Advanced Micro Devices, Inc. (AMD) - Q1 '26 Earnings Recap
1. NEWS SUMMARY
- AMD's Q1 data center segment hit $5.8B, up 57% YoY, with server CPU revenue growing 50%+ for the fourth consecutive record quarter; agentic AI is shifting CPU-to-GPU ratios toward 1:1 from the prior 1:4 to 1:8 norm, prompting AMD to double its 2030 server CPU TAM forecast from $60B to over $120B at a 35%+ CAGR, with server CPU revenue guided to accelerate to 70%+ YoY growth in Q2.
- MI450/Helios customer forecasts are now exceeding AMD's own internal plans, with multi-gigawatt pipeline expanding beyond the anchor Meta (6 GW, custom MI450-based GPU, H2 2026 shipment start) and OpenAI deals; Su expressed confidence in delivering tens of billions of dollars in annual Data Center AI revenue in 2027, and Helios rack-scale systems — positioned to rival Nvidia Grace Blackwell and Vera Rubin — are on track to ship H2 2026 with Samsung supplying HBM4 for MI455X.
- AMD disclosed it is using agent-based coding workflows internally to accelerate ROCm development cadence, citing day-0 support for Google Gemma 4, Qwen, and Kimi as software stack maturity signals; EPYC-powered cloud instances grew nearly 50% YoY to over 1,600, and AMD's new Verano CPU — purpose-built for AI infrastructure — is claimed to deliver over 2x throughput per socket versus leading ARM-based AI solutions.
- Q1 revenue was $10.25B (+38% YoY) vs. $9.89B consensus; non-GAAP EPS $1.37 vs.
- AMD stock surged ~16% on earnings day and hit an all-time high of $421.39; Goldman Sachs upgraded to Buy with a $450 target (from $240) citing agentic AI CPU tailwinds, Bernstein upgraded to Outperform with a $525 target (from $265) projecting EPS above $14 in 2027 and $20 in 2028, and BofA raised its target to $500 from $450; the stock is up roughly 76% YTD, trading at ~53x forward P/E, which Zacks flags as stretched relative to Nvidia at ~25x.
2. FINANCIAL HIGHLIGHTS
| Quarter | Period End | Revenue | YoY Growth | Gross Margin | Operating Margin | EV / Sales |
|---|---|---|---|---|---|---|
| Q2 FY2024 | 2024-06-29 | $5.8B | 8.9% | 49.1% | 4.6% | 44.7x |
| Q3 FY2024 | 2024-09-28 | $6.8B | 17.6% | 50.1% | 10.6% | 38.8x |
| Q4 FY2024 | 2024-12-28 | $7.7B | 24.2% | 50.7% | 11.4% | 26.3x |
| Q1 FY2025 | 2025-03-29 | $7.4B | 35.9% | 50.2% | 10.8% | 22.3x |
| Q2 FY2025 | 2025-06-28 | $7.7B | 31.7% | 39.8% | -1.7% | 30.3x |
| Q3 FY2025 | 2025-09-27 | $9.2B | 35.6% | 51.7% | 13.7% | 27.9x |
| Q4 FY2025 | 2025-12-27 | $10.3B | 34.1% | 54.3% | 17.1% | 34.0x |
| Q1 FY2026 | 2026-03-28 | $10.3B | 37.8% | 52.8% | 14.4% | 32.0x |
3. EARNINGS CALL HIGHLIGHTS
Essence
AMD's Q1 2026 call was unambiguously bullish, with Lisa Su projecting confidence bordering on aggression across both CPU and GPU AI infrastructure narratives. The dominant theme was a structural re-rating of the server CPU TAM — from $60B to over $120B by 2030 at a 35%+ CAGR — driven by agentic AI workloads demanding far more CPU orchestration compute than previously modeled, a narrative that caught analysts off guard and generated the most substantive Q&A of the call. On the GPU side, Su disclosed that MI450/Helios customer forecasts are now exceeding AMD's own internal plans, with multi-gigawatt pipeline expanding beyond the anchor Meta and OpenAI deals, and she raised the long-term Instinct growth target to exceed 80% CAGR into 2027. Analyst sentiment was constructive but probing on supply constraints, gross margin dilution from Instinct ramp, and competitive pressure from ARM CPUs, all of which management addressed with measured confidence. The overall posture was that AMD sees itself at a genuine inflection point, not just a cyclical upturn, with deepening co-engineering relationships and rack-scale integration positioning it as a systemic AI infrastructure partner.
Highlights
- AMD dramatically raised its server CPU TAM forecast from ~$60B (18% CAGR) to over $120B by 2030 (35%+ CAGR), explicitly attributing the revision to agentic AI driving CPU-to-GPU ratios toward 1:1 or even CPU-heavy configurations, up from traditional 1:4 or 1:8 host-node ratios — a direct signal that inference and agentic orchestration workloads are creating a new, large compute category.
- Su disclosed that MI450/Helios lead customer forecasts are now 'exceeding our initial plans' for 2027, with 'a growing number of new customers engaging on large-scale deployments, including additional multi-gigawatt opportunities' beyond the publicly announced Meta and OpenAI partnerships — raising AMD's stated confidence to 'tens of billions of dollars in annual Data Center AI revenue in 2027.'
- The Meta partnership was detailed further: AMD will deploy a custom GPU accelerator based on the MI450 architecture co-designed for Meta's next-gen AI workloads, integrated into the Helios rack-scale platform pairing Instinct GPUs with EPYC Venice CPUs, with shipments on track for H2 2026 — the first public disclosure of a custom MI450 derivative.
- AMD accelerated its ROCm development cadence by deploying 'agent-based coding workflows' internally to speed performance improvements and model support, citing day-0 support for Google Gemma 4, Qwen, and Kimi as evidence of software stack maturity — a notable admission that AI-assisted development is now part of AMD's own engineering pipeline.
- On the competitive CPU landscape, Su framed ARM as 'more point products relative to a portfolio' and positioned Venice's Verano — described as AMD's first EPYC CPU purpose-built for AI infrastructure — as delivering over 2x throughput per socket versus leading ARM-based AI solutions, signaling AMD intends to compete directly in the AI-optimized CPU segment rather than cede it.