AMD Data Center Revenue Doubles as Helios AI Racks Begin to Ramp

AMD's data center revenue more than doubled in the second quarter, rising 107% from a year earlier to $6.7 billion, as demand for its EPYC server processors and Instinct AI accelerators surged, the company reported on August 4. Data center now makes up 58% of AMD's revenue. For enterprise buyers, the results show that a second large supplier of AI hardware is scaling, which matters for pricing, availability and how much leverage customers have when they plan AI infrastructure.

What did AMD report?

Total revenue was $11.5 billion, up 50% year over year, according to AMD's earnings release filed with the Securities and Exchange Commission. Gross margin was 54% on a GAAP basis and 56% on a non-GAAP basis.

The data center segment carried the quarter. Its growth rate accelerated sharply from the 57% pace of the prior quarter, which Saxo had flagged in an earnings preview as the number to beat. Client processor revenue rose 23% to $3.1 billion, embedded revenue rose 19% to $977 million, and gaming fell 31% to $779 million.

For the third quarter, AMD guided to revenue of about $13 billion, plus or minus $300 million. At the midpoint that is roughly 41% growth from a year earlier and 13% from the second quarter, with non-GAAP gross margin expected around 56%.

Who is buying AMD's AI systems?

The release reads like a list of the largest AI buyers. AMD cited a strategic partnership with Anthropic to deploy up to 2 gigawatts of MI450 Series GPUs in AMD's Helios rack-scale systems, along with a multiyear collaboration to optimize Instinct GPUs and AMD's ROCm software using Claude. Microsoft expanded plans to deploy Helios racks at scale on Azure and added new virtual machines built on EPYC.

AMD also listed Helios deployments with Cirrascale, HUMAIN, Meta, OpenAI, Oracle, Tensorwave, Vultr and others. It named a collaboration with Cerebras that pairs Helios with Cerebras' wafer-scale engine for low-latency inference, and one with Cisco that combines AMD's Ryzen AI Halo systems with Cisco networking and security.

CEO Lisa Su said in the release that the second half should bring three things at once: EPYC demand accelerating, Instinct deployments scaling and the start of the Helios ramp.

Why does AMD's growth matter to enterprise AI buyers?

Most enterprises will never buy a Helios rack directly. They will rent AI capacity from cloud providers and model vendors, many of which now appear on AMD's customer list. A credible second source of accelerators gives those providers more room to add capacity and negotiate on price, and some of that should eventually show up in what enterprises pay for inference.

It also changes the software conversation. AMD's ROCm stack has long been seen as the weak point against Nvidia's CUDA. Anthropic's agreement to help optimize ROCm with Claude is a bet that AI-assisted engineering can narrow that gap faster. Enterprises running their own models should test whether their frameworks and inference servers run well on both vendors' hardware before committing.

AI Connect speakers have urged that kind of flexibility for some time. Dave Mariani, Co-founder and CTO of AtScale, argued during a AI Connect webinar on AI infrastructure that scales that infrastructure should stay open because "we don't know where this stuff is going." Panelists in an earlier session on future-proof AI infrastructure warned about vendor lock-in for the same reason.

What should enterprise leaders watch next?

The first test is execution. Helios is only beginning to ramp, and AMD's third-quarter guidance assumes that rack-scale systems ship on schedule. Delays would ripple through cloud capacity plans.

The second is cost. More supply does not automatically mean cheaper AI for end users, especially when the largest buyers are booking capacity years ahead. "Compute costs money," Robert Blanchard, Principal Data Scientist at SAS, said at AI Connect's AI Demo Day this week, explaining why his team replaced some LLM steps with cheaper deterministic models. That discipline, matching the workload to the right infrastructure, will matter more than which chip sits in the rack.

The third is the competitive response. With data center revenue now the majority of AMD's business, watch for sharper pricing and bundling from both AMD and Nvidia as they compete for the same hyperscaler and model-lab orders in the second half.

Frequently asked questions

How much did AMD's data center revenue grow in Q2 2026?

Data center revenue rose 107% year over year to $6.7 billion, making up 58% of AMD's $11.5 billion total revenue.

What is AMD Helios?

Helios is AMD's rack-scale AI system built around Instinct GPUs. AMD said Helios is beginning to ramp, with deployments at Meta, OpenAI, Oracle, Microsoft Azure and others.

What is AMD's Q3 2026 guidance?

AMD expects third-quarter revenue of about $13 billion, plus or minus $300 million, with non-GAAP gross margin around 56%.

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