tapebrief

AMD · Q3 2025 Earnings

Neutral

AMD

Reported November 4, 2025

30-second summary

SENTIMENT: Constructive 30-second take: Revenue grew 36% YoY to $9.25B in Q3 FY2025, beating the prior guide's $8.7B midpoint by $546M with Data Center +22%, Client +46%, and Gaming +181% YoY. Data Center operating income recovered to $1.07B from a Q2 loss that included an $800M MI308 inventory write-down, and the MI-308 China line was zero — so the segment is back to underlying profitability without any China contribution. Q4 is guided to $9.6B (+25% YoY, +4% QoQ), continuing the strong YoY growth trajectory; the modest +4% sequential reflects the unusually large +20% Q3 ramp versus the ~13% AMD originally guided, not an underlying demand softening.

Headline numbers

EPS

Q3 FY2025

$1.20

Revenue

Q3 FY2025

$9.25B

+36.0% YoY

Gross margin

Q3 FY2025

52.0%

Free cash flow

Q3 FY2025

$1.53B

Operating margin

Q3 FY2025

14.0%

Key financials

Q3 FY2025
MetricQ3 FY2025YoYQ2 FY2025QoQ
Revenue$9.25B+36.0%$7.68B+20.3%
EPS$1.20$0.48+150.0%
Gross margin52.0%40.0%+1200bps
Operating margin14.0%-2.0%+1600bps
Free cash flow$1.53B$1.18B+29.7%

Guidance

AMD beat Q3 FY2025 revenue and growth guidance significantly, but Q4 guidance reveals an abrupt deceleration with YoY growth expected to fall to ~25% and sequential growth to ~4%, suggesting inventory build or demand normalization despite record quarterly results.

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
RevenueQ3 FY2025$8.4B–$9.0B (midpoint $8.7B)$9.246B+$0.546B above high end of rangeBeat
Year-over-year revenue growthQ3 FY2025approximately 28% YoY36% YoY+8 percentage points above guidanceBeat
Sequential revenue growthQ3 FY2025approximately 13%20.3%+7.3 percentage points above guidanceBeat
Non-GAAP gross marginQ3 FY2025approximately 54%54%in-lineMet

New guidance

MetricPeriodGuideYoY
Non-GAAP gross marginQ4 FY2025approximately 54.5%
RevenueQ4 FY2025$9.3B–$9.9B (midpoint $9.6B)approximately 25% YoY
Year-over-year revenue growthQ4 FY2025approximately 25%
Sequential revenue growthQ4 FY2025approximately 4%

Segment performance

Q3 FY2025
SegmentQ3 FY2025YoY
Data Center$4.341B+22.0%
Client$2.75B+46.0%
Gaming$1.298B+181.0%
Embedded$0.857B-8.0%
Data Center Operating Income$1,074M
Client and Gaming Operating Income$867M

Profitability

Q3 FY2025
SegmentQ3 FY2025
Non-GAAP Gross Margin54%
Non-GAAP Operating Margin24%
Free Cash Flow Margin17%
Adjusted EBITDA$2,431M
Operating Cash Flow Margin (continuing operations)19%

Management tone

Q1 (unobserved) → Q2 "Data Center stalls, Client/Gaming carry" → Q3 "Record everything, Q4 guide sustains strong YoY growth"

Transcript prepared remarks were not provided; tone analysis is limited to Q&A and the guidance disconnect.

The clearest tonal shift this quarter is confidence on the AI ramp. Last quarter management refused to quantify MI revenue and Data Center reported a segment loss (driven by the $800M MI308 inventory write-down — underlying DC was profitable); this quarter Data Center returned to $1.07B of operating income, the MI355 ramp is described as having proceeded well with a "sharp ramp" into Q3, and Su volunteered a first-gigawatt MI450 deployment timeline of H2 2026 with OpenAI. The Q4 guide's +25% YoY growth is consistent with that confidence, even if it's 3 points below the ~28% YoY guided for Q3 — and the Q3 +$546M beat suggests management is setting a base they can comfortably clear.

