Akamai disclosed a broad infrastructure agreement with Anthropic on September 24, under which the AI company committed approximately $11.6 billion for dedicated cloud computing capacity and managed support services over seven years.
The agreement was actually signed on September 18 but disclosed publicly only six days later in Akamai's filing with the US Securities and Exchange Commission. The agreement can expand by up to another $9 billion, bringing its potential value to approximately $20 billion.
What Anthropic is buying from Akamai
According to Akamai's 8-K filing, the company will provide dedicated cloud capacity and managed support services to Anthropic.
The agreement focuses on Anthropic's CPU workloads.
The fact that this concerns CPU, not just GPU, is interesting because large AI companies need broad infrastructure beyond training and inference on accelerators. Storage, networking, preprocessing, orchestration and supporting services can also create significant demand for general computing resources.
The public filing does not specify the exact service commencement dates for every part of the agreement.
$11.6 billion, with room to expand
Anthropic committed approximately $11.6 billion through two project plans with initial terms of seven years.
In addition, the structure permits further purchases of up to $9 billion.
This means the agreement could reach approximately $20 billion, but distinguish the existing commitment from the option to expand it. The additional $9 billion is not a purchase already made.
Actual payments also depend on delivery conditions, availability and termination clauses included in the project plans.
Akamai also gives Anthropic a stock warrant
As part of the deal, Akamai issued Anthropic a warrant potentially covering approximately 7.7 million common shares on an as-converted basis.
According to the filing, this represents approximately 5% of Akamai's outstanding shares, at an exercise price of $111.33 per share.
Approximately 2% is linked to the initial commitment, and the additional portion, approximately 3%, can vest as Anthropic purchases up to another $9 billion in services.
This makes the agreement more than an ordinary customer-supplier deal. Akamai is also giving Anthropic some economic exposure to the infrastructure provider's success.
Akamai expects to invest billions in infrastructure
Akamai estimates it will need approximately $5.5 billion in capital expenditure related to the agreement to support the project.
The company also noted an additional approximately $1.7 billion in 2026 for components, including memory.
The capex scale underscores the size of the commitment. This is not a standard cloud purchase that can be supplied solely from existing capacity. Akamai needs to expand infrastructure to meet the requirements.
Why Anthropic is seeking additional infrastructure
Companies operating frontier models need computing capacity across several layers.
GPU attracts most attention, but a service at scale also requires CPU, networking, storage and distributed capacity.
The Akamai deal indicates that Anthropic is expanding its supplier base and seeking infrastructure that can grow alongside its workloads.
For Akamai, this is an opportunity to expand its position in the AI infrastructure market beyond the CDN and security businesses for which it is primarily known.
What remains unknown
The public filing does not provide a full schedule for bringing all the capacity online.
It is unknown whether Anthropic will use the entire $9 billion expansion option.
Actual payment amounts may also change depending on delivery and service availability.
The $20 billion figure should therefore remain defined as a potential ceiling, not an existing commitment.
Why the deal matters
An $11.6 billion agreement over seven years is a significant indication of the pace at which AI companies are building infrastructure.
This case also combines a cloud commitment, large supplier capex and a warrant that could give Anthropic a significant economic stake in Akamai in an unusual way.
The important story is not just the agreement's value. It shows how competition over AI infrastructure is expanding beyond GPU and traditional hyperscale providers, bringing players such as Akamai into long-term deals with leading model companies.