AMD · US
AMD.US: Q1'26 update — #2 AI supplier + EPYC dual engine
AMD posted Q1 2026 revenue of $10.3B with Data Center ~$5.78B (+57%), Non-GAAP EPS $1.37 and 55% GM. Q2 guide ~$11.2B ±$0.3B (~+46% YoY). The “#2 AI accelerator + server CPU share” thesis is clear, but training share and ROCm vs CUDA still lag NVDA; valuation looks rich ahead of the 2026-08-04 print. See also INTC and the CPU industry deep dive.
Cite a section with a deep link, e.g. /en/r/amd-us-research#thesis
Q1 2026 snapshot
- Q1 revenue
- USD 10.3B
- Data Center (+57%)
- USD 5.8B
- Non-GAAP EPS
- USD 1.37
- Non-GAAP GM
- 1
Q2 guidance
- Revenue midpoint
- USD 11.2B
- Guide YoY
- 0
- Non-GAAP GM guide
- 1
- Earnings date
- 2026-08-04
Thesis
AMD’s near-term pricing debate has shifted from PC/gaming cycles to Data Center: EPYC share recovery plus Instinct as a credible #2 AI accelerator. Q1 2026 validated that mix with $10.3B revenue and DC ~$5.78B (+57% YoY). Q2 guide of $11.2B ±$0.3B (+46% YoY) and Non-GAAP GM ~56% shows the company still tilting product mix toward higher-margin data center.
The premium holds only if:
- Instinct delivery is repeatable — rack-level shipment slope at hyperscalers / sovereign AI, not one-off benchmarks;
- ROCm cuts CUDA migration friction — training share still lags NVIDIA by a wide margin; inference and price/performance are the realistic wedge;
- EPYC keeps taking share from Intel — cash-flow ballast and share certainty under the AI narrative.
Lisa Su stresses multi-generation roadmaps and customer co-design; MI450 / Helios is the 2H26–2027 story for closing the gap with NVDA. For server CPU TAM and industry structure see the CPU industry deep dive.
Business
Model: Fabless design of x86 CPUs (EPYC / Ryzen) and Instinct GPU accelerators; Data Center is the primary growth engine.
Flagships: (1) Instinct MI — AI elasticity; (2) EPYC — share-recovery anchor; (3) Client Ryzen — PC/AI-PC base.
Competitiveness: Server CPU share vs Intel; credible #2 AI price/performance vs NVDA.
Strategy (12–24m): MI450 / Helios rack delivery; ROCm maturity; mix shift to higher-margin DC.
Value chain
Mid-stack fabless design monetizing CPU/GPU silicon; manufacturing outsourced.
Upstream: TSMC, HBM, advanced packaging. Downstream: hyperscalers / sovereign AI / enterprise. Risks: packaging/HBM bottlenecks; customer concentration — echoed in Risks.
Valuation
AI themes drove a strong YTD move in 2026. As of this note (2026-07-29), the stock sits in a rich band ahead of the 2026-08-04 Q2 print—spot prices move with the tape. Valuation already prices some MI / DC upside, so risk/reward is asymmetric around the earnings window.
Valuation frame
| Metric | Status | Read |
|---|---|---|
| 盈利增长 | 加速 | Q1 DC +57%；Q2 指引 ~+46% YoY |
| 毛利率 | 改善 | Q1 Non-GAAP 55% → Q2 指引 ~56% |
| 估值 | 偏高（Q2 前） | AI 主题已部分定价 MI / DC 高增 |
| 相对 NVDA | 折价 | 训练份额与 CUDA 生态差距仍大 |
| 催化剂 | 临近 | 2026-08-04 Q2 财报 |
| 风险收益 | 不对称 | 超预期上行 vs 不及预期回调 |
Versus NVDA: AMD still trades at a discount, but that discount maps to lower training share, a thinner software moat, and a wide GM gap (AMD Non-GAAP ~55–56% vs NVDA ~75%). Do not apply NVDA multiples mechanically.
Financial trend (~24 months)
Eight-quarter revenue and gross margin with YoY and QoQ. Semi/AI hardware seasonality is secondary to supply, ASP, and order timing.
Revenue · last 8 quarters
Interactive chart available in the reader.
Gross margin · last 8 quarters
Interactive chart available in the reader.
Operations
Q1 2026 operations
Q1 P&L and segments
| Item | Q1 2026 | Note |
|---|---|---|
| 营收 | $10.3B | DC 占比超 56% |
| Data Center | ~$5.78B（+57%） | EPYC + Instinct 双驱动 |
| GAAP EPS | $0.84 | — |
| Non-GAAP EPS | $1.37 | — |
| GAAP 毛利率 | 53% | — |
| Non-GAAP 毛利率 | 55% | 产品结构改善 |
Segment mix (illustrative)
| Segment | Revenue mix | Watchpoint |
|---|---|---|
| Data Center | ~56%+ | Q1 ~$5.78B；EPYC + Instinct |
| Client | ~20% | 锐龙 / AI PC；宏观敏感 |
| Gaming | ~12–15% | 独立显卡周期 |
| Embedded | ~8–10% | 工业 / 通信 |
Takeaways:
- DC >56% of revenue is now the growth engine and valuation anchor; Client/Gaming still fund the base but carry less investment weight.
