Intel Server Supply Exceeds Expectations, but Long-Term Market Share Challenges Remain

Core Conclusion

Intel’s Q1 2026 results and Q2 guidance materially beat consensus—DCAI revenue $545M above estimate, gross margin 41% vs. 34.9%—driven by supply-constrained server CPUs and ASIC strength. However, this cyclical tailwind masks structural share erosion: key AI customers (NVIDIA, AWS Trainium, Google TPU) are migrating head-node CPUs away from Intel to ARM or in-house designs. Product leadership in servers is not expected before end of CY27. At ~53x forward EPS, risk/reward is balanced; we maintain Equal-weight with a $73 target, preferring memory stocks with longer-duration supply constraints and no share risk.

What the Market May Be Underpricing

The market extrapolates near-term CPU shortages into sustained revenue momentum, while ignoring the pace of share loss in AI data centers. Management’s CPU-to-GPU ratio improvement narrative (8:1→4:1) is misleading—Bianca cards already have a head-node CPU per two GPUs plus host node, and the vast majority of Intel CPUs sit outside AI clusters. Most important CPUs in AI face share headwinds: NVIDIA moving some racks to Vera (ARM), Trainium shifting head node to ARM, both TPU and Trainium moving host node from x86 to ARM. Within x86, AMD is likely to favor its own CPUs and can secure additional wafers.

Evidence Chain

• Q1 outperformance: Non-GAAP revenue $13.577B (MS est $12.699B), DCAI $5.052B (+22% y/y, $545M above), CCG $7.727B (+1% y/y, $223M above). Gross margin 41% (MS est 34.9%). EPS $0.29 (MS est $0.03). Q2 guidance midpoint $14.3B vs. Street $13.105B, gross margin 39% (Street 36%). Management says undershipping demand by >$1B.
• Foundry losses stubborn: $2.4B in Q1, flat over four quarters despite revenue growth. Breakeven target of 7 quarters away appears challenging; any new foundry wins would push it back. No clarity on Terafab structure—Tesla builds fabs, Intel’s value capture unclear.
• Product leadership gap: 18A ramping for notebook client, but no performance leadership in desktop until year-end, server until CY27. Internal foundry advantage not yet proven in flagship data center CPU.
• Customer shifts: ASIC strength in Q1 ($325M, up 30% q/q) continues in Q2, but these are mostly custom chips. NVIDIA partnership positive for NVLINK/CPU integration, but bulls immediately assume foundry component—management insists focus is product.
• EPS revisions: CY26 EPS raised to $1.12 from $0.68; CY27 to $1.73 from $1.34.

Key Divergences and Risks

• TAM expansion overstated: CPU is more important for agentic workloads, but the volume base is small relative to total Intel CPU shipments. AI growth applies to a minor portion of units.
• AMD share resilience: AMD likely articulates ability to get more wafers; no expectation of conceded share loss over intermediate term.
• Foundry breakeven risk: Any future foundry wins require further capex, delaying breakeven. Stock issuance for CHIPS Act was effectively Intel issuing equity for already-committed funds.
• Downside scenario: Cycle turns in 2027, share losses continue. Bear case $43 (40x CY27 EPS $1.07).
• Upside catalyst: Sustained shortage beyond 12 months, product roadmap execution, foundry partnerships de-risk the story. Bull case $110 (40x CY27 EPS $2.75).

Valuation and Trading Implications

Price target raised to $73 (from $56) by applying 42x to CY27 EPS $1.73—above the high end of large-cap logic peers, reflecting operating leverage potential and foundry optionality. At $66.78, stock trades at ~60x CY26 EPS. Leverage in the model is high given depressed margins, but we remain cynical on longer-term market share. We see better risk/reward in memory (e.g., Micron) where supply constraints have longer duration and no share headwind. Maintain Equal-weight; would require evidence of server performance leadership to become more constructive.

Appendix: Key Estimate Changes (MS ModelWare)