MANY MINDEDAMD and Anthropic strike a 2-gigawatt AI compute deal
AMD will invest up to $5 billion in Anthropic under a deal that has Anthropic deploying up to 2 gigawatts of AMD Instinct MI450 GPUs in Helios server systems. The chips will train and run Anthropic's Claude models. The first gigawatt phase is scheduled to begin in the first half of 2027.
The systems pair MI455X GPUs with AMD EPYC Venice processors and AMD networking. Anthropic already uses AMD's MI355X GPUs, and the two companies plan a multi-year effort to use Claude to improve AMD's ROCm software, while AMD adopts Claude internally across its development teams. AMD has signed similar arrangements with Meta and OpenAI.
The significance for knowledge abundance is supply diversification. Frontier AI has depended heavily on a single chip vendor, and a second large-scale supplier can ease bottlenecks and, over time, pressure prices for the compute that underpins AI tools. The two-gigawatt scale also ties this closely to energy demand, since that capacity requires substantial power.
The caveats are real. Both the $5 billion investment and the 2-gigawatt figure are stated as "up to" ceilings, not commitments. Critics note the circular structure of these deals, where a chipmaker invests in a customer that then buys its chips. And whether AI labs can independently cover bills of this magnitude remains unresolved. The 2027 start date leaves ample room for the terms to shift.
Source: The Decoder