Micron’s Contracts Raise Floors but Market Still Prices a Peak

Micron Technology posted record $41.46 billion Q3 revenue in fiscal 2026, up 346% from a year earlier, with non-GAAP gross margin at 84.9% and diluted EPS of $25.11. The same night it disclosed 16 multi-year take-or-pay Strategic Customer Agreements that lock roughly $100 billion in minimum revenue through 2030.

Those contracts look like the end of memory’s familiar boom-and-bust pattern. The stock’s single-digit forward multiple says investors still see a peak.

Record Quarter Rewrites the Scale

Fiscal Q3 ended May 28. Revenue hit $41.46 billion versus $9.30 billion a year ago and $23.86 billion the prior quarter. Non-GAAP operating income reached $33.68 billion. Cash, marketable investments and restricted cash stood at $30.2 billion; net cash was about $24.4 billion after debt paydown.

Data center revenue alone topped $25 billion, an annualized run rate over $100 billion. HBM4 shipments already exceeded $1 billion. Management guided fiscal Q4 revenue to $50 billion plus or minus $1 billion at roughly 86% gross margin and non-GAAP EPS near $31.

• $41.46B Q3 revenue (+346% YoY)
• 84.9% non-GAAP gross margin
• $25.11 non-GAAP EPS
• $18.3B adjusted free cash flow

Business-unit splits show the breadth. Cloud Memory delivered $13.8 billion at 83% gross margin. Core Data Center added $11.5 billion at 87%. Mobile and Client matched that $11.5 billion at 87%. Automotive and Embedded contributed $4.6 billion at 79%.

CEO Sanjay Mehrotra called the results proof of memory’s strategic value in the AI era and said the multi-year agreements will enhance durability and predictability.

Sixteen Contracts Rewrite the Playbook

The 16 strategic customer agreements details are the real shift. Fourteen carry cumulative minimum revenue of about $100 billion at contract prices over their remaining terms. Customers committed roughly $22 billion in cash deposits and related financial support.

Typical term runs five years, calendar 2026 through 2030. Automotive deals are usually three years. The signed set covers roughly 20% of Micron’s DRAM volume and one-third of NAND over the period. Four very large customers and three medium-sized ones dominate; the rest are smaller automotive names. Management expects half or more of total company revenue under SCAs once the full planned set is complete.

These are take-or-pay. Customers must purchase the contracted volumes or pay anyway. That structure is rare in DRAM and NAND history.

How the Floors and Deposits Work

Largest agreements generally set a ceiling at current calendar-Q2 market prices for existing products and a floor that holds through the full term. Several smaller SCAs use fixed prices or no bands. When all planned deals close, agreements with fixed or near-current ceilings are expected to cover about 40% of revenue. Pricing inside the bands stays between floor and ceiling.

The floors are set to deliver gross margins well above any peak quarterly margin Micron printed in prior cycles. Next-generation products such as newer HBM, DDR6 and LPDDR6 carry separate premium negotiations. Deposits give customers skin in the game and give Micron cash to fund expansion with lower risk.

• Take-or-pay volume commitments, multi-year and binding
• Price floors supporting margins above historical peaks
• Ceilings tied near current spot for many large deals
• $22 billion customer cash and financial commitments
• Target half-plus of revenue under SCAs

Mehrotra said customers recognize that AI roadmaps now depend on reliable long-term access to advanced memory. Supply shortages will take years to ease.

The Multiple Still Prices a Peak

After the print the stock rallied then gave back gains. By late July it traded near $820, roughly 32% off the post-earnings high near $1,213. Market capitalization sat around $926 billion. Trailing P/E hovered near 20. Forward P/E sat at forward P/E near 5.8 times.

That multiple is the market’s plain statement. It is willing to pay for the contracted floors and the near-term shortage. It is not willing to capitalize today’s $25-$31 quarterly EPS as a permanent run rate. The same dynamic appeared in the Micron’s recent slide into bear territory earlier in the AI memory run, when valuation compressed even as fundamentals improved.

Crowd conversation on X captured the split. High-engagement posts noted that $18-22 billion in deposits prove customers fear multi-year shortage. Other voices pointed out that low single-digit forward multiples historically mark peak earnings in memory, not the start of a new normal.

Past Cycles Never Had This Floor

Traditional DRAM and NAND cycles ran three to four years. Oversupply crushed prices, producers cut capital spending, shortages returned, prices spiked, new fabs were announced, and the loop repeated. Peak gross margins in prior up-cycles rarely cleared the mid-50s to low-60s for long. Micron’s current 85% band is roughly double those old peaks.

