Micron estimates raised by Baird on surging DRAM pricing and AI demand

Baird raised its Micron price target to $500 from $443, kept an Outperform rating, and now models EPS of ~$60 for fiscal 2026 and ~$100 for fiscal 2027. The firm expects DRAM pricing to more than double QoQ in Q1 and rise another >40% in Q2 amid tightening supply and surging HBM demand for AI; HBM is forecast to be ~11% of DRAM production in 2026 and data centers ~53% of DRAM demand. Baird projects server DDR5 gross margins to average ~85% in 2026 (approaching 90% by year-end) and NAND margins rising to ~70% in Q2 and high-70s by year-end, noting DRAM bit growth of ~25% in 2027 while advanced packaging capacity (COWOS) could grow ~60% YoY.

Analysis

The market is treating memory as a two-speed industry: a concentrated, high-value HBM stream tied to AI infrastructure and a commoditized DRAM/NAND stream tied to legacy client devices. That bifurcation creates asymmetric payoffs—firms with scale and advanced packaging integration capture outsized margin expansion while commodity-focused suppliers face demand erosion in PCs/phones and inventory markdown risk.

Because demand is concentrated among a handful of hyperscalers and accelerator vendors, inventory dynamics will amplify volatility: deliberate build cycles by a few buyers can create multi-quarter tightness followed by abrupt destocking once capacity and packaging come online. This concentration also creates single-event catalysts—each hyperscaler guide or major GPU architecture announcement can swing pricing and utilization far more than broad end-market indicators.

On the supply side, the fastest lever to relieve HBM tightness is packaging substrate and OSAT throughput, not wafer starts alone; that creates a multi-year lead/lag where bit growth and usable HBM supply can diverge materially. Capital allocation will respond to the current margin signal, risking a cyclical overshoot in 2027–2028 where nominal bit growth outpaces actual high-bandwidth demand.

Geopolitics and tooling cadence are tail risks: export controls or equipment bottlenecks (EUV/packaging tools) can extend premiums, while faster-than-expected substrate capacity adds would flip the setup. Monitor hyperscaler capex cadence, packaging fab announcements, and quarterly inventory disclosures as high-signal, short-to-medium-term catalysts.

Key Decisions for Investors

• Initiate a tactical long in Micron (MU): allocate 2–3% NAV to MU equity or Jan-2028 LEAP calls to capture margin re-rating driven by HBM skew; target 50–80% upside in 12–24 months, place a hard stop at -30% to limit exposure to a sharp inventory-led reversal.
• Pair trade to isolate memory vs GPU exposure: go long MU (1.5% NAV) and sell (write) NVDA 6–9 month calls (~50% premium capture) sized to neutralize delta—rationale is to capitalize on MU’s margin leverage while hedging beta from broad AI hardware demand; expected asymmetric payoff if memory pricing outperforms GPU unit elasticity, risk = NVDA gap-up volatility.
• Hedge hyperscaler inventory risk with short-dated put spreads on hyperscalers (AMZN, MSFT, GOOGL): buy 3–6 month 5–8% OTM put spreads sized to cost ~0.5–1% NAV total. These protect against near-term multiple compression if cloud customers pause builds after heavy 2026 stocking.
• Buy select enterprise/stack-exposed longs (ORCL or META, 1–2% NAV) as convex, lower-volatility exposure to secular AI spend: these tend to steady revenue if hyperscalers pause capex and can capture incremental on-prem or application-side monetization. Target 20–40% upside in 12 months; trim into announcements of resumed hyperscaler purchasing.