Anthropic has committed $11.6 billion to Akamai over seven years for cloud infrastructure, marking one of the largest infrastructure deals in AI history and signaling a strategic pivot toward CPU-based computing over GPUs. The agreement includes an unusual equity component where Akamai will grant Anthropic a stake of up to 5% of its stock, with the percentage increasing proportionally as Anthropic's spending grows. The deal could balloon to roughly $20 billion if expansion clauses trigger.
This arrangement breaks from typical cloud vendor relationships. Rather than paying purely for capacity, Anthropic gains both computing resources and financial upside through Akamai shareholding. The structure aligns incentives: as Anthropic scales, it owns more of Akamai, creating mutual investment in each other's success. For Akamai, the deal locks in a massive revenue stream from a high-growth customer while retaining control. For Anthropic, it secures guaranteed capacity at predictable costs while building a financial stake in infrastructure.
The CPU focus carries strategic weight. While GPU clusters dominate large language model training, CPUs excel at inference, the phase where models generate responses to user queries. As Claude becomes more widely deployed across products and services, inference compute often exceeds training compute in volume and cost. Akamai's global edge network and CPU infrastructure position it well to handle distributed inference workloads at scale.
The timing reflects Anthropic's growth trajectory. The startup has raised over $7 billion in funding and operates one of the largest AI models by parameter count. Growing inference demands from enterprise deployments, API access, and potential product integrations strain existing infrastructure arrangements. Securing multi-year capacity through Akamai reduces uncertainty and ensures Anthropic controls its own computing destiny rather than relying on shared cloud provider resources where capacity constraints could limit growth.
Akamai's involvement also signals changes in the AI infrastructure market. Hyperscalers like AWS, Google Cloud, and Azure traditionally dominated AI workload hosting. This deal suggests specialized infrastructure providers can capture meaningful AI spending by offering differentiated compute solutions. Akamai's strengths in edge computing, content delivery, and global distribution address needs that generic cloud providers may not prioritize. The equity component demonstrates how competitive the infrastructure bidding has become.
The potential $20 billion ceiling reveals Anthropic's internal projections for seven-year spending, implying expected growth in deployment scale and user volume. For context, Nvidia's entire annual data center revenue reached roughly $47 billion in 2024, so Anthropic's commitment to one infrastructure vendor represents substantial scale within a narrower segment.
Questions remain about execution. Akamai must rapidly scale CPU capacity to meet Anthropic's needs without service degradation. Anthropic must integrate Akamai infrastructure into its systems while maintaining flexibility to add or modify vendors if requirements shift. The equity stake means long-term financial alignment but also creates complexity if either party's strategic direction changes.
This deal foreshadows broader infrastructure competition as AI workloads mature. Winners will offer not just raw compute but global distribution, cost efficiency, and reliability at scales that match the largest AI companies' ambitions.