While AI giants scramble for Nvidia GPUs, a cloud company born as a CDN is quietly making big money in the AI compute race. On September 24, Akamai announced a seven-year, $11.6 billion cloud agreement with Anthropic — the largest contract in Akamai's history — sending its shares up more than 20% in after-hours trading.
The deal centers on CPU workloads. Unlike the GPU clusters used to train large models, Anthropic has committed to buying $11.6 billion of cloud services over seven years, using Akamai Cloud's distributed infrastructure spanning thousands of points of presence to handle growing CPU compute demand — agent inference, content delivery and other non-training workloads. The contract can expand by a further $9 billion, bringing the potential total to about $20 billion. Notably, this is more than six times the $1.8 billion partnership disclosed in May; Anthropic's compute appetite is clear.
As part of the strategic alignment, Akamai issued Anthropic a warrant to purchase non-voting convertible preferred stock representing 7.7 million shares of common stock on an as-converted basis — up to about 5% of Akamai's outstanding common stock, at an exercise price of $111.33 per share. About 2% vests with the current $11.6 billion commitment, with the remaining 3% tied to further expansion. Akamai expects roughly $5.5 billion in capital expenditures tied to the contract, plus about $1.7 billion in additional 2026 spending to secure supply.
The deal highlights a fact often overlooked: AI's compute demand is far from GPU-only. Anthropic has already committed $4.5 billion to lease Nscale's West Virginia data center campus, and has locked in infrastructure with Nvidia, Amazon and now Akamai — model training is just the beginning; agents' massive inference and invocation demands are creating another enormous compute track. For a "non-hyperscaler" player like Akamai, landing the largest AI customer's long-term contract means formally entering the inner circle of the AI infrastructure race. As the second battlefield beyond GPUs — distributed CPU compute — gets carved up by hundred-billion-dollar commitments, AI's compute map is expanding faster than most imagine.