Autoresearch: Micron HBM / HBM4 / sold-out status into Q3 FY2026 earnings (Jun 24, 2026)

Generated by /autoresearch on 2026-06-12. Focused 1-round pass (time-sensitive market topic — no Grokipedia anchor: encyclopedic sources don't carry June-2026 HBM pricing). Treat as raw material — review before promoting. Context: vault/projects/stock-market

Summary

The HBM/CoWoS bottleneck is tightening, not loosening, heading into Micron's make-or-break Q3 FY2026 call on June 24, 2026. Micron's entire 2026 HBM capacity was booked under binding contracts before the year began, and the contract structure has shifted from quarterly negotiation to three-to-five-year fixed-price supply agreements — a structural change that locks pricing and volume and is what makes the "sold out" claim durable rather than a single-quarter spike. CoWoS (the advanced-packaging gate downstream of HBM) is sold out through 2027 with an 18–24-month expansion lead time. HBM4 entered volume production in Q1 2026 — roughly a quarter ahead of guidance — shipping for Nvidia's Vera Rubin platform. The bull case is increasingly consensus; the marginal new information on Jun 24 is therefore about 2027 contract commentary, capex guidance, and the one bear tell (QoQ DRAM inventory build), not whether 2026 is sold out.

Findings

2026 sold out under binding contracts; contracts lengthening to multi-year

Micron's "entire 2026 HBM capacity was booked under binding contracts before the year began" and the company is "sold out of HBM for the next several quarters" (TechTimes, Jun 8 2026). The more important structural point: "the entire 2026 HBM supply is sold out under multi-year contracts" and "clients are now securing three-to-five year supply agreements" — a shift away from the quarterly negotiation that historically defined memory pricing cycles (Investing.com analysis). This is the load-bearing change: it converts memory from a spot-priced commodity into a contracted, visibility-rich revenue stream — the same thing the hbm-supply-bottleneck and helium-cliff-to-hbm-supply-crunch theses lean on.

Whole industry sold out; price hikes, not discounts

SK Hynix has already sold out its entire 2026 HBM supply; Micron's 2025 and 2026 HBM capacity is fully booked; and Samsung is raising HBM prices high-teens to low-twenties percent on 2026 contracts while planning a ~50% capacity surge to ~250k wafers/month by end-2026 (TrendForce, Dec 2025). The three suppliers "would rather hold supply tight and pricing high than flood the market and trigger the next bust" (TechTimes) — disciplined oligopoly behavior, the opposite of the 2022 memory bust.

CoWoS is the binding downstream gate, sold out through 2027

TSMC's CoWoS capacity "remains sold out through 2025 and into 2026, with CoWoS capacity effectively sold out through 2027," and "expanding CoWoS capacity takes 18–24 months" (Fusion Worldwide). This is exactly the hbm-cowos-as-binding-bottleneck chain — HBM and CoWoS are co-binding, and new HBM capacity (SK Hynix Yongin cluster) doesn't meaningfully add availability until 2027 (AI CERTs).

HBM4 ramped a quarter early

HBM4 "entered high-volume production in Q1 2026 — one full quarter ahead of management guidance" (Investing.com), volume shipments began March 2026, with 36GB 12-high stacks at 2.8 TB/s for Nvidia's Vera Rubin (TechTimes). HBM4's customer-specific logic base die deepens the lock-in vs. commoditized HBM3.

The Jun 24 setup — what's priced, what isn't

Fiscal Q2 (reported Mar 18) was a blowout: 74.4% gross margin, ~68% operating margin, $12.20 adjusted EPS (beat by ~33%). For Q3, the cited revenue estimates "span an unusually wide range, from about $33.7B to $40.9B" — i.e., analysts can't pin the number, which is itself the opportunity/risk. What to watch: HBM revenue growth, whether gross margin holds near the high, capex guidance, and 2027 supply-contract commentary (TechTimes). The single best bear tell is QoQ DRAM inventory buildup, which "proved prescient before the 2022 peak" (Investing.com).

Contradictions and open questions

• Revenue-estimate dispersion is huge ($33.7B–$40.9B cited) — sources don't converge on a Q3 consensus number; treat any single point estimate skeptically.
• Is the bull case now consensus? Multiple sources call MU "underpriced," but the sold-out narrative is widely reported — the edge is narrowing to the 2027 contract length/price disclosure and the DRAM-inventory tell, not the 2026 sold-out fact.
• Memory-cyclical risk remains the structural watch: the contract-lengthening blunts but does not remove the cycle; a DRAM inventory build would be the first crack.

Provenance

Rounds run: 1 (focused — time-sensitive; strong convergence across sources, early exit).

Sub-questions: (1) HBM4 qualification + pricing status; (2) sold-out 2026/2027 + contract structure; (3) CoWoS bottleneck tightening/loosening; (4) Q3 FY2026 consensus expectations.

Anchor source: none (time-sensitive market topic; Grokipedia not applicable).

URLs fetched (2 deep-fetched, 2 via search aggregation):

• TechTimes — "Micron Stock Climbs as HBM Sells Out: June 24 Earnings Decides the AI Memory Trade" (Jun 8 2026) — news — earnings preview + sold-out + margins.
• Investing.com — "Micron Faces a Rerating Moment…" — analysis — multi-year contracts, HBM4 early ramp, DRAM-inventory tell.
• Fusion Worldwide — "Inside the AI Bottleneck: CoWoS, HBM, 2–3nm through 2027" — industry — CoWoS sold out through 2027.
• TrendForce — Samsung 50% HBM capacity surge / price hikes — industry — pricing direction.
• AI CERTs — HBM supply crunch lasts until 2027 — industry — Yongin 2027 timing.

Tools used: WebSearch, WebFetch. Generated: 2026-06-12.