AMD's Helios chips land Microsoft, and Anthropic may be next

Nvidia’s iron grip on the AI chip market is starting to look a little less absolute. Microsoft is expanding its use of AMD hardware in Azure, betting on a new platform called Helios that’s slated to ship in the second half of 2026. The pitch is straightforward: give cloud customers a real alternative to Nvidia’s GPU systems for running large AI models. AMD CEO Lisa Su called the deal a major milestone, and Microsoft’s Satya Nadella framed it as a matter of customer choice. Fair enough — nobody wants a single supplier calling all the shots.

But the more interesting thread might be what’s happening over at Anthropic. A public GitHub profile belonging to an AMD director suggests the Claude-maker is actively testing AMD hardware. According to SemiAnalysis, AMD has given Anthropic its highest priority rating, the same tier reserved for giants like Meta. That’s not a casual tire-kicking exercise. It signals that a company known for pushing AI boundaries is seriously evaluating whether it can build its future on something other than Nvidia silicon.

The catch, as always with AMD, is software. Anthropic is still in the testing phase, and sources indicate AMD engineers are working to resolve concerns about software quality before anything goes public. If they pull it off, a formal announcement could come at AMD’s “Advancing AI” conference. That would be a coup — not just another cloud partnership, but a top-tier AI lab putting its weight behind the ecosystem.

Nvidia isn’t sweating yet. The company still commands enormous pricing power, and its CUDA software moat remains deep. But the ground is shifting. Between Microsoft’s Helios commitment, OpenAI’s existing AMD partnership, and now Anthropic circling, the alternatives are getting more credible. The question isn’t whether Nvidia faces competition anymore. It’s how fast that competition gets good enough to matter.