tapebrief
AMD · Q3 2025
AMD · Q1 2026

AMD · Q4 2025 Earnings

Bullish

AMD

Reported February 3, 2026

30-second summary

30-second take: Q4 revenue grew 34% YoY to $10.27B, beating the high end of the prior guide by $370M with Data Center +39%, Client +34%, and Gaming +50%; headline non-GAAP gross margin printed 57% vs. ~54.5% guide, but ex-MI308 inventory release and China revenue, adjusted GM was ~55% — roughly in line with the guide. Q1 FY2026 is guided to $9.5–10.1B (mid $9.8B), implying +32% YoY at the midpoint with a typical seasonal step-down in client/gaming/embedded, plus a ~$290M MI308 China headwind (from $390M in Q4 to ~$100M in Q1). Underneath the optics, management re-endorsed >35% revenue CAGR over 3–5 years and a tens-of-billions AI revenue path for 2027, with Data Center segment guided to grow sequentially into Q1 despite normal seasonality elsewhere.

Headline numbers

EPS

Q4 FY2025

$1.53

Revenue

Q4 FY2025

$10.27B

+34.0% YoY

Gross margin

Q4 FY2025

54.0%

Free cash flow

Q4 FY2025

$2.08B

Operating margin

Q4 FY2025

17.0%

Key financials

Q4 FY2025
MetricQ4 FY2025YoYQ3 FY2025QoQ
Revenue$10.27B+34.0%$9.25B+11.1%
EPS$1.53—$1.20+27.5%
Gross margin54.0%—52.0%+200bps
Operating margin17.0%—14.0%+300bps
Free cash flow$2.08B—$1.53B+36.1%

Guidance

AMD beat Q4 FY2025 guidance significantly across revenue, margins, and growth rates, but Q1 FY2026 guidance implies material sequential contraction and a deceleration to 32% YoY growth, signaling headwinds from China export controls on AI accelerators.

Guidance is issued one quarter forward. The Prior-guide column references the guide issued last quarter for the period just reported; the New-guide column is for next quarter.

Actuals vs prior guidance

MetricPeriodPrior guideActualΔResult
RevenueQ4 FY2025$9.3B–$9.9B (midpoint $9.6B)$10.27B+$0.37B above high end of guideBeat
Non-GAAP Gross MarginQ4 FY2025approximately 54.5%57%+250 bps above guideBeat
Year-over-year Revenue GrowthQ4 FY2025approximately 25%34%+9 percentage points above guideBeat
Sequential Revenue GrowthQ4 FY2025approximately 4%11%+7 percentage points above guideBeat

New guidance

MetricPeriodGuideYoY
Non-GAAP Gross MarginQ1 FY2026approximately 55%—
MI308 China SalesQ1 FY2026approximately $100 million—
RevenueQ1 FY2026$9.5B–$10.1B (midpoint $9.8B)approximately 32%

Segment performance

Q4 FY2025
SegmentQ4 FY2025YoY
Data Center$5.38B+39.0%
Client$3.097B+34.0%
Gaming$0.843B+50.0%
Embedded$0.95B+3.0%
Data Center Revenue (Q4)$5.38B—
Data Center YoY Growth39%—
Client & Gaming Revenue (Q4)$3.94B—
Client & Gaming YoY Growth37%—

Capacity & utilization

Q4 FY2025
SegmentQ4 FY2025
AMD Instinct MI308 Revenue to China (Q4)$390M

Profitability

Q4 FY2025
SegmentQ4 FY2025
Non-GAAP Gross Margin57%
Non-GAAP Operating Margin28%
Free Cash Flow Margin20%

Management tone

Q2 "Client and Gaming carry while Data Center stalls" → Q3 "Record everything, Q4 guide sustains strong YoY growth" → Q4 "AI inflection point with multi-year demand super cycle"

Two quarters ago AMD refused to quantify MI revenue and posted a $155M Data Center operating loss tied to the MI308 write-down; one quarter ago Data Center returned to $1.07B of operating income with the first concrete OpenAI gigawatt timing; this quarter Data Center grew 39% YoY to $5.38B with $1.75B of segment operating income, and management is anchoring publicly to "tens of billions in annual AI revenue in 2027" and ">60% annual data center growth over 3–5 years." The press release statement that "our AI business is accelerating, with the launch of MI400 series and Helios representing a major inflection point" is the most forward-committed language AMD has used in this cycle — a clear escalation from Q3's "clear path to targets." The shift signals management believes the supply chain, the customer roster, and the product roadmap are now aligned enough to bear public quantification, not just directional confidence.

The CPU narrative has shifted across three quarters from background share-gain story to co-equal AI infrastructure pillar. Q2's commentary framed EPYC strength as enterprise-driven and ASP-led; Q3 added the hyperscaler-AI-spawned general-purpose compute thread; this quarter management states "demand for EPYC CPUs is surging as agentic and emerging AI workloads require high-performance CPUs to power head nodes and run parallel tasks alongside GPUs." That is a material repositioning — AMD is now claiming CPU and GPU are jointly compounding on the same AI demand curve, rather than CPU benefiting incidentally. The CFO reinforced this by guiding Q1 server CPU revenue up sequentially in a normally seasonally-down quarter, and Lisa Su flagged that the server CPU order book has strengthened "especially over the last 60 days." If correct, it lifts the long-range Data Center revenue base meaningfully above what a pure-GPU TAM model would suggest.

The Instinct adoption story has broadened materially in tone. Q2 was OpenAI-anchor-tenant with HUMAIN as the sovereign proof point; Q3 added Oracle as a named MI450 launch partner; Q4 cites "eight of the top 10 AI companies use Instinct to power production workloads" plus multiple neocloud providers launching MI350 offerings. Combined with the MI400 portfolio expansion to MI455X (superclusters), MI430X (HPC/sovereign), and MI440X (enterprise), and HPE and Lenovo publicly announcing Helios rack plans, the customer concentration concern that dominated Q3's Q&A appears to be receding.

