Anthropic has committed $11.6 billion over seven years to Akamai Technologies, the largest contract in the content-delivery company's history and one of the most unusual AI infrastructure deals of the year. Unlike the GPU-heavy agreements that dominate headlines, this one centres on CPUs, the general-purpose chips that increasingly carry the load of AI agents, tool calls and orchestration.
The Deal Terms
Akamai announced the agreement on September 24. Under it, Anthropic will use Akamai Cloud's distributed infrastructure and software to handle its growing CPU workloads. Key terms disclosed in Akamai's announcement and SEC filing include:
- $11.6 billion over seven years, with an option to expand by a further $9 billion, taking the potential total to roughly $20 billion.
- A master services agreement dating to May 5, 2026, with two project plans forming the $11.6 billion commitment signed on September 18.
- A deal more than six times larger than the roughly $1.8 billion arrangement between the companies that Bloomberg reported in May.
The commitment is not unconditional. Akamai's filing says it depends on meeting delivery and service-availability requirements, and either party can terminate under certain conditions.
An Equity Kicker for Anthropic
Akamai also issued Anthropic a warrant to buy non-voting convertible preferred stock equal to up to 7.7 million common shares, about 5% of shares outstanding, at an exercise price of $111.33 per share.
Roughly 2% is expected to vest under the initial commitment. The remaining 3% vests in 1% steps each time Anthropic commits another $3 billion in cloud spending. The structure echoes vendor-financing arrangements seen elsewhere in the AI build-out, where suppliers hand customers equity upside in exchange for long-term volume.
Akamai Spends to Deliver
Delivering on the contract requires heavy investment. Akamai expects about $5.5 billion in capital expenditure tied to the commitment, including an extra $1.7 billion in 2026 to secure and pre-purchase critical components such as memory. It signed a hardware supply agreement with Lenovo on September 23 and authorised contract manufacturer Jabil to buy about $1.7 billion of memory the following day.
The revenue arrives later. Akamai said the deal does not change its 2026 guidance and expects $150 million to $300 million in revenue from it in the second half of 2027 as services ramp. The contract adds to more than $2.8 billion in multi-year cloud commitments Akamai had already announced this year.
Investors cheered. Akamai shares jumped more than 20% after hours on Thursday and were still trading higher on Friday, with 24/7 Wall St. reporting that CoreWeave and Cloudflare also ticked up in sympathy. Akamai CEO Tom Leighton said the company was thrilled Anthropic had chosen its platform to build and operate AI infrastructure.
Why CPUs Matter Now
The choice of CPUs is the most telling part of the deal. Training and serving large models still runs on GPUs and custom accelerators, but agentic AI creates a large and fast-growing layer of conventional compute around the model:
- Sandboxes where agents execute code and browse the web.
- Tool calls to databases, APIs and enterprise systems.
- Orchestration that schedules, monitors and retries multi-step tasks.
- Pre- and post-processing of data flowing in and out of models.
As products like Claude Code and enterprise agents scale, each model call can spawn many CPU-bound operations. Akamai's globally distributed footprint, originally built to serve web content close to users, is well suited to running those workloads at the edge.
The timing also stands out. The underlying agreement was signed in May, when the relationship was reported at a fraction of today's size. Its expansion to $11.6 billion within four months shows how quickly Anthropic's demand for supporting compute has grown as its agent products have scaled, and how much suppliers are willing to invest up front to capture that demand.
Why It Matters
The deal shows that the AI infrastructure race is broadening beyond Nvidia GPUs and hyperscaler data centers. For Anthropic, it diversifies supply at a time when compute is scarce, and adds a potential equity stake in a key supplier. For Akamai, it is a transformational contract that repositions a mature CDN business as an AI cloud player.
For the wider market, it is another sign that agent workloads are creating demand for conventional compute. Cloud providers without top-tier GPU fleets may still find a lucrative role in the AI economy, provided they can build fast enough and lock in memory and hardware supply ahead of the competition.
