wall on the post-print rally, and closed at $1,227.90 with one-month implied volatility elevated to 103%.

Snapshot: June 24 through June 26, 2026.

The options market repriced Micron faster than the equity market. On June 25, 2026, dealers went from long-gamma pinning to a broken-through call wall in a single day. The stock cleared the $1,200 call wall on the post-print rally and closed the session up 16.75%. One-month implied volatility elevated to 103%. That is what a re-rating looks like in the options tape.

Long-gamma pinning below the flip, short-gamma amplification above it. Spot cleared the $1,200 call wall on the post-print rally. Source: BargoAI options positioning snapshot, June 26, 2026.

These are point-in-time levels from the trading sessions around the print. Options positioning shifts every day. For current MU price, gamma walls, IV term structure, and signed flow, use the live BargoAI dashboard.

What this means for your portfolio

If you hold NVIDIA, AMD, Broadcom, ASML, or Micron, here is how a serious AI and semiconductor investor should think about the Micron Q3 FY2026 print:

• Micron itself is priced like a cyclical while growing like a monopoly. Every quarter that 80%+ gross margins hold, the argument for a multiple re-rating gets stronger.
• The AI capex debate just got extended by 12 to 18 months. Mehrotra's "beyond calendar 2027" language reprices everything downstream in the AI-semi complex.
• NVIDIA's 2027 HBM4 supply risk is meaningfully lower than the market assumed heading into the Micron print. That removes one bear thesis on NVDA.
• ASML and the equipment trio (AMAT, LRCX, KLAC) got real order-book evidence, not sentiment. Micron's raised capex flows through their P&L.
• The China DRAM (CXMT) headline risk is real but sized wrong. It hits the low-margin 15% to 20% of Micron's mix, not the HBM profit pool.
• Watch two things that could break the thesis: Samsung getting qualified at NVIDIA for HBM3E, and the memory cost-down curve.

This is not investment advice. All live financials and signals are on bargo.ai.

Sources

• Micron Q3 FY2026 press release and investor materials, investors.micron.com/news-releases
• Micron 8-K SEC filings, SEC EDGAR CIK 723125
• SemiAnalysis, "China's CXMT Is Set to Challenge DRAM Incumbents," June 2026, semianalysis.com
• Deepwater Asset Management commentary by Gene Munster, June 24, 2026, deepwatermgmt.com
• BargoAI research and live data, bargo.ai

Related reading

• Nebius vs CoreWeave: same market cap, very different moats
• BargoAI Micron dashboard

Reviewed by the Bargo editorial desk. All P&L figures, management quotes, and quarterly metrics are historical from Micron's Q3 FY2026 earnings release on June 24, 2026 and do not update. Live price and market data are best consumed via the BargoAI dashboards linked above. This is research and educational content; it is not investment advice. Do your own work.