The MI308 China disclosure is the one piece of harder language. Breaking out the $390M Q4 contribution and pre-telegraphing $100M for Q1 is a new transparency posture that didn't exist in Q2 or Q3. The implication is that AMD now expects the China line to be regulatorily volatile enough that investors need to see it discretely — a tacit acknowledgement that export controls are the binding variable on near-term Data Center revenue smoothness.

Recurring themes management leaned on this quarter:

AI infrastructure inflection with MI450/Helios as watershed momentCPU criticality in agentic AI workloads alongside GPU accelerationEnterprise adoption acceleration across data center, PC, and embeddedMulti-generational product roadmap execution and supply chain planningOperating leverage and margin expansion despite aggressive OPEX investmentBroadening Instinct ecosystem from training to inference across multiple form factors

Risks management surfaced:

China licensing dynamics and MI325/MI308 regulatory uncertaintySupply chain tightness for HBM and wafer capacity despite mitigation effortsCompetitive threats from NVIDIA's ARM CPU offerings and SRAM-based inference architecturesGaming semi-custom decline in 2026 due to seventh-year console cycle maturityPC market TAM contraction from inflationary commodity and memory pricing pressures

Answers to last quarter's watch list

Whether Q4 prints another large beat vs. the ~$9.6B guide — Q4 revenue of $10.27B was a meaningful beat vs. the $9.6B midpoint, exceeding the high end of the range by $370M.
Resolved positively
Q4 Data Center segment growth rate — Data Center grew 39% YoY in Q4, a re-acceleration from +22% in Q3. Note approximately $390M of the segment came from MI308 China sales; ex-China growth was approximately 30% YoY, still meaningful re-acceleration but less than the headline.
Resolved positively
Data Center operating margin trajectory — Data Center segment operating income was $1.75B (33% margin) vs. $1.16B (30%) YoY — meaningful margin expansion driven by higher revenue and the MI308 inventory reserve release, partially offset by continued AI hardware/software investment.
Resolved positively
Named MI450 anchor tenants beyond OpenAI and Oracle — The press release cites "eight of the top 10 AI companies use Instinct to power production workloads" and management referenced "a number of other customers" in active discussions on at-scale multi-year MI450 deployments, plus public Helios commitments from HPE and Lenovo. Specific named MI450 hyperscaler commitments beyond OpenAI were not disclosed, but the customer breadth signal strengthened materially.
Continue monitoring
Analyst Day (Nov 11) TAM and long-range financial targets — Management reaffirmed the ">35% revenue CAGR over 3–5 years," "tens of billions in AI revenue in 2027," and ">$20 annual EPS" targets from Analyst Day, and added ">60% annual data center growth" framing with Lisa Su noting this is "certainly possible in 2026." The long-range structural case is intact and quantified.
Resolved positively
Embedded segment inflection — Embedded grew +3% YoY in Q4 and +11% sequentially, ending the YoY decline streak after five consecutive quarters and confirming the inflection management guided to.
Resolved positively

What to watch into next quarter

Whether Q1 prints above the $9.8B midpoint despite the $290M MI308 step-down — a print above $10.0B would suggest underlying ex-China Data Center is still accelerating and the guide carries the usual conservatism; below $9.8B is the first material miss in this cycle.

MI308 China line item disclosure cadence — AMD broke out $390M Q4 and ~$100M Q1 guide, and is forecasting zero beyond Q1. Watch whether the Q1 actual lands at $100M, $0, or higher, and whether MI325 licenses come through — both are direct read-throughs on export policy direction.

Ex-China Data Center YoY growth — Q4 was approximately 30% ex-China. With Q1 MI308 China dropping to ~$100M, the headline number compresses but the underlying line should sustain or accelerate; watch for 30%+ ex-China growth as the validation point.

Server CPU sequential strength and supply — Q1 server CPU is guided up sequentially against normal seasonal decline, with management adding TSMC capacity. Watch whether the order-book strength Lisa Su flagged over the last 60 days converts to continued sequential server CPU growth through Q2, and whether pricing/mix inflects with Venice launch in 2H.

Non-GAAP gross margin trajectory through Q1 and Q2 — Q1 guided to 55% (flat vs. Q4 adjusted, +130bps YoY). Watch whether Q2 guide holds 55% or compresses as MI450 ramp mix begins, and how Hu frames the 2H GM cadence — Q4 GM will be "driven largely by mix" per the CFO.

Named MI450 hyperscaler commitments at H1 2026 events — the broadening Instinct customer narrative needs at least one additional hyperscaler-scale MI450 commitment in H1 2026 to fully defuse the OpenAI-concentration concern flagged in Q3.

First MI450/Helios production revenue disclosure — MI450 revenue starts in Q3 2026 and ramps to "significant volume" in Q4 per Lisa Su. Any pull-forward signal in Q2 commentary materially de-risks the 2027 "tens of billions" anchor; any slip is the first crack in the multi-year story.

Sources

  1. AMD Q4 FY2025 press release (SEC filing): https://www.sec.gov/Archives/edgar/data/2488/000000248826000014/q42025991.htm
  2. AMD Q4 FY2025 earnings call transcript (prepared remarks and Q&A)
  3. Tapebrief AMD Q3 FY2025 brief
  4. Tapebrief AMD Q2 FY2025 brief

Get the next brief, free.

We publish analyst-grade earnings briefs the same day or morning after every call — headline numbers, segment KPIs, Q&A highlights, and tone analysis. Free during beta.

This is not investment advice.