Wednesday night, Micron Technology reported $41.5 billion in quarterly revenue, more than four times the year-ago figure and $5.8 billion above the Wall Street consensus, and the stock surged more than 13% in after-hours trading to $1,188. The numbers are extraordinary. They are not the story. The story is the 16 contracts buried further down the earnings deck that may have permanently ended the most destructive recurring event in semiconductor history.
For thirty years, the DRAM industry has operated on a cycle that destroyed and rebuilt billions of dollars in shareholder value on a rolling basis. Prices surge, manufacturers race to add capacity, supply overwhelms demand, prices collapse, capital exits, the cycle resets. Samsung and SK Hynix have both lived through four of these full rotations since 2000. Micron’s stock went from $103.38 to $1,213.56 in twelve months and still trades at 9.5 times forward earnings, partly because every investor who has ever owned a memory stock has the scar tissue of watching that cycle reverse.
The 16 Strategic Customer Agreements Micron announced on Wednesday may be the mechanism that breaks it. They are structured as take-or-pay contracts: customers are legally obligated to purchase committed volumes regardless of whether market prices fall. The agreements cover approximately 20% of Micron’s DRAM volume and one-third of its NAND volume, running from calendar 2026 through 2030, with automotive contracts on a three-year term. The company disclosed that total remaining performance obligations across all signed SCAs amount to approximately $100 billion, with $22 billion in upfront cash deposits and financial commitments already received from customers. Per the earnings call prepared remarks filed with the SEC, Micron CEO Sanjay Mehrotra described the agreements as “fundamentally transforming” the company’s business model.
Put that in context. In fiscal Q3 2025, one year ago, Micron reported $9.3 billion in revenue. Wednesday’s figure of $41.5 billion represents a 346% increase year-over-year and a 74% increase from the immediately prior quarter of $23.9 billion. Gross margin reached 84.9%, up ten percentage points sequentially and more than double a year earlier, according to the Micron 8-K filed with the SEC on June 24. Operating income was $33.7 billion on an 81.2% operating margin. Free cash flow hit $18.3 billion, a quarterly record. Q4 guidance of $50 billion in revenue, plus or minus $1 billion, sits approximately $6.4 billion above the analyst consensus of $43.6 billion.
What the Contracts Actually Mean
The SCA structure has a specific precedent in reserved capacity models in cloud computing, where hyperscalers commit to multi-year compute capacity at locked prices in exchange for preferential allocation. The difference is that AWS Reserved Instances carry no obligation to pay for unused capacity. Micron’s take-or-pay contracts do. Customers who signed are exposed to the contracted volume floor regardless of what happens to memory spot prices between 2026 and 2030.
That is the mechanism that breaks the cycle. In every prior downturn, falling demand triggered a sequential response: customers cancelled orders, spot prices fell below contract prices, manufacturers absorbed the inventory glut, then cut capacity. The SCA structure removes the first step. If a hyperscaler has committed to take-or-pay terms on 20% of Micron’s DRAM through 2030, cancelling in a downturn is not a customer decision. It is a legal and financial event with specific consequences. The floor price provisions, which Goldman Sachs noted in its post-earnings research note guarantee Micron gross margins above any level achieved in any prior cycle, mean the company retains pricing power even in soft demand environments.
Morgan Stanley’s Joseph Moore flagged the significance on the call, asking management to clarify the $22 billion in customer commitments and specifically the $18 billion in cash deposits. TheStreet’s live earnings blog summarised the exchange plainly: this is insurance against the AI bubble bursting. If AI demand wanes materially, Micron’s contracted customers absorb the financial exposure, not Micron’s income statement. Only 20% of DRAM volume and one-third of NAND volume is under the SCA framework at present. Management indicated they expect that figure to reach 50% or more of total revenue over time.
The Estimate Miss That Nobody Can Explain
The magnitude of the consensus miss deserves specific attention because it points to a structural problem in how Wall Street is modelling the memory market. Analysts tracking Micron through LSEG had consensus revenue of $35.69 billion. The actual figure was $41.46 billion, a beat of $5.77 billion or 16.2%. Adjusted EPS of $25.11 came in 23.8% above the $20.28 consensus, per 24/7 Wall Street data.
