AMD Data Center Sales Nearly Double as MI450 GPUs Reach Big AI Buyers

AMD's data center business nearly doubled in the third quarter as its new Instinct MI450 GPUs began shipping into rack-scale systems used by some of the largest AI buyers. The chipmaker reported on Tuesday, November 3, that record revenue beat estimates and guided to a sharp jump in the fourth quarter. For enterprise leaders planning AI capacity, the results point to a second large supplier of AI accelerators at a time when compute access, cost and vendor concentration are board-level concerns.

What did AMD report for the third quarter?

Revenue rose 45% from a year earlier to a record $13.4 billion, AMD said in its earnings release. Data center revenue reached a record $8.5 billion, up 97%, driven by demand for EPYC server processors and the launch of the Instinct MI450 Series. GAAP gross margin was 54%, and non-GAAP earnings were $2.09 per share, up 74%.

The rest of the business was mixed. Client and gaming revenue fell 8% to $3.7 billion, as a 7% rise in client processors was offset by a 40% drop in gaming. Embedded revenue rose 29% to $1.1 billion.

Revenue topped the consensus estimate of $13.09 billion, and adjusted earnings beat the $1.95 per share analysts expected, Beinsure reported, citing LSEG data. Data center sales now make up 64% of the total. The shares rose about 8% in extended trading after three straight days of declines.

How strong is AMD's fourth-quarter outlook?

AMD guided to fourth-quarter revenue of about $17 billion, plus or minus 3%, which the company said implies roughly 66% growth from a year earlier and 27% from the third quarter. That compares with an LSEG average estimate of $16.2 billion, according to Beinsure. Guidance for a non-GAAP gross margin of about 55% came in slightly below the StreetAccount consensus of 55.3%.

"Demand for our Instinct and EPYC products continued to strengthen," chief executive Lisa Su said in the release. Patrick Moorhead of Moor Insights & Strategy told Yahoo Finance, as cited by Beinsure, that AMD is now mainly a data center company with a PC business attached, and that investors had at first applied Intel's weaker PC outlook to AMD despite its much larger data center exposure.

Who is running AI workloads on AMD's Helios systems?

AMD said Anthropic, Meta, Microsoft, OpenAI and Oracle Cloud Infrastructure are running models and workloads on the first Helios rack-scale systems, which combine MI450 GPUs with sixth-generation EPYC CPUs. The release did not give deal terms for any of those customers.

The company also listed several channels that matter for enterprises that do not build their own clusters. HPE said cloud provider Vultr will deploy the first Helios systems built by HPE. Oracle's OCI Compute E7 instances will use sixth-generation EPYC processors. In Europe, the EuroHPC Joint Undertaking picked MI430X GPUs and EPYC processors for the LUMI-AI supercomputer.

What does a stronger AMD mean for enterprise AI infrastructure?

For most enterprises, AMD's quarter matters less as a stock story than as a supply story. When frontier labs and hyperscalers commit to a second GPU platform, the software stack, cloud instances and integrator support around it tend to mature, which widens the options for everyone else.

AI Connect speakers have long urged teams to keep those options open. In a AI Connect webinar on AI infrastructure at scale, AtScale co-founder and CTO Dave Mariani argued that infrastructure choices should stay open because "we don't know where this stuff is going." Earlier panels on self-hosted models and vendor lock-in made the same case for testing several vendors before committing.

Practical steps for infrastructure and finance leaders:

• Ask cloud providers about AMD-based instances. As OCI and Vultr add AMD capacity, price and availability comparisons across accelerators become possible for inference workloads.
• Test portability early. Models and pipelines built tightly around one vendor's tools are harder to move. Proofs of concept on a second platform reveal the real switching cost.
• Measure utilization, not just price. More supply does not help if GPUs sit idle. At The AI Enterprise Conference 2026, Everpure's Kaycee Lai warned that when storage cannot keep GPUs busy, "that's not a good thing to tell your CFO." Tracking cost per outcome keeps the focus on value delivered.

The next markers are how fast Helios moves from the largest AI labs to enterprise-facing clouds, and whether AMD's fourth-quarter guidance holds as investors reassess AI infrastructure spending.

Frequently asked questions

How much did AMD's data center revenue grow in Q3 2026?

Data center revenue rose 97% from a year earlier to a record $8.5 billion, driven by EPYC CPUs and the launch of Instinct MI450 GPUs.

What is AMD Helios?

Helios is AMD's rack-scale AI system combining Instinct MI450 Series GPUs with sixth-generation EPYC CPUs. AMD said Anthropic, Meta, Microsoft, OpenAI and Oracle are running workloads on the first units.

What does AMD's growth mean for enterprise AI buyers?

A second large accelerator supplier can widen cloud instance choices and pricing options, but enterprises should test software portability and GPU utilization before shifting workloads.

Filed under: Data & AI Infrastructure, AI Strategy & ROI