AMD invests $5 billion in Anthropic to boost AI infrastructure

What happened?

AMD has announced an investment of up to $5 billion in Anthropic, the artificial intelligence startup behind Claude, to expand its computing infrastructure. As part of the deal, Anthropic will deploy up to 2 gigawatts (GW) of AMD Instinct MI450 GPUs using AMD's new Helios rack-scale system. Deployment will begin with the first GW in the first half of 2027, according to AMD's official statement and reports from The Wall Street Journal and The Verge. This investment adds to Anthropic's previous agreements with Google, Broadcom, Amazon, SpaceX, and TeraWulf, reflecting a strategy of supplier diversification to secure computing capacity in a market where demand exceeds supply.

Historical context

AMD has been trying for years to gain market share from NVIDIA in the data center GPU segment, dominated by NVIDIA with over 80% of the market. In 2023, AMD launched the Instinct MI300 series, which achieved some success with clients like Microsoft and Oracle, but mass adoption remains elusive. The deal with Anthropic represents AMD's largest individual investment in AI infrastructure, surpassing previous agreements such as the $1 billion deal with Microsoft in 2023. For Anthropic, this move is part of a broader trend: since 2023, the startup has signed multi-billion dollar agreements with Google ($2 billion), Amazon ($4 billion), and SpaceX (cloud computing capacity), in addition to a 20-year lease with TeraWulf for a data center in Kentucky. The need to secure long-term computing capacity has become critical for AI companies, especially after OpenAI and Microsoft jointly invested over $10 billion in infrastructure. The deployment of 2 GW is roughly equivalent to the capacity of two large nuclear power plants, posing significant energy and environmental challenges. According to the IEA's 2024 report, data centers already consume 1-2% of global electricity, and this figure is expected to double by 2030.

Why is this important?

This agreement is strategic for both players. For AMD, it represents its largest investment in the AI market, aiming to erode NVIDIA's dominance, which currently controls over 80% of the data center GPU market. For Anthropic, it secures large-scale computing capacity to train and run its language models, in a context where infrastructure demand exceeds supply. Additionally, it diversifies its partnerships, adding to previous agreements with Google, Broadcom, Amazon, SpaceX, and TeraWulf. Dependence on a single supplier (NVIDIA) has been a risk for many AI startups; Anthropic seeks to mitigate it with multiple partners. AMD's Helios system, which promises more efficient GPU integration, directly competes with NVIDIA's solutions like DGX. According to Bernstein analysts, AMD's share of AI GPUs could rise from the current 10% to 15-20% by 2028 if it successfully executes deals like this one.

Market consequences

AMD's move could accelerate competition in the AI hardware market, pressuring NVIDIA to innovate and adjust prices. For Anthropic, this investment reinforces its ability to scale, though it also raises questions about dependence on multiple technology partners. Geopolitically, the deal underscores the importance of AI infrastructure as a strategic asset, with implications for technological sovereignty and energy security. Moreover, the race to secure computing capacity is leading to increasingly large agreements: in 2024, Microsoft and OpenAI announced a $100 billion plan for a data center, while Google invested $2 billion in Anthropic. This deal could also influence regulation: the European Union and the United States are considering laws to ensure equitable access to AI infrastructure. For investors, the news reinforces the thesis that AI hardware will remain a high-growth market, with a Compound Annual Growth Rate (CAGR) estimated at 30% through 2030 according to Gartner.

What readers should know

• The $5 billion investment is for infrastructure, not equity; AMD is not acquiring a stake in Anthropic.
• The deployment of 2 GW is roughly equivalent to the capacity of two large nuclear power plants, posing significant energy and environmental challenges.
• Anthropic has signed similar agreements with Google, Amazon, and SpaceX, indicating a multi-supplier strategy to secure computing capacity.
• AMD's Helios system promises more efficient GPU integration, directly competing with NVIDIA's solutions like DGX.
• According to The Verge, rumors suggest Meta is also considering leasing computing capacity from Anthropic, underscoring the high demand for infrastructure.

"This deal is a game-changer in AI infrastructure. AMD shows it is willing to compete aggressively, and Anthropic secures the power needed for its ambitions." — Analyst at TheVortiq.