Earlier attempts at long-term agreements rarely carried hard take-or-pay terms or large upfront deposits. Spot pricing still dominated. When demand slowed, contracts were renegotiated or quietly ignored. The new SCAs shift more of that risk onto customers who need the bits for AI accelerators, high-end phones, PCs and autonomous vehicles.

The structural case rests on three legs: AI systems are architecturally hungry for bandwidth and capacity; greenfield fabs take years and face labor, permitting and power bottlenecks; process transitions deliver slower bit growth and higher cost per bit. HBM’s wafer-intensive nature further squeezes conventional DRAM supply. Mehrotra said the company currently lacks line of sight to when supply catches demand.

Capacity Wall Arrives After 2027

Micron itself expects tight conditions to persist beyond calendar 2027, with only gradual improvement in 2028. Its own U.S. expansion and peers’ Korean and other projects begin to contribute later in the decade. Industry wafer-start growth remains mid-single digits against demand that has been compounding far faster.

That timeline is why the contracts matter now and why they do not settle the long-term debate. The uncontracted half of revenue, still roughly 50% or more until the full SCA set closes, stays exposed to spot. If AI infrastructure spending cools, or if models become more memory-efficient, pricing pressure can reappear there first. New capacity that arrives in 2028-2030 will test whether the floors hold or whether renegotiation pressure builds.

Skeptics note that every prior supercycle ended the same way: returns far above cost of capital pulled in capital, supply arrived in a wall, and margins collapsed. The absolute dollars committed this time are larger, which could make the eventual downswing steeper even if the floor contracts cushion Micron relative to history. Michael Burry’s short against the AI memory rally earlier reflected exactly that peak-cycle skepticism.

Who Gains Visibility and Who Keeps Spot Risk

Customers who signed gain multi-year supply certainty at known price bands. That matters when a single missing HBM stack can idle an expensive GPU rack. They also accept the obligation to take or pay and the capital tied up in deposits. Smaller buyers or those who stayed on spot retain flexibility but face allocation risk and full price volatility. The earlier memory rally that squeezed buyers already showed how quickly spot markets can punish the unhedged.

Micron gains revenue visibility, margin floors, and customer-funded expansion. It also accepts concentration: a handful of large hyperscalers and AI chip designers now drive a bigger share of the high-value mix. HBM qualification cycles are long, which raises switching costs in Micron’s favor, yet also raises the impact if one major platform slows.

The contracts do not erase cyclicality. They compress the amplitude for the covered portion and push more of the residual risk onto the uncontracted half and onto the 2028-and-beyond supply calendar. That is why the stock can print software-like margins and still trade at a cyclical multiple.

Micron has rewritten the near-term playbook with binding floors and deposits that past cycles never offered. The market’s pricing of those earnings as temporary remains the clearest signal that the longer cycle question is still open.

Frequently Asked Questions

What exactly do Micron’s 16 Strategic Customer Agreements cover?

They are mostly five-year take-or-pay contracts (three years for most auto deals) running through 2030 that commit customers to specific DRAM and NAND volumes, including HBM where applicable. Fourteen of the 16 carry roughly $100 billion in cumulative minimum revenue at contract prices; the set currently represents about 20% of DRAM volume and one-third of NAND, with a target of half or more of total revenue once fully ramped.

How do the price floors in the SCAs compare with past cycle peaks?

Management stated the floors are set to support gross margins well above any peak quarterly margin Micron achieved in prior cycles. Prior up-cycle peaks typically sat in the mid-50% to low-60% range for extended periods; current overall non-GAAP gross margin is already 84.9% and Q4 guidance is about 86%.

How much cash have customers committed under the new deals?

Micron projects $22 billion in cash deposits and related financial commitments under the signed SCAs. Roughly $18 billion of that is described in market commentary as cash, giving the company prepaid capital that de-risks its own expansion spending.

What portion of Micron revenue remains exposed to spot pricing?

Until the full planned SCA portfolio is complete, roughly half of revenue sits outside the contracted structure. Even after full ramp, the remaining share and any volume above contracted minimums will still move with market prices, though next-generation products carry separate premium negotiations.

When does Micron expect industry supply to begin catching demand?

The company expects tight conditions to last beyond calendar 2027, with only gradual supply improvement starting in 2028. It currently has no line of sight to the point when supply fully catches rising demand, citing long fab lead times, skilled labor shortages, permitting, energy infrastructure and slower bit growth from advanced nodes.

Disclaimer: This article is for informational purposes only and does not constitute investment advice. Past performance is no guarantee of future results. Conduct your own research before making any investment decisions.