AMD's Data Center Revenue Doubles to $6.7B as MI400 and Anthropic's 2GW Deal Land

AMD reported second-quarter 2026 revenue of $11.5 billion, up 50% year over year and 13% sequentially, with data center revenue more than doubling to $6.7 billion — now roughly 58% of total company revenue. CEO Lisa Su credited the jump to accelerating demand across “Epyc demand accelerates, Instinct deployments scale and Helios begins to ramp.”

The headline product news: AMD’s Instinct MI400 series is now shipping, led by the MI455X for large-scale AI training and inference and the MI430X aimed at HPC and sovereign AI workloads. Alongside the chips, AMD launched Helios, its rack-scale AI server system, which it’s calling the most powerful AI server rack on the market. Helios is already deployed or being deployed by Anthropic, Cirrascale, HUMAIN, Meta, Microsoft, OpenAI, Oracle, Tensorwave, and Vultr.

The Anthropic relationship is the most concrete signal of AMD’s competitive position against Nvidia. AMD confirmed a strategic partnership under which Anthropic will deploy up to 2 gigawatts of MI450-series GPUs in Helios racks, with the first gigawatt beginning deployment in the first half of 2027. The two companies are also collaborating directly on GPU and ROCm software development — a meaningful vote of confidence in AMD’s software stack from a frontier lab that could have gone all-in on Nvidia’s CUDA ecosystem instead.

Guidance for Q3 came in at roughly $13 billion, plus or minus $300 million. AMD said it expects data center segment revenue to more than double again in 2027, and Su reportedly locked in yearly AI rack launches through 2030, with the MI500 series confirmed for 2027 — a public roadmap commitment that puts AMD on a comparable annual cadence to Nvidia’s.

Despite beating estimates on both revenue and guidance, AMD shares fell more than 8% on the report, a reaction that reads as investors pricing in near-term margin questions around the capital intensity of gigawatt-scale rack deployments rather than doubting the demand itself. For infrastructure teams evaluating AI hardware, the practical takeaway is that ROCm is now backed by a frontier lab at meaningful scale — the software gap that’s historically kept AMD out of serious consideration next to Nvidia is closing faster than the stock reaction suggests.

Sources

Related reading