On customer concentration, Su was more direct than last quarter, explicitly framing 2027-2028 supply planning around "multiple customers at similar scale" to OpenAI — a tacit acknowledgement that the OpenAI deal has shifted concentration risk meaningfully and that diversification is now an active engineering constraint, not just a talking point.

Q&A highlights

Vivek Arya · Bank of America Securities

Asked about CPU-GPU mix in Q3/Q4, transition strategy from MI355 to MI400, expectations for continued growth in H1 2026, and visibility into OpenAI allocation across multiple suppliers given power and CapEx constraints.

Lisa stated MI355 ramped well in Q3 without MI308 sales; strong server CPU demand expected into 2026 from hyperscalers forecasting significant CPU builds for AI workloads; MI355 continues ramping in H1 2026 with MI450 series launching in H2 2026 with sharper ramp; planning multiple quarters ahead with OpenAI on power and supply chain availability; first gigawatt MI450 deployment starting H2 2026 progressing well.

MI355 sharp ramp proceeded well in Q3First gigawatt MI450 deployment starts H2 2026MI355 expected to continue ramping in H1 2026MI450 series ramps sharply in H2 2026

Timothy Arcuri · UBS

Asked about market impact from OpenAI announcement one month after deal, whether it influenced engagement with other customers, and risk of single-customer concentration given potential 50%+ of data center GPU revenue from OpenAI in 2027-2028.

Lisa noted OpenAI deal was in works for considerable time; has driven increased customer interest alongside Helios rack showcase at OCP; observed acceleration in customer engagement at higher scale; managing concentration risk by dimensioning supply chain for multiple customers at similar scale in 2027-2028 timeframe; commitment to broad customer base as foundation of business.

OpenAI deal was multi-quarter effort in developmentObserved acceleration of customer interest and broader engagement post-announcementPlanning supply chain to support multiple customers at similar scale as OpenAI2027-2028 goal is multiple large-scale customers not concentrated on single account

Thomas O'Malley · Barclays

Asked about discrete vs. system sales crossover timing for Helios, customer response at OCP, and whether component availability or data center infrastructure/power would be the binding constraint for 2026 deployments.

Lisa reported phenomenal OCP reception with customers bringing engineering teams; early MI450 customers expected to focus on rack-scale solutions with other form factors available; great excitement expanding with OpenAI and OCI announcements; on constraints, noted entire ecosystem must plan together on power, silicon, memory, packaging, and supply chain over two-year horizon; supply chain well-positioned to deliver significant growth; tightness expected but ecosystem adapts as power and supply open up.

Early MI450 customers will focus on rack-scale Helios solutionsOther form factors for MI450 also availableOCP reception described as phenomenal with strong customer engineering engagementSupply chain prepared for significant growth rates

Joshua Buchwalter · Cowen

Asked about sustainability of CPU demand trends driven by AI workloads, supply chain constraints, whether server CPU business is becoming aseasonal, expectations for H1 2026 seasonality, and competitive status of Rockham 7 software stack vs. competitors.

Lisa stated broadening CPU demand observed across 2025 with large hyperscalers forecasting significant CPU builds into 2026; described demand as durable multi-quarter phenomenon tied to AI workload growth requiring general-purpose compute; sees positive demand environment into 2026; has supply to support growth; Rockham 7 represents significant performance step forward with day-zero support for new models; most customers starting with AMD experiencing smooth onboarding; continued investment in libraries and frameworks, particularly for emerging workloads combining training, inference, and reinforcement learning.

CPU demand broadening across multiple quartersLarge hyperscalers forecasting significant CPU builds into 2026Demand described as durable multi-quarter phenomenonRockham 7 delivers significant performance step forward

Aaron Rakers · Wells Fargo

Asked about unit growth vs. ASP expansion dynamics in server CPU cycle, and view on updated AI silicon TAM opportunity beyond the previously cited $500 billion.

