Akamai Signs $11.6B Cloud Deal With Anthropic — AI Compute Race Expands Beyond GPUs to 'CPU Infrastructure'

Known for its content delivery network (CDN) and cybersecurity business, Akamai has stepped onto the stage as a compute supplier for frontier AI. On September 24, the company said it had struck a multi-year cloud computing deal with Anthropic worth at least $11.6 billion over seven years. The agreement can expand by up to $9 billion more, bringing the potential total to roughly $20 billion. The striking detail: the target is not Nvidia GPU training clusters, but Anthropic's "growing CPU workloads."

The Deal — $11.6B Over Seven Years, Up to $20B

The terms are straightforward. Anthropic will use Akamai Cloud's distributed infrastructure and software to handle its rising CPU workloads, and Akamai will scale up to meet that demand. The company said it plans roughly $5.5 billion in capital expenditures over the buildout tied to this commitment, including an additional $1.7 billion in 2026 to secure supply-chain components such as memory.

The two also tied their interests together through equity. Akamai issued Anthropic a warrant for about 5% of its common stock — the right to buy roughly 7.7 million shares at $111.33 each. About 2% of that vests alongside the current $11.6 billion commitment, while the remaining ~3% vests in stages as the deal expands in $3 billion increments. In other words, the bigger the contract grows, the more of Akamai that Anthropic can own.

Why CPUs, Not GPUs?

Say "AI infrastructure" and most people picture Nvidia GPUs. But this deal is aimed at CPU workloads, not GPU training. Training frontier models and running massive matrix math takes GPUs and TPUs — yet operating those models as a live service demands enormous amounts of general-purpose CPU compute. Orchestrating and routing API requests, data pre- and post-processing, retrieval and augmentation, agent tool execution, safety filtering, logging and monitoring all run on CPUs.

As Claude usage has exploded, this "surrounding" compute has itself become vast, and Anthropic needed a partner that could handle it close to users across a geographically distributed cloud. That is exactly Akamai's strength: infrastructure spread across thousands of points of presence worldwide is well suited to low-latency serving and elastic scaling. The deal, in short, is about a different layer of the AI stack than GPUs — the serving and operations infrastructure that pushes models out into the world.

Akamai's Reinvention — From CDN and Security to 'AI Cloud'

Akamai was co-founded in 1998 by Tom Leighton, an MIT mathematician, and long built its business on CDN and web security. After acquiring cloud-hosting provider Linode in 2022, it pushed into distributed cloud under the "Akamai Connected Cloud" banner. The Anthropic deal is the culmination of that pivot — Akamai's bid to establish itself as a compute supplier for the AI era.

"Anthropic is advancing the AI revolution and we are thrilled they chose Akamai's capabilities for building and operating AI infrastructure at scale," said Dr. Tom Leighton, Akamai's co-founder and CEO. In an AI-cloud market dominated by the hyperscalers (AWS, Google, Microsoft), Akamai is betting that its distributed footprint gives it a differentiated way to carve out a niche.

What It Means

The deal signals two things. First, AI compute demand is spilling well beyond GPU training into the serving and operations layer — the fact that one frontier lab's "surrounding" compute alone can anchor a deal in the tens of billions is proof of that. Second, AI is becoming a new growth axis for incumbent infrastructure players like Akamai, once a CDN and security company.

Execution is the question. Akamai has to absorb roughly $5.5 billion in capex while protecting margins, and Anthropic has to fill that infrastructure with real, growing usage. Because the equity warrant vests in stages, whether the contract expands from $11.6 billion toward $20 billion will be the first real test of how much confidence sits behind this partnership.