AMD Q3 revenue rises 36% to $9.2b

AMD reported record revenue of US$9.2 billion for Q3 2025, up 36% year-on-year, with net income of US$1.2 billion and diluted earnings per share of US$0.75.

Gross margin was 52%, while operating income reached US$1.3 billion.

On a non-GAAP basis, net income came to US$2 billion and earnings per share were US$1.20.

Data center segment revenue rose 22% year-on-year to US$4.3 billion, supported by demand for EPYC processors and Instinct GPUs.

Client and gaming segment revenue reached US$4 billion, up 73% year-on-year, driven by Ryzen processor sales and increased gaming GPU demand.

Embedded segment revenue dropped 8% year-on-year to US$857 million.

The company said Q3 results did not include revenue from shipments of Instinct MI308 GPUs to China.

AMD expects Q4 revenue of about US$9.6 billion, plus or minus US$300 million, with a non-GAAP gross margin of around 54.5%.

🔗 Source: AMD

🧠 Food for thought

Implications, context, and why it matters.

AMD data center growth relies on GPUs as mix stays vague

AMD posted $4.3 billion in data center revenue that bundles EPYC server CPUs (the company’s server processors) with Instinct AI GPUs (its data center accelerators), while it does not disclose each slice 1.
Fifth-gen EPYC Turin made up nearly half of EPYC sales, so CPUs still power a large share of data center results 1.
Oracle offers MI355X instances as AMD signs multi-year deals with OpenAI for 6 gigawatts of Instinct GPUs 1. The first gigawatt of MI450 accelerators should launch in the second half of 2026 1.
AMD said supply looks solid, while MI308 shipments to China are excluded from Q3 results plus Q4 guidance due to U.S. export controls 2 1.

Clouds add MI300X capacity and openings for service vendors

Oracle launched MI355X instances for real-time AI, and AMD’s Developer Cloud (a hosted environment for testing plus running workloads on AMD GPUs) offers MI300X VMs with 192 GB plus 8x with 1536 GB 1 3.
Integration vendors can help enterprises moving off Nvidia migrate to Radeon Open Compute (ROCm), AMD’s open software stack for GPU computing 1. ROCm7 lifted inference up to 4.6x and training 3x 1.
Managed service providers (MSPs) and AI infrastructure consultants can offer services around AMD GPUs as MI300X pricing eases with MI350X plus MI355X rollouts 4. Time-bound GPU allocations plus multi cloud control planes create demand for orchestration and placement optimization that help teams choose where to run workloads at lower cost 4.