Embedded design wins in 2025: $17B, up ~20% YoY; cumulative since Xilinx acquisition >$50B (CEO)
Official guidance Q1 2026: revenue ~$9.8B ± $300M (including ~$100M MI308 China sales); non-GAAP gross margin ~55%; non-GAAP opEx ~$3.05B; non-GAAP other net income ~$35M; effective tax rate 13%; diluted shares ~1.65B (CFO)
Long-term targets: data center segment revenue >60% annual growth over 3–5 years; AI business to tens of billions in annual revenue in 2027; total revenue >35% CAGR over 3–5 years; annual EPS >$20 in strategic timeframe (CEO)
Prepared Metrics
Metric
Value
Speaker/Context
Q4 data center segment revenue
$5.4B
CFO – record, 39% YoY
Q4 client business revenue
$3.1B
CFO – record, 34% YoY
Q4 diluted EPS
$1.53
CFO – record, 40% YoY
Q4 free cash flow
$2.1B
CFO – record
Full-year 2025 revenue
$34.6B
prepared – 34% YoY
Full-year 2025 EPS
$4.17
CFO – record, up 26% YoY
Q1 2026 revenue guidance midpoint
~$9.8B
CFO – ±$300M
Q1 2026 non-GAAP gross margin guidance
~55%
CFO
Q&A Batch (1-5 of 12)
Q1 — Aaron Rakers
Topic: MI450/Hellios ramp and 2027 AI revenue trajectory
Key points:
MI450 development on track for second-half launch and production start.
OpenAI relationship is strong; ramp starts H2 2026 into 2027 and is on track.
Multiple customers interested in fast MI450 ramp across both inference and training.
Mgmt endorses “tens of billions of dollars” of data center AI revenue for 2027.
Mgmt stance: Bullish — product progress and customer engagement proceeding very well.
Q2 — Tim Arcuri
Topic: Q1 guidance details and data center GPU ramp shape in 2026
Key points:
Q1 revenue guided down ~5% sequentially; data center overall is up.
Server CPU (seasonal) down high-single-digit %; client, embedded, gaming all decline seasonally.
Data center GPU (including China) up sequentially in Q1.
MI355 ramps in H1; MI450 starts revenue in Q3, significant volume in Q4 2026.
Full-year: server CPU order book strengthened over last 60 days; strong growth throughout year.
Mgmt stance: Bullish — “very bullish on the year”; data center has strong growth vectors.
Q3 — Vivek Arya
Topic: China MI308 sales outlook; possibility of 60%+ data center growth in 2026
Key points:
China MI308: $100M revenue forecast for Q1; no additional China revenue forecast beyond Q1 due to dynamic situation.
Licenses for MI325 submitted; no additional revenue assumed.
Data center drivers: EPYC (Turin, Genoa) plus Venice launch in H2; MI450 ramp in H2.
Long-term target of >60% data center growth is “certainly possible” in 2026.
Mgmt stance: Bullish — major growth drivers in CPU and GPU make 60%+ possible.
Q4 — CJ Muse
Topic: Server CPU capacity, supply tightness, and gross margin progression through 2026
Key points:
Server CPU TAM expected to grow strong double-digits in 2026, driven by AI/CPU relationship.
Supply capacity for server CPUs already increased, enabling higher Q1 server guide.
Q1 gross margin guided at 55% (+130bp YoY), driven by favorable mix (Turin, MI355, client upstack, embedded recovery).
MI450 ramp in Q4 will reduce gross margin temporarily due to mix; more color later.
Topic: MI450 form factor, rack revenue recognition, and ramp risk
Key points:
MI450 series includes an 8-way GPU form factor, but vast majority in 2026 will be RackScale solutions.
Revenue recognized when AMD ships to rack builder.
Development and testing (silicon + rack scale) on track; customer input helps parallel testing.
Expect second-half launch to be on schedule; no specific ramp risks called out.
Mgmt stance: Neutral/bullish — “so far, so good”; no issues flagged, expects to be on track.
Q&A Batch (6-10 of 12)
Q6 — Stacy Rasgon
Topic: OpEx trajectory and leverage; China revenue margin impact; 2025 Instinct revenue
Key points:
OpEx in 2025 grew as revenue increased; mgmt expects OpEx to grow slower than revenue in 2026, especially in H2 with revenue inflection.
$100M China revenue in Q1: mgmt did not confirm if it drops through at zero cost basis or is a margin headwind (Jean Hu began answering but cut off).
Mgmt declined to provide 2025 Instinct annual revenue, but noted Q4 data center AI number (ex-China) still showed Q3-to-Q4 growth.
Mgmt stance: Bullish on OpEx leverage in 2026, citing free cash flow generation and revenue growth; neutral on China revenue specifics (no direct answer given).
Q7 — Lisa Su (continuation)
Topic: Data center AI revenue modeling (follow-up to Q6)
Key points:
No business-level guidance provided; but Q4 data center AI number (excluding China’s non-recurring ~$100M) still showed sequential growth from Q3 to Q4.
Mgmt stance: Neutral – provides directional help for models but no specific numbers.
Q8 — Joshua Buchalter
Topic: Client segment growth; PC TAM and inflationary memory costs; Instinct competitive positioning vs. SRAM/ASIC architectures
Key points:
Client segment performed well in 2025 with strong ASP mix and unit growth.
For 2026: PC TAM expected down slightly due to commodity inflation (including memory); mgmt models H2 as “a bit subseasonal” vs H1; AMD expects to grow PC business despite market decline, focusing on enterprise and premium.
On Instinct vs. SRAM/ASIC: mgmt views workload-optimized products as natural evolution; AMD’s triplet architecture allows optimization across inference/training; inference is a significant opportunity alongside training ramp.
Mgmt stance: Bullish on client growth in 2026 despite TAM headwinds; neutral on competitive architecture shifts (sees AMD’s flexible stack as well-positioned).
Q9 — Ben Reitzes
Topic: OpenAI MI450 ramp timeline; x86 vs. ARM in CPU market (agentic workloads)
Key points:
MI450 ramp for OpenAI is on schedule to start in H2 2026; Helios platform doing well; deep co-development ongoing.
Multiple other customers also excited about MI450 series and ramping in same timeframe.
CPU market: high-performance CPUs needed for agentic workloads (AI agents driving traditional CPU tasks); EPYC optimized for cloud/enterprise; mgmt sees multiyear CPU cycle as part of AI infrastructure.
Mgmt stance: Bullish on MI450 ramp and broad customer base; bullish on EPYC’s role in AI infrastructure vs. ARM.
Q10 — Tom O'Malley
Topic: Memory procurement timelines (HBM); system architecture evolution (rack-scale vs. other form factors)
Key points:
HBM and wafer procurement done over multiyear time frame; mgmt has been planning for significant CPU and GPU ramp for past couple of years; multiyear agreements extend beyond 2026.
System architecture: no one-size-fits-all; rack-scale architecture is best for high-end inference/training; enterprise AI may use other form factors; AMD investing across spectrum.
Mgmt stance: Bullish on supply chain readiness for 2026 growth; neutral on architecture direction (flexible approach).