Lisa explained TURN has more content driving ASP growth as it ramps, but Genoa demand remains strong as hyperscalers unable to migrate everything to latest generation immediately; broad-based CPU demand across workloads driven largely by AI-spawned traditional compute needs; noted strong pull for Venice in early engagements indicating importance of general-purpose compute; on TAM, stated $500 billion view will be updated at Analyst Day but AI compute TAM 'just going up' and larger opportunity exists over next few years.

TURN ramping very quickly with higher ASP/contentGenoa demand continues strong across hyperscalersCPU demand driven by AI workloads spawning traditional computeStrong pull and early engagement on Venice

Answers to last quarter's watch list

MI-350 quarterly revenue disclosure / Data Center YoY ex-China — Data Center grew 22% YoY in Q3 with zero MI-308 China contribution, clearing the "mid-20s%" bar I set, though not by much. AMD still refused to break out MI-specific revenue, but Su confirmed MI355 ramped well in Q3 and the return to $1.07B in segment operating income (after the Q2 MI308 charge distortion) validates the underlying ramp. Status: Resolved positively
Data Center segment operating income recovery — Q3 Data Center operating income was $1,074M vs. -$155M in Q2 (which included an $800M MI308 write-down) and $1,041M in the year-ago quarter. Underlying segment profitability is restored, though YoY margin compressed 460bps as R&D investment scaled. Status: Resolved positively
MI-308 China license decision — No revenue from MI-308 in Q3 and no commentary on license status. The China line remains structurally absent and Q4 guidance does not include it. Status: Resolved negatively (no path back disclosed)
Q4 Client growth trajectory — Client grew 46% YoY in Q3. Q4 segment-level commentary from Jean Hu indicates Client revenue increasing sequentially with Gaming down strong double digits post-holiday semi-custom build. Status: Continue monitoring
MI-400 customer commitments — Su confirmed first-gigawatt MI450 deployment with OpenAI starts H2 2026 and that supply is being dimensioned for "multiple customers at similar scale" by 2027-2028. OpenAI is now the anchor tenant. Additional named hyperscalers were not disclosed, though Oracle was confirmed as a lead MI450 launch partner with tens of thousands of GPUs starting in 2026. Status: Resolved positively

What to watch into next quarter

Whether Q4 prints another large beat vs. the ~$9.6B guide — a third consecutive >$200M beat reframes the Q4 guide as conservatism and supports a higher 2026 growth setup.

Q4 Data Center segment growth rate — Q3 printed +22% YoY; Jean Hu guided to double-digit sequential growth in Data Center in Q4 driven by server plus MI350 ramp. Watch whether Data Center re-accelerates YoY as MI355 scales.

Data Center operating margin trajectory — 24.7% in Q3 is 460bps below year-ago; watch whether MI355 mix and scale drive margin recovery or whether ongoing R&D intensity keeps margins compressed through the MI450 ramp.

Named MI450 anchor tenants beyond OpenAI and Oracle — Su flagged 2027-2028 supply planning for "multiple customers at similar scale." Watch for specific hyperscaler MI450 commitments on the Q4 call; absence of additional names by then makes the OpenAI concentration risk acute.

Analyst Day (Nov 11) TAM and long-range financial targets — AMD's updated AI silicon TAM, MI450 revenue trajectory framing, and any 2027 revenue anchor will reset the structural valuation case.

Embedded segment inflection — five consecutive quarters of YoY decline now (-8% in Q3), though +4% sequentially and guided to double-digit sequential growth in Q4. Watch whether Q4 confirms stabilization.

Sources

  1. AMD Q3 2025 press release (SEC filing): https://www.sec.gov/Archives/edgar/data/2488/000000248825000163/q32025991.htm
  2. AMD Q3 2025 earnings call Q&A
  3. Tapebrief AMD Q2 2025 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.