AMD invests $5bn in Anthropic to close Nvidia software gap
AMD is investing up to $5 billion in Anthropic and supplying 2 gigawatts of compute power, a move that could finally break Nvidia’s stranglehold on the AI hardware market by solving AMD’s long-standing software problems.
AMD will invest up to $5 billion in Anthropic and deploy 2 gigawatts of its Instinct MI450 Series GPUs to run the Claude AI model. The first gigawatt of these Helios rackscale solutions is scheduled to ship in the first half of 2027. Anthropic is already operating AMD’s MI355X chips and will transition to the newer MI455X accelerators, paired with EPYC "Venice" CPUs and Pensando networking.
The hardware deployment is substantial, but the engineering collaboration may matter more to the broader market. Anthropic will use Claude to optimise workloads specifically for AMD's chips and accelerate the development of ROCm, AMD's software stack. Because ROCm's shortcomings are the primary reason developers default to Nvidia hardware even when AMD offers competitive specs, using Anthropic's AI to close this software gap could resolve the bottleneck that has restricted AMD's market share for years.
For Anthropic, the deal secures a genuinely multi-vendor compute strategy at a time when computing capacity remains the primary constraint on frontier AI development. The AI company already relies on Google and Broadcom, Amazon (for 5 gigawatts of Trainium chips), CoreWeave (for Nvidia GPUs), and SpaceX. Anthropic chief compute officer Tom Brown said the AMD addition gives the company the ability to "map the right workloads to the right hardware," thereby reducing the risk of over-reliance on any single chip vendor.
AMD’s $5 billion equity investment mirrors a strategy Nvidia pioneered by taking stakes in AI infrastructure firms like Nebius to lock in hardware demand. By funding a major customer, AMD is ensuring its chips run a meaningful share of the world's most advanced AI models rather than being sidelined. Lisa Su called the Anthropic deal "a major platform for the next generation of AI infrastructure," though the true test will be whether these Helios systems can compete with Nvidia’s NVL72 racks at production scale.
For businesses and investors tracking the AI supply chain, this partnership signals a potential shift in market dynamics. If AMD successfully uses Claude to bridge the software gap, enterprise buyers could eventually face a genuine alternative to Nvidia. That outcome would reduce pricing power in the GPU market and alter the economics of AI deployment across European and global industries.