- GM expansion reflects higher Instinct ASP mix and better server CPU structure.
- EPYC is the more certain layer; Instinct is the AI elasticity—and the volatility source.
AI accelerators
Instinct MI300 / MI350 has won some hyperscaler deployments; training remains NVIDIA-dominated. AMD’s more credible wedges:
- Inference and cost-efficient clusters — $/FLOP, open stack, multi-vendor strategies;
- System delivery — rack, networking, liquid cooling, turnkey (Helios);
- Software maturity — ROCm coverage and migration tooling.
Watch Instinct revenue mix and ASP, quarterly units, hyperscaler commentary (Microsoft, Meta, Oracle, etc.), and repeatable sovereign-AI wins. If AI CapEx plateaus, AMD’s elasticity can exceed pure-CPU names—and also exceed names with thicker software moats.
CPU / Client
EPYC keeps taking cloud/enterprise share from Intel—the steadier half of the dual engine. Versus Lip-Bu Tan’s demand-anchored, finance-first reset at Intel, AMD’s fabless product cadence still shows up on the server side. Foundry/policy context: INTC research.
Client (Ryzen) benefits from AI-PC / NPU OEM deals but still swings with macro and PC cycles. CPU cash flows buffer the AI story; valuation cannot be pure-AI.
Competition
Strategy / product / risk map
| Category | Detail | Implication |
|---|---|---|
| AI GPU | Instinct MI300/MI350；MI450/Helios 在路上 | 训练份额仍远落后 NVDA |
| 服务器 CPU | EPYC 持续抢 Intel 份额 | 确定性较高的增长源 |
| 软件生态 | ROCm vs CUDA | 迁移成本是最大壁垒 |
| 客户 | Microsoft、Meta、Oracle 等 | 大客户验证是关键 |
| 竞争 | NVDA 主导 + 云 ASIC 自研 | 推理场景竞争更激烈 |
| 估值 | AI 主题推高 YTD | Q2 不及预期回调风险 |
Three axes:
| Axis | AMD | NVDA | Intel |
|---|---|---|---|
| AI accel | MI challenger; inference / PPA more realistic | CUDA full-stack dominance | Gaudi / internal still catching up |
| CPU | EPYC / Ryzen share repair | Grace and ARM host paths | Product repair; foundry narrative heavier |
| Leadership | Lisa Su: roadmap + co-design | Platform lock-in | Lip-Bu Tan: demand-anchored, finance-first |
Since 2014, Lisa Su’s playbook has been: simplify the product set, bet on high-end x86 and GPUs, then scale into data center. The current mandate is taking AI accelerators from viable to scalable delivery. CEO-style contrast is now part of sell-side framing.
Peer comparison
| Company | Share / role | GM | Strength | Weakness |
|---|---|---|---|---|
| AMD (self) | AI #2 + EPYC share | ~55% | CPU share + MI ramp | CUDA / training gap |
| NVIDIA | AI GPU leader | ~75% | CUDA ecosystem | Premium valuation |
| Intel | x86 repair | ~39% | Installed base | Foundry / AI lag |
Management
Lisa Su CEO since 2014; CFO Jean Hu since early 2023. No material CEO/CFO change in 24m. Stability: steady — focus is scaling Instinct delivery.
Key management (24m)
| Role | Name | Since | 24m change |
|---|---|---|---|
| CEO / Chair | Lisa Su | 2014-10 | No change |
| CFO | Jean Hu | 2023-01 | No change (appointed Jan 2023) |
| EVP Technology & Eng. | Mark Papermaster | 2011 | No material disclosure change |
Outlook
Company guide · Q2 2026
| Item | Guide | Note |
|---|---|---|
| 营收 | ~$11.2B ±$0.3B | 约 +46% YoY |
| Non-GAAP 毛利率 | ~56% | 较 Q1 继续扩张 |
| 财报发布 | 2026-08-04 | 关键验证窗口 |
Framework:
- Hitting ~$11.2B would re-rate the full-year path and reinforce the #2 AI supplier narrative.
- Non-GAAP GM ~56% shows mix improvement, still far below NVDA ~75%.
- MI450 / Helios commentary (orders, sampling, capacity, software milestones) will be price-sensitive on the call.
- Any “AI slowdown” narrative would punish Instinct expectations first.
Scenarios
| Scenario | Conditions | Implication |
|---|---|---|
| Bull | Q2 >$11.5B and GM ≥56%; MI450 large wins / early ramp; EPYC share accelerates; ROCm breakthrough | #2 AI narrative strengthens; upside continues |
| Base | Q2 roughly hits $11.2B; DC stays 50%+; MI ramps steadily | Range-bound rich valuation; hold, reassess after Q2 |
| Bear | Miss guide; Instinct slows; NVDA price/architecture pressure; AI theme fades | 20–30% multiple compression possible |
Positioning frame (research only): AI satellite, not core. Prefer not to chase into Q2. Track DC/Instinct, GM, MI450/Helios, customers, EPYC share.