RAM supply set to worsen, says Micron, as CEO celebrates ‘much higher’ prices

Micron said memory demand will exceed supply through at least 2028, with most of its 2027 production already sold and customers expected to pay materially higher prices. FY2026 Q4 revenue rose to $54.25B from $11.3B, annual revenue reached $133.2B versus $37.4B, and net income increased 895% to $85B; cloud-memory and datacenter gross margins expanded to 83% and 90%, respectively. The company guided for FY2027 Q1 revenue of $61.5B plus or minus $1.5B and 86.25% gross margin, supported by AI-driven HBM and datacenter SSD demand, although shares ended after-hours near their prior close.

Analysis

The investable implication is not simply higher MU earnings: a multi-year allocation market shifts bargaining power from GPU/system vendors to memory suppliers, making booked capacity and qualified HBM stacks strategic assets. MU’s valuation can sustain a higher cycle multiple if forward contracts demonstrate that pricing is contractual rather than spot-driven; the key sensitivity is whether incremental capex converts into usable HBM yield on schedule, not wafer-start announcements. The claimed financial figures should be reconciled against the company’s filed results before sizing—an unverified margin/revenue data point would make the apparent upside signal non-actionable.

For NVDA, constrained HBM is both a moat and a throughput risk. Secured supply can preserve system availability versus AMD and smaller accelerators, but rising memory content raises GPU bill-of-materials and potentially limits unit growth if cloud customers hit AI-capex budgets; this is more likely to show up over 1-3 quarters than immediately. Dell and HPE are plausible second-order losers because they compete on system delivery while absorbing component inflation, whereas WDC and SNDK could benefit over 6-18 months if AI inference drives durable SSD demand rather than a one-off inventory build.

Consensus may be underestimating the risk that scarcity-induced margin peaks trigger a capital-spending response across MU, Samsung Electronics and SK Hynix. The contrarian view is that 2028 tightness is too distant to capitalize today: memory equities historically re-rate on the next supply addition, customer inventory digestion, or a single hyperscaler capex reset well before physical balance returns. Watch HBM qualification/yield, contract-price renewals, hyperscaler capex guidance, and NAND inventory days rather than headline supply forecasts.

Key Decisions for Investors

• Maintain or initiate a 3-6 month long MU only after validating reported results and forward-booking disclosures in primary filings; target a 15-20% upside from multiple expansion if FY27 gross-margin guidance holds, with a 8-10% stop or exit on any material HBM yield/qualification delay.
• Express the component-inflation spread through long MU / short HPE or DELL over the next two earnings cycles. The thesis is memory pricing passing through faster at the supplier than at server integrators; cover the short if either OEM demonstrates sustained gross-margin expansion despite higher memory costs.
• Keep NVDA as a quality long but do not add solely on the memory-scarcity narrative. Add after evidence of secured HBM supply and maintained shipment guidance; reduce if cloud-capex guidance weakens or accelerator lead times lengthen despite demand, which would indicate supply is capping revenue conversion.
• Place a 6-12 month watch alert on WDC and SNDK rather than chase immediately: initiate only if enterprise SSD pricing and exabyte shipments rise concurrently for two reporting periods. A rebound driven solely by NAND price inflation without shipment growth would be a late-cycle warning, not confirmation of structural AI storage demand.