Anthropic Spreads AI Compute Across 3 Vendors as CFO Says 30%-40% of Time Goes to Capacity
Updated
Updated · 24/7 Wall St. · May 15
Anthropic Spreads AI Compute Across 3 Vendors as CFO Says 30%-40% of Time Goes to Capacity
3 articles · Updated · 24/7 Wall St. · May 15
Krishna Rao said Anthropic’s compute planning is existential: buy too little and it falls behind frontier rivals; buy too much and it risks going out of business.
30%-40% of the CFO’s time goes to capacity decisions because small weekly or monthly demand shifts can compound into radically different outcomes over one to two years.
Anthropic uses Nvidia GPUs, Amazon Trainium and Google TPUs to keep workloads fungible and avoid being trapped by any one supplier’s roadmap risk.
Nvidia said Anthropic will initially adopt 1 gigawatt of Grace Blackwell and Vera Rubin capacity, while Amazon is training Claude on more than 500,000 Trainium2 chips and Google remains both a TPU supplier and investor.
The strategy underscores why AI infrastructure spending is surging across Anthropic’s backers and suppliers, even as the same compute buildout creates a financial overcapacity risk for frontier labs.
With Anthropic's billion-dollar compute bets, could overcapacity trigger an AI infrastructure bubble before 2028?
How will Anthropic sustain Claude's performance and reliability as it juggles rapid scaling and multi-vendor compute strategies?
Could Anthropic's push for proprietary chips and hardware co-design upend NVIDIA's dominance in the AI hardware market?
Anthropic’s $80 Billion Compute Surge: Diversifying AI Infrastructure from Cloud to Orbit
Overview
In May 2026, Anthropic made a bold move to strengthen its position in the competitive AI industry by dramatically increasing its investment in computational resources. The company announced a multi-vendor strategy, committing $30 billion to Azure capacity through partnerships with Microsoft and Nvidia, and an additional $50 billion for US AI infrastructure. This approach ensures Anthropic can rapidly scale its operations and maintain robust, diverse compute infrastructure for its advanced AI models like Claude. By adopting this strategy, Anthropic aims to secure the resources needed to lead in AI development and deployment.