MU stock surges 15% on record Micron Q3 revenue and HBM4 ramp

Micron Technology Inc. (NASDAQ: MU) shattered fiscal third-quarter expectations on Wednesday after market close, reporting revenue of $41.46 billion against a consensus estimate of roughly $35.25 billion and non-GAAP earnings per share of $25.11 versus the $20.28 analysts had penciled in. The Boise-based memory manufacturer simultaneously disclosed 16 multi-year Strategic Customer Agreements representing approximately $100 billion in minimum contracted revenue and $22 billion in upfront customer cash, and guided fiscal fourth-quarter revenue to $50 billion plus or minus $1 billion, roughly $6 to $7 billion above Wall Street expectations. Micron Technology shares surged about 15 percent in extended trading, with the stock having already appreciated roughly 700 percent over the prior twelve months to push the company past a $1 trillion market capitalization. The numbers individually are remarkable. Taken together, they describe a structural rewrite of how the memory industry will be modelled for the rest of this decade.

What the Micron Technology fiscal third-quarter revenue blowout actually means for the semiconductor cycle

The headline revenue figure of $41.46 billion is up roughly 346 percent year over year from $9.30 billion in the prior-year quarter, and it represents the fifth consecutive quarterly record for Micron Technology. That kind of growth rate in a hardware business is not normal under any framework that has historically been applied to memory manufacturers. The DRAM and NAND industries have for four decades been modelled as deep cyclicals with brutal pricing volatility, capacity gluts every three to four years and operating margins that frequently turned negative at the trough. Investors who continued to apply that framework through calendar 2025 have been comprehensively wrong, and the third-quarter print closes the argument.

GAAP gross margin reached 84.6 percent versus 37.7 percent a year ago, a figure that places Micron Technology in the same gross-margin neighbourhood as enterprise software companies rather than commodity hardware suppliers. The economic transformation behind that number is the high-bandwidth memory shift. HBM sits physically adjacent to AI accelerators inside hyperscaler data centers, only three companies in the world manufacture it at scale, and demand has comprehensively outrun supply for more than 18 months. When a commodity becomes a bottleneck, it stops behaving like a commodity, and the pricing power flows directly to the bottom line.

The second-order observation is that GAAP net income climbed to $28.24 billion in the quarter from just $1.89 billion in the prior-year period. A nearly fifteen-fold increase in quarterly net income is a number that forces every memory-sector analyst to rebuild their forward models from scratch. Sell-side estimates for fiscal 2027 will be revised aggressively higher in the coming days, and the question for institutional buyers is whether the consensus catches the actual run rate or continues to lag for several quarters as analysts struggle with the credibility of straight-line projections off this base.

The third signal sits in the segment composition. Data center revenue exceeded $25 billion in the quarter, putting Micron Technology on an annualized data center run rate of more than $100 billion, with enterprise SSD revenue alone hitting $5 billion and more than doubling sequentially. The Mobile and Client Business Unit matched Core Data Center dollar-for-dollar at $11.52 billion, generating an 87 percent gross margin and an 86 percent operating margin. The Automotive and Embedded Business Unit quadrupled year over year to $4.63 billion at a 79 percent gross margin. The margin profile across every segment now sits at levels that the memory industry has historically never sustained, and the breadth of the strength suggests this is not a single hyperscaler concentrating purchases ahead of a capacity expansion.

Why the 16 Strategic Customer Agreements may permanently break the memory cycle

The most strategically important disclosure from the Micron Technology print is not the revenue or earnings number. It is the structural shift embedded in the 16 multi-year Strategic Customer Agreements covering approximately 20 percent of DRAM volume and up to one-third of NAND volume. These are take-or-pay contracts with predefined price bands, multi-year duration and $22 billion in upfront customer cash already received by Micron Technology. The customer base for these agreements is the hyperscaler and enterprise infrastructure community that has spent the past 18 months unable to secure adequate HBM and high-capacity DRAM supply at any price.

The economic logic for the customers is straightforward. Hyperscalers have collectively earmarked something north of $725 billion in artificial intelligence infrastructure capital expenditure for calendar 2026, and the operational risk of having that capital expenditure stranded because memory supply runs out is unacceptable at boardroom levels. Paying $22 billion upfront to lock in supply through 2028 is rational risk management when the alternative is multi-billion-dollar GPU clusters sitting idle. The economic logic for Micron Technology is even cleaner. Roughly $100 billion in minimum contracted revenue with price floors removes the cyclical downside that has historically destroyed memory-sector valuations during inventory corrections.

