The race to build infrastructure for artificial intelligence has largely been told as a story about graphics processors. A new agreement between Akamai Technologies and Anthropic points to a less-discussed part of the market that is growing fast: the conventional central processing units, or CPUs, that AI agents rely on to get work done.
Akamai announced after the US market close on Thursday, September 24, that it had signed an $11.6 billion, multi-year agreement with Anthropic, the developer of the Claude family of AI models. Under the seven-year deal, Akamai will provide cloud infrastructure and services to support Anthropic's CPU workloads on its distributed infrastructure.
The agreement could expand further. Anthropic can increase its spending by up to an additional $9 billion, which would take the total value to roughly $20 billion.
Investors reacted strongly. Akamai shares rose as much as 17 per cent in after-hours trading on Thursday and surged about 15 per cent in early trading on Friday, September 25, before settling to close up about 8.8 per cent, according to TheStreet. Heading into the session, the company had a market capitalisation of about $15.9 billion, making the contract large relative to its size.
"Anthropic is advancing the AI revolution and we are thrilled they chose Akamai's capabilities for building and operating AI infrastructure at scale," said Tom Leighton, Akamai's co-founder and chief executive.
Why CPUs, and why now
Training and running large AI models depends heavily on specialised accelerators such as graphics processing units (GPUs). But as AI systems evolve into agents that carry out multi-step tasks, a growing share of the work happens outside the model itself.
An AI agent that writes and tests code, browses the web, calls software tools or processes files needs general-purpose computing to execute those actions. Those tasks, from running code in isolated environments to handling data and orchestrating workflows, are typically performed on CPUs. As agent usage grows, so does demand for large pools of CPU capacity distributed close to users.
TechCrunch described the agreement as a bet on "a less-hyped corner of AI infrastructure". Neither company disclosed exactly which workloads the capacity will support.
The terms in detail
The deal includes several unusual features. According to Akamai's announcement and reporting by TechCrunch, the commitment is contingent on Akamai meeting delivery and service-availability requirements, and either party can end the agreement under certain conditions.
Akamai also issued Anthropic a warrant for non-voting convertible Series B preferred stock. The warrant could give Anthropic exposure to up to about 7.7 million shares, or approximately 5 per cent of Akamai's common stock, at an exercise price of $111.33 per share. Roughly 2 per cent vests with the initial $11.6 billion commitment, with about 1 per cent more vesting for each additional $3 billion of cloud services Anthropic purchases, up to a further 3 per cent.
Warrants of this kind have become more common in large infrastructure deals, aligning the interests of a supplier and a major customer. For Anthropic, it provides upside if Akamai's value rises, partly because of the business Anthropic brings. For Akamai, it helps secure a long-term, high-value customer.
The capital commitment
Serving the contract will require significant investment. Akamai expects capital expenditure of roughly $5.5 billion to $5.6 billion related to the commitment. It anticipates an increase of about $1.7 billion in 2026 capital spending, partly to secure supply chain components, including memory, in advance.
