
AMD has landed the kind of AI infrastructure deal it needs if it wants investors to believe the Nvidia era will not be a one-company monopoly. The company will invest up to $5 billion in Anthropic, while Anthropic plans to buy up to 2 gigawatts of AMD’s latest AI chips beginning in 2027.
Reuters reported that the deal centres on AMD’s upcoming Instinct MI450 chips and Helios rack-scale systems. The first 1GW deployment is expected in the first half of 2027, with the broader commitment potentially reaching 2GW. The arrangement also includes engineering collaboration, with AMD using Anthropic’s Claude models in its software and product workflows.
This is not just a chip sale. It is part hardware order, part financing relationship and part strategic alignment. MarketWatch says the overall partnership could be worth tens of billions of dollars, while AMD’s investment will be tied to deployment milestones.
For Anthropic, the logic is obvious. Claude needs more compute. Every major AI lab is now fighting for chips, power, data-centre space and cloud capacity. Depending only on Nvidia gives buyers less leverage, so companies are looking for credible alternatives before the bottleneck gets worse.
For AMD, the deal is even more important. Nvidia still owns the AI accelerator market in both mindshare and software maturity. AMD’s challenge has been proving that its chips can win real production deployments, not just benchmark attention. Anthropic gives it a marquee customer with huge demand and strong technical credibility.
This follows AMD’s broader Helios push. TechBooky recently covered how Microsoft became a Helios AI rack customer, giving AMD another cloud-scale validation point. Add Anthropic to Microsoft, OpenAI, Meta and Oracle, and AMD can now argue that its AI story is moving from promise into deployment.
There is also a wider industry pattern here. AI startups increasingly need partners with balance sheets, not only vendors with chips. If AMD can help Anthropic secure hardware and possibly support future data-centre leases, it becomes more than a supplier. It becomes part of the financial scaffolding behind the AI buildout.
That comes with risk. These circular arrangements, where chip companies invest in AI companies that then buy their hardware, will attract scrutiny. Investors will ask whether demand is organic, subsidised or somewhere in between. But in a market where compute access determines model progress, companies are willing to get creative.
The simple takeaway is that Nvidia is still ahead, but AMD is no longer knocking politely. With Anthropic and Microsoft tied to Helios, AMD now has a more credible route into the highest end of AI infrastructure.