Q4 guidance of $50 billion represents a $6.4 billion beat against consensus before the quarter has started. That is not a modest upside. That is the analytical community systematically failing to model a business that is operating in conditions it has no historical template for. Bank of America, which raised its price target to $1,550 from $1,500 following results, maintained a buy rating and cited the SCAs specifically as a structural change rather than a cyclical tailwind. Goldman Sachs also raised its target. The 16 signed agreements, Goldman noted, feature both price floor and ceiling provisions, not merely floor protections, meaning the framework offers customers predictability in both directions.
The Supply Picture Through 2027
Micron disclosed on Wednesday that it “currently does not have line of sight as to when memory supply will be able to catch up with increasing demand” and guided investors to expect “tight conditions to persist beyond calendar year 2027.” That statement came from prepared remarks filed with the SEC, not ad-hoc commentary, and it captures the structural dynamic that is the foundation of both the SCA contracts and the current pricing environment.
The supply constraint is architectural. Every wafer directed toward High Bandwidth Memory production removes approximately three equivalent wafers from commodity DRAM supply, according to Counterpoint Research analysis cited by Sourceability in March. As SK Hynix, Samsung, and Micron have redirected capacity toward HBM to serve AI accelerator demand, DDR5 and general-purpose DRAM supply has contracted sharply even without any decrease in DDR5 demand. Micron’s Idaho ID1 fab is not expected to begin wafer output until mid-2027. ID2 follows in late 2028. Samsung’s Pyeongtaek P5 facility reaches production in 2028. New capacity is not a near-term factor. Counterpoint Research projects the global memory market to exceed $1.5 trillion by 2027, with server memory rising from under 50% of the total in 2025 to 57%.
SK Hynix confirmed on Thursday that it plans to raise approximately $29.4 billion through a Nasdaq ADR listing scheduled for July 10, issuing 17.79 million new shares to fund construction of its Yongin semiconductor cluster and an advanced packaging facility in Indiana, according to Reuters. SK Hynix shares surged 12% on Thursday on a combination of the listing news and the read-across from Micron’s results. Rolf Bulk, head of semiconductors and infrastructure at Futurum Group, told CNBC that the Micron beat “is a very positive read-across for SK Hynix, who are exposed to the exact same market dynamics.” SK Hynix trades at 6.97 times forward earnings against Micron’s 9.5 times, per Dow Jones Market Data, a valuation gap that the US listing is explicitly designed to close through expanded analyst coverage and institutional access.
The Market Structure Question SK Hynix Creates
There is a tension in Wednesday’s events that the earnings coverage has largely missed. Micron’s results validated the memory super-cycle thesis with the most comprehensive beat in the company’s history. The SK Hynix $29.4 billion Nasdaq listing does two things simultaneously: it validates the same thesis by demonstrating that the world’s dominant HBM producer believes its equity is worth raising capital against, and it introduces the largest new US-listed semiconductor equity offering in years into a market that had just watched memory stocks fall sharply on AI valuation fears before the Micron numbers restored confidence.
The Roundhill Memory ETF, which holds SK Hynix, Samsung, and Micron as its top three positions, surged 10% in after-hours trading following Micron’s results, per CNBC data, after falling 14.25% on Tuesday. That two-day range captures precisely the volatility that the SCA structure is designed to remove from Micron’s revenue line, even if it remains endemic to the equity market’s read on the sector.
For broader market context on where semiconductor and AI equity valuations sit after Tuesday’s selloff and Wednesday’s recovery, current risk asset positioning across US markets reflects the same AI-driven rotation that has defined 2026’s equity narrative. The commodity side of the AI infrastructure equation connects to the supply chain analysis published here in March, when the AI materials trade was first becoming visible in commodity pricing. For the energy cost side of memory fab economics, where natural gas prices affect operating costs in Korea and Idaho, Brent pricing remains a relevant input to the operating cost models of all three major producers. For the monetary policy backdrop against which Micron’s $50 billion Q4 guidance will be discounted, BofA now expects the Fed to hike three times in 2026 under Kevin Warsh in September, October, and December, per Reuters. Gold’s positioning near all-time highs reflects the same stagflation hedge dynamic that makes Micron’s locked-in contract revenue structurally more valuable in a rising-rate environment.
The memory industry has spent thirty years proving that no commodity can sustain its pricing peak. Wednesday night, Micron suggested it may have found the mechanism that tests that assumption. Sixteen contracts, $100 billion in minimum obligations, take-or-pay terms through 2030. The cycle’s next test begins in Q4, when guidance says revenue should reach $50 billion.
That number was $9.3 billion twelve months ago. The analysts who set consensus at $43.6 billion already know how this has been going.