The strategic implication is that the memory industry has just acquired a contracted revenue base that more closely resembles enterprise software subscription economics than commodity hardware cycles. Investors who have spent careers waiting for the memory cycle to roll over now have to consider the possibility that the cycle has been structurally lengthened, that pricing volatility will be muted by contract floors, and that the bear case for memory stocks now needs to incorporate a scenario where artificial intelligence demand actually declines rather than simply expecting normal cyclical mean reversion. Neither of those repricings has fully flowed through to sector multiples yet.

The execution risk in this construct is concentrated in supply ramp. HBM3E and HBM4 are fully booked through calendar 2027, with demand extending into 2028, but the bottleneck for incremental capacity is fabrication infrastructure that takes three to four years to bring online. Micron Technology began construction in January 2026 on its $100 billion multi-decade semiconductor manufacturing complex in Clay, New York, but that capacity does not arrive in any meaningful volume until fiscal 2028. Until then, the company is selling everything it can produce, which is excellent for pricing but limits the upside from incremental volume growth.

How HBM4 ramp positioning Micron Technology against Nvidia Vera Rubin reshapes the AI infrastructure stack

The HBM4 ramp is the technical centerpiece of the bull case, and the third-quarter commentary tightened the position considerably. Volume shipments of HBM4 commenced in calendar first-quarter 2026 for Nvidia’s Vera Rubin platform, with yield ramps tracking ahead of HBM3E. That last phrase carries disproportionate weight. Memory yield ramps that exceed prior-generation curves are rare in the industry, and they translate directly into faster gross-margin expansion as more die per wafer reach saleable quality. The implication is that the gross-margin trajectory through fiscal 2027 may run ahead of even the elevated current levels, which is not a scenario most sell-side models have built in.

The competitive structure of the HBM market is the second piece of the puzzle. Only three suppliers manufacture HBM at meaningful scale. The relative positioning between Micron Technology, SK Hynix and Samsung Electronics determines who captures the disproportionate economic rent from the AI buildout. Micron Technology’s positioning as the only United States-headquartered HBM manufacturer carries strategic significance beyond the pure economics, particularly in an environment where semiconductor supply-chain resilience has become a national-security framing for procurement decisions across both hyperscalers and defense customers.

The third element is the demand vector beyond HBM itself. Robotics was flagged on the call as a 20-year demand horizon, the artificial intelligence personal computer category is raising recommended memory specifications to a 32 gigabyte minimum, and flagship smartphone DRAM requirements are rising sharply. None of these vectors individually rival the data center build in dollar terms, but together they create a structural floor under DRAM demand that did not exist in prior cycles. The bear case for memory has historically rested on PC and mobile unit declines absorbing data center strength. The Micron Technology numbers suggest those two demand pools are now decoupled enough that mobile and client weakness no longer offsets data center strength even at the margin.

What the Micron Technology capital expenditure step-up signals about supply discipline across the memory industry

Micron Technology will spend approximately $10 billion in fiscal fourth-quarter capital expenditure and intends to step up further into fiscal 2027, with full-year fiscal 2027 capex expected to land in the low to mid forty billions before continuing to rise. That spending profile is substantial in absolute terms, and it raises the question that has historically destroyed memory cycles. Will capacity expansion eventually outrun demand, trigger pricing collapse and erase the current margin structure.

The structural answer this cycle differs from prior cycles in three ways. First, the technical complexity of HBM and leading-edge DRAM has lengthened the time required to bring incremental capacity online from roughly 18 months in earlier cycles to three to four years now. Second, the capital intensity of each incremental wafer has risen substantially, meaning that suppliers face higher financial penalties for building speculative capacity that subsequently sits idle. Third, the Strategic Customer Agreement structure now in place removes some of the volume that would otherwise flow into spot market dynamics and the resulting price discovery, which dampens the feedback loop that historically converted oversupply into rapid price collapse.

Management commentary on capital returns reinforced that the company intends to direct incremental cash flow toward share repurchases rather than aggressive dividend increases or speculative merger activity. Free cash flow is expected to exceed $30 billion in the fiscal fourth quarter alone, which gives Micron Technology the financial firepower to fund the Clay, New York complex, sustain leading-edge research and development, and return material capital to shareholders simultaneously. The Clay complex is a multi-decade build, but the upfront customer cash from the Strategic Customer Agreements meaningfully de-risks the financing of the early-phase construction.

