AMD and Anthropic announced a strategic partnership on July 22, 2026 under which Anthropic will deploy up to 2 gigawatts of AMD Instinct MI450 Series GPUs. According to AMD’s press release, deployment of the first gigawatt begins in the first half of 2027, and AMD has committed to a strategic equity investment of up to $5 billion in Anthropic.
On this page
The hardware
Anthropic will deploy AMD Helios rack-scale systems. AMD lists the components as follows:
| Component | As stated by AMD |
|---|---|
| GPUs | AMD Instinct MI455X, part of the Instinct MI450 Series |
| CPUs | AMD EPYC “Venice” |
| Networking | AMD Pensando |
| Software | ROCm |
| Rack platform | AMD Helios rack-scale solutions |
| Total scale | Up to 2 GW |
| First gigawatt | Deployment begins in the first half of 2027 |
AMD says the deployment “builds on Anthropic’s existing use of AMD Instinct MI355X GPUs,” so the agreement extends an existing supplier relationship to a newer GPU generation and a full-rack system rather than starting one. The Instinct line-up is covered in The Model Press AI chip catalog.
Equity and engineering
AMD “has committed to make a strategic equity investment of up to $5 billion in Anthropic in the future,” according to the release. The two companies are also starting a multi-year engineering collaboration in which teams “will use Claude to optimize workloads for AMD Instinct GPUs and accelerate ROCm software development.” AMD says it will also broadly adopt Claude across its engineering and product development teams.
The arrangement runs in both directions: Anthropic commits to deploy AMD hardware at gigawatt scale, AMD commits capital to Anthropic, and Anthropic’s model is used inside AMD to work on the software stack that the deployment depends on.
What changes
A gigawatt is a measure of power, so the 2 GW figure describes the electrical scale of the planned deployment rather than a GPU count. The first of the two gigawatts starts deploying in the first half of 2027.
For AMD, the agreement places Helios and the MI450 Series in a frontier lab’s deployment plan at gigawatt scale. For Anthropic, it adds a GPU supplier commitment sized in gigawatts, on hardware whose software layer, ROCm, the two companies say they will work on together.
The deployment is tracked in The Model Press AI data-center project tracker.
