DataAMD -9.70%NOV 05, 2026

AMD data-center revenue rises 97%; Q4 sales forecast reaches $17 billion

CPU and GPU demand drove the quarter, giving small technology providers a reason to revisit customer workloads—not promise cheaper or available capacity.

BY LINH VINH · DATA INFRASTRUCTURE

AMD reported November 3 that Q3 2026 data-center revenue rose 97% year over year to $8.5 billion, driven by EPYC processors and the launch of Instinct MI450 Series GPUs. It isn’t a standalone measure of AI accelerator sales: server CPUs contributed, too.

For the quarter ended September 26, total revenue reached $13.4 billion, up 45% annually and 16% sequentially. GAAP net income was $2.4 billion, or $1.46 per diluted share. Non-GAAP diluted earnings rose 74% to $2.09 per share. GAAP operating income increased 123% to $2.8 billion; operating margin reached 21%, versus 14% a year earlier.

AMD’s SEC-filed earnings presentation shows data-center revenue also grew 27% sequentially. Segment operating income increased 168% annually to $2.9 billion, with margin rising to 34% from 25%. Client and Gaming revenue fell 8% to $3.7 billion: client revenue rose 7% to $2.9 billion, while gaming fell 40% to $773 million on lower semi-custom and graphics revenue. Embedded revenue increased 29% to $1.1 billion.

Fourth-quarter revenue guidance is approximately $17 billion, plus or minus 3%, implying approximately 66% annual and 27% sequential growth. That’s guidance, not booked sales. Forecast non-GAAP gross margin is approximately 55%, against 56% in the third quarter.

“We expect growth to accelerate further in the fourth quarter as we scale Helios and ‘Venice’ production,” said Lisa Su, chair and CEO of AMD.

AMD says Anthropic, Meta, Microsoft, OpenAI and Oracle Cloud Infrastructure are running models and workloads on its first Helios rackscale systems, powered by Instinct MI450 Series GPUs and sixth-generation EPYC CPUs. Oracle Cloud Infrastructure announced its Compute E7 instances will use sixth-generation EPYC CPUs for enterprise applications and CPU services surrounding models. Google Cloud and Microsoft announced general availability of storage-optimized instances using fifth-generation EPYC CPUs. AMD also reported releasing ROCm 10, with updated libraries, compilers, frameworks, tools and model support, alongside developer experiences through ROCm.AI.

The practical inference for owners of small B2B cloud, hosting and AI-service firms: use the results to reopen customer discovery, not promise cheaper compute. Ask which workloads need capacity soon, then verify availability, regional access, compatibility and pricing with the relevant provider before proposing services.

The structural signal is broader infrastructure demand across CPUs and GPUs, with supplier diversification worth discussing. Supplier revenue doesn’t establish what an individual customer can buy, where they can run it or what they’ll pay.

Sources

About the author

Linh Vinh

DATA INFRASTRUCTURE

Linh Vinh covers vector databases, retrieval, feature stores, and the plumbing layer of enterprise AI. She has been writing about data infrastructure for several years and files on infra vendor consolidation, pricing pressure, and technical architecture shifts.