What the Micron Technology AI memory cycle means for the broader semiconductor sector and hyperscaler economics

The read-across from this print rolls through several adjacent sectors. The most direct beneficiaries are SK Hynix and Samsung Electronics, where similar HBM economics should be flowing through to financial results across the same demand cycle. The relative positioning of the three suppliers will determine which captures the largest share of the contracted revenue economics, but all three should benefit from the broader structural shift away from pure spot pricing.

The second-order read-across is the equipment makers. Applied Materials, KLA Corporation, Lam Research and the Dutch lithography equipment supplier ASML should see continued strength in memory-related tooling orders through fiscal 2027 and 2028 as Micron Technology and its peers ramp capital expenditure. Within that group, advanced packaging tooling for HBM stacking is the most directly exposed segment, and the order books at suppliers focused on that capability should be reviewed against the contracted volume implications of the Micron Technology disclosure.

The third-order read-across is the hyperscaler cost structure itself. Hyperscaler capital expenditure is increasingly being absorbed by memory rather than compute. If memory contract pricing remains elevated through 2027 and 2028, the unit economics of hyperscaler AI services will shift, and the pressure to extract pricing power from end customers, whether enterprises consuming AI APIs or consumers using AI-augmented products, will intensify. That dynamic ultimately determines whether the artificial intelligence capital expenditure cycle continues at current trajectory or moderates as hyperscalers recalibrate their return-on-invested-capital expectations.

The risk to the entire thesis is concentrated in two scenarios. The first is a sharp deceleration in hyperscaler artificial intelligence capital expenditure, whether driven by macro stress, regulatory intervention or simple recognition that the current capex trajectory exceeds the addressable revenue opportunity. The second is a geopolitical disruption that affects semiconductor supply chains broadly, including Taiwan-related contingencies that would ripple through every node of the artificial intelligence infrastructure stack. Neither scenario is the base case, but both warrant explicit weight in any forward valuation framework for Micron Technology at current levels.

Key takeaways on what the Micron Technology fiscal Q3 print means for the company, peers and the broader AI infrastructure ecosystem

• Fiscal third-quarter revenue of $41.46 billion and non-GAAP earnings per share of $25.11 represent a 16 percent and 24 percent beat respectively, with revenue up 346 percent year over year and gross margin expanding to 84.6 percent from 37.7 percent.
• The 16 Strategic Customer Agreements covering approximately $100 billion in minimum contracted revenue and $22 billion in upfront customer cash structurally lengthen the memory cycle and remove much of the historical pricing downside risk.
• Data center revenue exceeded $25 billion in the quarter on an annualized run rate above $100 billion, with enterprise SSD revenue alone hitting $5 billion and more than doubling sequentially.
• HBM3E and HBM4 are fully booked through calendar 2027 with demand extending into 2028, and HBM4 yield ramps for Nvidia Vera Rubin are tracking ahead of HBM3E, which positions Micron Technology for further gross-margin expansion.
• Fiscal fourth-quarter revenue guidance of $50 billion plus or minus $1 billion sits roughly $6 to $7 billion above consensus, and non-GAAP earnings per share guidance of $31 represents another sequential step up.
• Free cash flow is expected to exceed $30 billion in the fiscal fourth quarter alone, providing financial firepower for the Clay, New York complex, sustained research and development investment, and continued share repurchases.
• Capital expenditure stepping up to the low to mid $40 billions in fiscal 2027 funds the multi-decade supply ramp without triggering the historical oversupply pattern, because incremental capacity takes three to four years to arrive at scale.
• The market capitalization above $1 trillion, achieved on a roughly 700 percent twelve-month share-price appreciation, forces sell-side analysts to materially raise fiscal 2027 and fiscal 2028 estimates and likely triggers multi-quarter earnings revision momentum.
• The structural read-across favors SK Hynix and Samsung Electronics in memory, the semiconductor equipment suppliers in tooling, and creates margin pressure on hyperscaler unit economics as memory costs absorb a larger share of artificial intelligence capital expenditure.
• The principal risks remain hyperscaler capex deceleration, geopolitical supply-chain disruption and the eventual normalization of memory pricing as Clay-era capacity arrives in fiscal 2028 and beyond.