Micron's Supercycle Is a Contract Story, Not a Demand Story

The most quoted number in Micron's story this month is a trade, and it is the least informative one. David Tepper's Appaloosa cut its Micron stake by 41% in the quarter that ended in June, selling roughly 690,000 shares, at a moment when the stock had more than tripled in the quarter and was up about 1,300% since the fund built the position in mid-2023. A trim into a parabolic move is profit-taking, not a thesis. Tepper kept Micron as his second-largest holding, worth around $1.1 billion, while adding to NVIDIA for a sixth consecutive quarter and piling into other AI infrastructure names — a rotation within the AI trade, not out of it.

What the trading list obscures is the change that matters to a Micron holder, and it is not demand. Micron has rewritten its revenue model around fixed-price, multi-year contracts — a supply-and-pricing mechanism that changes what this cycle means for margins and for the stock.

The revenue is real, and pricing is doing the work

Micron's quarterly revenue has gone from running near $8 billion through much of 2025 to $13.6 billion in late 2025 to $23.9 billion in its most recent quarter, with earnings per share of $12.20 and gross margin near 72%. Operating margin landed around 66%. Those are not demand-side numbers; unit volumes did not climb anywhere near that fast. The lever is price — specifically the price of high-bandwidth memory, the DRAM stacked for AI accelerators, which Micron has said is sold out for 2026 and fully booked through 2027.

This is the memory-pattern point that matters. In the old model, revenue rose because the market grew and everyone added wafers, and the inevitable oversupply crushed prices. Since 2022 the three big memory makers — Micron, Samsung, SK Hynix — have capped capacity and aimed capital at technology migration rather than commodity volume, so the upcycle ran on supply discipline before it ran on demand.

What is new is that the discipline is now in contracts

Micron closed 16 Strategic Customer Agreements that together guarantee at least $100 billion in revenue through 2030, covering roughly half of its sales once fully ramped. HBM, its highest-priced product, is entirely contracted. That is a revenue floor with margins set well above prior cycle peaks — which is why Tepper, even while trimming, kept structural exposure through the long-term revenue guarantees.

Table 1 below summarizes the mechanism.

Driver	Old cycle	This cycle
Revenue engine	Market growth + capacity	Price, on shrinking capacity
Pricing	Spot, volatile	Contracted minimums to 2030
What is sold out	Nothing, eventually	HBM, through 2027
Cycle shape	Boom then price collapse	Floor on roughly half of sales

The catch is that a floor caps the ceiling

A fixed-price contract removes downside, but it also gives up upside: the agreements couple volume commitments with price controls that, cap pricing at recent peak levels. Micron is trading off the top of a historic cycle to insure the bottom. For a stock that is up about 630% over the past year and still trades near 21 times trailing earnings, with Wall Street consensus looking for roughly another 51% of upside, the question is no longer whether the floor holds — it is whether expectations keep assuming the ceiling rises too.

The old memory argument was a demand guess: trust that AI capex keeps coming. This cycle's answer is structurally different. Half the revenue is already locked at minimum prices through 2030, and capacity has been pointed at HBM rather than commodity wafers. The floor is real.

The variable that decides the case is what happens beyond the contracts

Micron is spending record capital — about $25 billion over the past year — while still generating positive free cash flow. Whether Micron and its two rivals hold the line on capacity additions is what determines whether those contracted minimums turn out to be a floor or roughly the ceiling. If capacity stays disciplined and the contract book keeps pricing near the peak, earnings can hold at these levels and the valuation is defensible against them. If the industry cracks and floods commodity DRAM again — the historical failure mode — the contract floor cushions the fall, but the 72% gross margin does not. That distinction is exactly what a fund manager's trade list cannot tell you.