NextFin News - The numbers are out, and they are unlike anything Micron or the memory industry has ever produced. Revenue came in at $54.23 billion for fiscal Q4 2026, up 379% year over year, against a Wall Street consensus of $51.07 billion. Adjusted EPS of $33.42 beat the $31.61 estimate. Adjusted gross margin hit 87.0%, above the 86.1% consensus and above Micron's own guidance of roughly 86%. For context: Micron's Q4 net income alone, $37.7 billion, exceeded the company's entire fiscal 2025 revenue of $37.38 billion. That is not a typo.
Micron closed Wednesday's regular session at approximately $1,065 before the results dropped. Afterhours, the stock briefly dipped toward $1,050 before recovering toward $1,078, reflecting the market's well-worn pattern with Micron: results too strong to dismiss, guidance too conservative to aggressively re-rate on, even when it beats.
The Segment Breakdown That Reframes the Narrative
The four segment lines tell the fuller story. Core Data Center delivered $18.00 billion, up 36% sequentially. Cloud Memory came in at $16.28 billion against an estimate of $15.14 billion. Mobile and Client added $13.11 billion. The standout surprise was Automotive and Embedded at $6.82 billion, beating the $4.73 billion estimate by 44% and rising 47% sequentially. That last number matters because it is the earliest financial evidence of what management is now openly calling the physical AI cycle.
For the full fiscal year 2026, Micron posted $133.2 billion in revenue and $85 billion in profit. Operating cash flow in the quarter alone reached $43.97 billion, against an estimate of $34.47 billion. The cash generation is substantial enough that even with H1 FY2027 capex guidance of $25 billion and additional spending expected in the back half, the balance sheet is not under pressure.
Q1 FY2027 Guidance: The Number That Will Move Everything Else
The most consequential disclosure was not Q4. It was Q1 FY2027 guidance of $61.5 billion in revenue, plus or minus $1.5 billion, against a consensus of $57.02 billion. Adjusted EPS guidance of $38.15, plus or minus $1.00, came in well above the $35.40 estimate.
The CFO added a statement that has immediate implications for every memory stock: "We anticipate Q1 to be the floor for gross margins in FY27, with higher gross margins beyond Q1 for the remainder of fiscal 2027." That is a direct rejection of the margin peak thesis that bears have been running since the summer. If Q1 at 86.25% is the floor, then the FY2027 consensus EPS of roughly $154, which has been the contested center of the valuation debate all year, is more likely a floor than a ceiling.
Management also stated that Micron expects sequential revenue growth in every quarter of FY2027. That framing, on top of a $61.5 billion Q1 base, sets an implied FY2027 run rate that makes the current forward multiple look structurally low.
No Line of Sight to Balance: What Management Actually Said
The most analytically significant statement from the call was this: "Even with additional industry DRAM clean room space plans, with robust demand trends including new upside requests from customers, we do NOT have line-of-sight to when supply and demand will return to balance."
That is a stronger statement than anything Micron has said in prior quarters. It aligns with SK Hynix's own guidance, disclosed at its July 29 results, that the company does not expect supply-demand equilibrium before 2030. Micron's senior advisor Sumit Sadana has previously noted that "meaningful new supply will only begin ramping in 2028." The gap between now and then is the investment thesis, stated plainly.
HBM pricing for calendar 2027 is confirmed as "much higher than 2026 prices." Micron is now in active collaboration with Nvidia on custom HBM4E implementations. The company has also signed 26 strategic customer agreements, up from 16 last quarter, worth $32 billion in financial commitments. More than 75% of 2027 output is already committed, and management noted that most active customer discussions have already shifted to 2028 capacity.
Physical AI: The Demand Driver Beyond the Data Center
Micron used this call to formally introduce physical AI as a category, not just a talking point. The company stated that humanoid robots could require more than 200 gigabytes of DRAM and multiple terabytes of storage per unit, with memory content "more than an order of magnitude greater" than current semi-autonomous systems. Multiple physical AI customers are already sampling next-generation Micron products.
Automotive and Embedded at $6.82 billion, up 44% above estimates, is the early revenue signal for this. The Automotive and Embedded segment has historically been the quietest part of Micron's business. A single quarter's outperformance could be noise. Combined with the explicit disclosure of physical AI customer sampling and a projected decade-long demand runway, it begins to look like the start of a demand layer that is entirely separate from the hyperscaler cycle.
If physical AI demand ramps in the way management described, it adds memory consumption outside data centers at scale for the first time. Millions of deployed humanoid robots each requiring server-class memory content is a structural shift that existing supply models have not yet been built to accommodate.
The Burry Counternarrative
Today is also the day Michael Burry disclosed put options on Micron, Nvidia, and Palantir, predicting a 1987-style crash and calling AI capital expenditure unsustainable. The disclosure coinciding with Micron's strongest quarter in history creates an unavoidable contrast. Burry's thesis rests on supply from Samsung, SK Hynix, and Micron eventually catching up to demand. Micron's management spent an entire earnings call explaining in granular detail why that supply is not arriving through at least 2027 and likely not through 2028. The SCAs, the committed 75% of 2027 output, the new upside requests from customers that management specifically named: these are not the conditions in which a supply flood materializes. They are the conditions that precede sustained pricing power.
The counterargument is that Nvidia's CFO already characterized memory pricing as "extreme" at Nvidia's own August earnings call. What Nvidia called extreme, Micron is now guiding will get more extreme through 2027. That framing creates real near-term tension: the highest-revenue AI infrastructure customer is absorbing higher memory costs that compress its own margins, which creates a ceiling on how far memory pricing can run before it starts distorting the downstream demand it depends on.
What This Quarter Confirms for the AI Infrastructure Complex
Micron's results are a direct read-through to every name in the AI infrastructure supply chain. The company confirmed that capital expenditure from hyperscalers is arriving faster than Micron can build capacity to serve it, and that customer discussions for 2028 are already active. That is the demand environment Alphabet described when it raised its 2026 capex guidance to $195-$205 billion and cited "accelerated capacity deliveries to meet stronger demand." It is the environment that validates SK Hynix's 76% operating margin despite a Q2 revenue miss that was supply-constrained rather than demand-constrained.
Full fiscal year operating cash flow and the capex guidance for FY2027 both signal that memory infrastructure buildout is entering a new phase. The first phase was contracted take-or-pay agreements and HBM sell-out through 2026. The second phase, now underway, is expansion of physical capacity across new fabs in the US, Japan, Singapore, and Taiwan against a demand base that management explicitly cannot forecast an end date for.
The one number to hold going into Q1 FY2027: the gross margin trajectory. If Q1 delivers at or above 86.25% and the CFO's "floor" framing proves accurate, the FY2027 EPS debate resolves upward, and a stock trading at roughly 6 to 7 times forward earnings against a company generating $85 billion in annual profit becomes a straightforward re-rating candidate.
更多独家洞察尽在 nextfin.ai.
