Micron's 2027 HBM Capacity Is Sold Out; Stock Trades 26% Below Peak

Micron has sold out its high-bandwidth memory (HBM, the stacked DRAM used in AI accelerators) capacity through 2027. The stock sits 26 percent below its peak.

Those two facts frame the trade. On one side: a fully booked order book for the industry's most strategic memory product, stretching three calendar years out. On the other: a share price that hasn't followed.

The gap between the two is the story. Investors have priced AI infrastructure heavily across the semiconductor complex, yet Micron's quotation suggests the market still treats the company as a cyclical memory producer rather than a structural AI beneficiary. The sold-out 2027 book directly challenges that framing.

What does sold-out capacity actually change?

It changes visibility. Memory is normally a spot-and-quarter business — pricing moves with inventory cycles, and suppliers rarely see demand beyond the current year with any confidence.

HBM is different. Because the product goes into AI accelerator modules designed years ahead, customers commit early and commit large. Selling out 2027 this far in advance means Micron's revenue for that capacity is effectively underwritten before a single wafer is run.

That has two practical consequences:

• Pricing power. When capacity is booked solid three years out, the supplier sets terms. Micron doesn't have to discount to fill fabs.
• Capital-allocation certainty. Micron can commit capex to HBM capacity expansion knowing the output is already spoken for.

What does this change for the stock?

The 26 percent discount to its peak is the reader's practical question, and the honest answer runs two ways.

The bull case: the market hasn't repriced Micron for the durability of its HBM franchise. If the 2027 sellout holds — and if HBM margins behave more like logic silicon than commodity DRAM — the discount closes. Simple math, big swing.

The bear case: the discount exists for reasons the order book can't fix. Memory investors have been burned before. Capacity cycles turn, AI spending pauses hit suppliers first, and a sold-out book tells you nothing about pricing in 2028, only that 2027 is full.

Both cases accept the same fact pattern. They disagree on the multiple.

Why HBM is the underwriting argument

HBM sits at the intersection of two hard constraints in AI compute: memory bandwidth and memory capacity. Accelerators are only as fast as the memory feeding them, and HBM — stacked die wired in a wide, fast interface — is how system designers solve that bottleneck.

That's why the product supports multi-year commitments at all. Buyers aren't speculating; they're matching HBM supply to accelerator roadmaps already locked for 2026 and 2027 launches.

For Micron, this underwrites the AI narrative in the literal sense. The company's AI exposure isn't a projection or a market-share hope. It's contracted.

The disconnect, in one line

Sold out through 2027 on the product side. Trading 26 percent below peak on the price side.

If the market comes around to treating multi-year HBM backlog as durable rather than cyclical, that spread is the opportunity. If the cycle turns first, the discount was earned. Either way, the order book — not the narrative — is now the thing to watch.