Anthropic Commits $11.6 Billion to Akamai for CPU Cloud Capacity

Akamai announced on September 24, 2026 that Anthropic has committed to spend $11.6 billion on its cloud infrastructure over seven years, the largest contract in the company's history and a substantial bet that the next phase of AI demand will run on ordinary processors as much as on the specialized chips that have dominated the boom so far.

Related reading: Ukraine Hits Two Yandex Data Centres, Reaching Russia's AI Compute · Oracle Sends Force Majeure Notice on $18 Billion New Mexico Stargate Campus.

The agreement is more than six times the size of a $1.8 billion deal between the two companies that Bloomberg reported in May, and Akamai said it adds to more than $2.8 billion in multi-year cloud infrastructure commitments announced across its customer base this year. The commitment is not ironclad. According to Akamai's securities filing, it depends on the company meeting certain delivery and service-availability requirements, and either party can terminate under specified conditions.

Akamai said the money supports Anthropic's "accelerating CPU workload demands" by using Akamai Cloud's distributed infrastructure and software. The company did not say what Anthropic will run on the capacity, but the emphasis on CPUs is the part of the announcement that signals a shift.

The CPU undercurrent in AI spending

Most of the conversation about AI infrastructure has centered on GPUs and the accelerators that train and serve large models. But AI agents that execute long chains of small tasks, running code, querying services and browsing the web, generate a different kind of load that general-purpose CPUs handle. As companies deploy agents rather than single chat responses, that demand compounds.

Akamai is positioning itself in that less crowded corner. Its network has long been a content-delivery business, and it has spent the past year building out cloud infrastructure aimed at workloads that do not need high-end accelerators. A seven-year commitment from one of the largest AI labs validates the approach, at least at the contractual level.

The financial profile is back-loaded. Akamai told investors it expects no revenue from the deal this year, $150 million to $300 million in 2027 beginning in the second half, and an annual run rate of about $1.7 billion by the end of 2028. To serve it, the company expects to spend roughly $5.5 billion building capacity, and it added about $1.7 billion to its 2026 capital spending to buy components such as memory in advance.

That timing means the market is being asked to trust a multi-year plan before the revenue arrives. Akamai's shares rose as much as 17 percent in after-hours trading on Thursday, according to The Wall Street Journal, a reaction that reflects both the size of the contract and the credibility a customer like Anthropic confers.

A warrant that flips the usual AI deal

The most unusual term is the warrant. Akamai issued Anthropic the right to buy nonvoting preferred stock convertible into 7.7 million common shares, equal to up to about 5 percent of the company, at $111.33 a share. Roughly 2 percent is expected to vest once Anthropic makes its first payment under the deal, with the remainder tied to further spending. Each additional $3 billion Anthropic commits unlocks about another 1 percent.

That structure inverts the pattern of circular AI deals, in which chipmakers and cloud providers invest in the labs that buy their products. Here, the supplier is giving its customer a potential equity stake that grows as the customer spends more. AMD used a comparable approach with OpenAI, tying warrants to chip-purchase milestones.

It is the first time Akamai has attached a warrant to a cloud deal. The mechanism aligns the two companies' incentives over a long horizon: Anthropic gets a path to a stake if it deepens the relationship, and Akamai gets a stronger reason to believe the spending will materialize. It also exposes Akamai to dilution if Anthropic earns the full 5 percent, and it could turn a customer into a significant shareholder.

Why Akamai is an unexpected winner

Akamai built its business on content delivery, moving images, video and web traffic through a network of servers close to users. Cloud infrastructure is a newer and more competitive line, and it puts the company up against the largest hyperscalers, which have far more capacity and far deeper relationships with AI labs. Winning a seven-year commitment from Anthropic is the clearest sign yet that Akamai's distributed model has found a niche.

That niche is workloads that do not need a data center full of GPUs. Akamai's network already reaches most of the world's internet users, and CPU capacity placed at the edge can serve the code execution, data handling and request routing that AI agents generate. It is a different kind of AI infrastructure from the accelerator arms race, and it is one where Akamai's existing footprint is an asset rather than a disadvantage.

The deal also gives Akamai a reference customer of the highest profile. If Anthropic's workloads run well on the platform, other AI companies weighing CPU capacity have a reason to look at the same option. The risk is that a single large contract now defines a business line that is still proving itself, and Akamai has committed capital to it before the revenue arrives.

The memory-price backdrop

One detail in Akamai's disclosures is easy to miss. The company said it added about $1.7 billion to this year's capital spending to buy components such as memory in advance. Memory prices have been volatile as AI demand competes with consumer hardware for supply, and buying ahead is a hedge against both scarcity and price swings.

That hedge carries its own risk. If prices fall, Akamai has paid more than it needed to. If they rise, the early purchase looks wise. Either way, the decision shows how much of an AI infrastructure deal is now a bet on the component supply chain rather than only on software and real estate.

Anthropic's widening web of compute partners

Anthropic is no stranger to these arrangements. Amazon, Google, Microsoft and AMD have all invested in the company, or agreed to, while selling it chips or cloud capacity. Chief executive Dario Amodei told The New York Times in December that Anthropic does not participate in such deals at the "same scale as some other players," a characterization the Akamai contract will test.

The deal also reflects the scale of the compute the lab now requires. Anthropic has expanded aggressively across infrastructure providers as it competes with OpenAI and others for capacity, and its roster of suppliers now spans the major clouds plus specialized providers. Diversifying away from any single vendor reduces dependency, though it also multiplies the number of long-term financial commitments the company carries.

For Akamai, the risk is concentration. A single customer anchoring a large share of a new business line can be a strength while the relationship grows and a vulnerability if it contracts. The delivery and service-availability conditions in the filing give Anthropic bargaining power, and the warrant gives it a claim on the upside.

The competitive picture

Akamai is not the only company chasing CPU-heavy AI workloads, but it is unusual in combining a global edge network with a cloud business. Hyperscalers sell general-purpose compute alongside their accelerators, and many AI labs already run code and data processing on the same platforms that host their models. Akamai's argument is that there is value in separating the two, especially when agents need to run close to users and when accelerator capacity is scarce or expensive.

If that argument holds, more suppliers will try to copy it, and the warrant structure may become a standard negotiating tool. If it does not, Akamai will be left with a large build-out aimed at a demand curve that did not arrive. The seven-year term is long enough to cover both outcomes, which is precisely why the market reacted the way it did.

What to watch next

The near-term questions are operational. Akamai has to build capacity fast enough to meet the schedule it has described, and it is buying memory and other components ahead of demand at a time when prices have been volatile. If it slips, the contract's conditions could give Anthropic room to renegotiate or exit.

The longer-term question is whether CPU-heavy AI workloads grow as Akamai and Anthropic are betting. If agentic systems keep expanding the volume of small, persistent computing tasks, the CPU market becomes a larger slice of AI infrastructure spending, and Akamai's early positioning looks prescient. If the demand stays concentrated in accelerators, the deal will look like expensive capacity bought on optimistic assumptions.

Either way, the contract marks a change in how AI infrastructure deals are structured. A seven-year, multi-billion-dollar commitment, sweetened with equity that vests on spending, is a template more suppliers may copy as they compete for a small number of very large AI customers.

Cover image: the Akamai Technologies headquarters in Cambridge, Massachusetts. Photo by Coolcaesar via Wikimedia Commons (CC BY-SA 3.0).