The Next Micron Downturn Is Already Under Construction

The prevailing thesis on Micron is simple: memory is no longer cyclical. The company has locked up its 2026 high-bandwidth memory supply under long-term contracts, revenue tripled last quarter, and margins are above 70%. The stock has surged roughly 680% over the past 12 months, crossing $1 trillion in market capitalization — the second memory company in history to reach that level.

The conclusion that investors have drawn is that AI demand has fundamentally changed the memory business, and the old boom-bust cycle is behind Micron.

That conclusion is increasingly wrong. It is not that memory has stopped being cyclical. It is that the cycle has been pushed forward by a mechanism that investors are not pricing: the capital expenditure Micron is spending now will create the oversupply that ends this upcycle. The timeline, not the demand, determines when the music stops.

The supply mechanism behind the earnings explosion

Memory pricing has not been driven by unit demand growth. It has been driven by supply discipline and structural capacity reallocation.

The three major DRAM manufacturers — Micron, Samsung, and SK Hynix — collectively limited expansion after the 2022 downturn. Then came HBM, the specialized memory that sits beside every AI accelerator chip. HBM requires roughly three times the wafer capacity of standard DRAM to produce. Every unit of HBM Micron ships directly displaces conventional DRAM supply. The result is not just an AI memory shortage. It is a general memory shortage that has pulled pricing up across every segment.

Goldman Sachs projects the DRAM market will swing to a 5.9% undersupply by 2027, with NAND reaching a 4.6% deficit. These are the widest shortages since the cycle of 2017. Inventories sit at multi-year lows and lead times for larger orders have extended beyond 40 weeks.

Micron's earnings reflect this supply squeeze. Fiscal second quarter revenue reached $23.86 billion, up 196% from the year-ago quarter. Gross margins expanded to roughly 75%, more than double the 36.8% rate a year earlier. Non-GAAP earnings per share were $12.20, nearly 40% above analyst consensus. The third quarter guidance projects revenue of approximately $33.5 billion and gross margins near 81%.

The market has responded as though this is a permanent regime change. Micron now trades near $1,000 per share, up 256% year-to-date.

The capital expenditure trap

The mechanism that will end this cycle is not a demand collapse. It is the infrastructure being built right now.

Micron's capital expenditures have accelerated at a pace that tells the story more clearly than any management comment. Fiscal 2025 capex was just over $13 billion. The full-year 2026 guidance was raised to over $25 billion, roughly double the prior year. For fiscal 2027, spending is expected to exceed $45 billion. Construction costs alone account for more than half of the fiscal 2027 increase, with construction spending projected to consume the entirety of fiscal 2026's capex budget in under three quarters.

Metric	FY2025	FY2026	FY2027
Capital Expenditures	~$13B	~$25B	>$45B
Gross Margin	37%	58%+	—
Revenue	~$26B	~$54B	—

This is not incremental equipment spending. This is factory construction. Micron's long-term U.S. investment plan totals $250 billion. The Idaho facility, initially funded under the CHIPS Act, is expected to reach initial production by mid-2027. The $100 billion New York campus is targeting wafer output in the second half of 2028. Manufacturing projects are also under development in Japan.

Here is what investors should understand: the capacity being ordered today does not produce chips tomorrow. New fabrication facilities require 12 to 18 months to bring online at the fastest, with advanced-node cleanroom-to-production timelines extending to three years or more. Micron management itself has stated it has no line of sight on when the current tight supply will ease.

That is the trap. The capex is accelerating precisely when margins are at their peak. The factories being poured today will flood the market in 2028 and beyond. And the stock is already trading at a valuation that assumes this earnings regime endures.

The contract illusion

Micron management has presented the company's new contract structure as a solution to cyclicality. In June, the company disclosed 16 multi-year Strategic Customer Agreements, with 14 of them locking in approximately $100 billion in minimum revenue through 2030 under take-or-pay terms. Customers have provided roughly $22 billion in deposits and related financial commitments.

The contracts include pricing floors designed to keep gross margins above any prior-cycle peak. Management expects roughly half or more of company revenue to be covered by these agreements.

This is real structural improvement. The earnings trough in the next downturn will be materially higher than in cycles past. Customers have assumed some of the demand risk that used to fall entirely on manufacturers.

But the contracts do not eliminate cyclicality. They raise the floor. They do not cap the ceiling. During extreme shortages like the one happening now, some agreements use fixed prices or price bands, meaning Micron forfeits some of the upside it is currently capturing. More importantly, the contracts guarantee volume — but they cannot guarantee that the next cycle will not see the three manufacturers simultaneously pouring new capacity into a market that contracts. When oversupply arrives, pricing floors will cushion the fall but will not prevent it.

The valuation disconnect

At roughly $1.15 trillion in market capitalization, Micron is priced for durability. The forward P/E based on current consensus estimates appears cheaper — but only because those estimates assume these margins persist through fiscal 2027 and beyond. Non-GAAP EPS of $12.20 in fiscal Q2 is extraordinary for a memory company. Those numbers are not a baseline. They are a peak.

The earnings estimates show the trajectory that has analysts updating their models: Micron reported $12.20 per share in fiscal Q2, and consensus for the next reported quarter is already near $20. The market is pricing in a compounding machine. But memory companies have never compounded at this level for more than two or three quarters before the cycle turned.

The question is not whether Micron will earn $100 billion or $150 billion this fiscal year. The question is what happens when the $45 billion capex program of fiscal 2027 transitions from construction costs to wafer output two years later. The market is discounting a supply response that the capex numbers make inevitable.

What changes the investment case

The key issue is not whether AI demand for memory will persist. It will. The more important question is whether the supply response will arrive faster than the demand growth can absorb it.

Three conditions determine the forward path. First, whether the other two manufacturers — Samsung and SK Hynix — maintain their current supply discipline or decide that 70% gross margins are an invitation to expand. Second, whether the construction timelines for new facilities slip, extending the shortage window. Third, whether the Strategic Customer Agreements hold if hyperscaler capex slows. The five largest cloud providers are collectively spending approximately $755 billion on infrastructure in 2026, but that number can revise down as quickly as it revised up.

If supply discipline holds and construction runs behind schedule, this cycle extends further and the current valuation is justified. If any of the three manufacturers accelerates capacity, or if the new fabs come online as planned, the market faces a structural inflection around 2028 that the stock price has not yet priced in.

Memory has always been cyclical. The contracts make the next trough less brutal. They do not make the cycle disappear. The capex numbers suggest the next turning point is already under construction.