Anthropic is committing $11.6 billion over seven years to Akamai’s cloud infrastructure, a deal that highlights the extraordinary scale of computing resources now being secured by leading AI laboratories and points to a less visible constraint in the industry’s expansion: demand for conventional CPUs.

The agreement, announced by Akamai on Thursday, is more than six times the size of the $1.8 billion arrangement between the companies reported in May. It is also the largest contract in Akamai’s history.

The scale of the commitment illustrates how aggressively Anthropic is building computing capacity as it develops and deploys sophisticated AI systems. While much of the industry’s attention has focused on the enormous demand for Nvidia GPUs and other specialized AI accelerators, Anthropic’s agreement with Akamai highlights the growing importance of general-purpose processors as AI agents perform more tasks beyond simply generating text or images.

The deal is not unconditional. According to an Akamai securities filing, the agreement depends on the company satisfying specified delivery and service-availability requirements, and either party can terminate the arrangement under certain circumstances.

The headline $11.6 billion figure represents a long-term commitment rather than revenue that Akamai will immediately recognize. Akamai expects to generate between $150 million and $300 million from the agreement in 2027, beginning in the second half of the year. Revenue is expected to reach an annualized pace of roughly $1.7 billion by the end of 2028.

For Akamai, the contract represents a major expansion opportunity but also requires a substantial upfront investment.

The company expects to spend about $5.5 billion to build the capacity required for Anthropic. It is also adding about $1.7 billion to its 2026 capital expenditure to secure components, including memory, ahead of demand. That creates a striking financial equation: Akamai is effectively committing billions of dollars of its own capital to prepare infrastructure for a customer whose payments will arrive over several years.

The Akamai agreement is the latest example of Anthropic securing computing capacity from multiple infrastructure providers as it scales its AI operations.

Anthropic has previously entered major arrangements involving Amazon, Google, Microsoft and AMD. Those relationships combine access to chips or cloud infrastructure with investments in Anthropic, creating complex financial ties between AI developers and the companies supplying the computing power required to train and operate their models.

The Akamai transaction introduces another variation.

Instead of Akamai taking an equity position in Anthropic, Anthropic receives a warrant that could eventually give it a stake in Akamai.

Under the agreement, Akamai issued Anthropic a warrant for nonvoting preferred stock convertible into 7.7 million common shares, equivalent to as much as about 5% of Akamai’s outstanding shares at a price of $111.33 per share. About 2% is expected to vest when Anthropic makes its first payment under the agreement. Additional portions are linked to Anthropic’s future spending.

For every additional $3 billion Anthropic commits to Akamai’s cloud services, roughly another 1% of Akamai becomes available to Anthropic. If all the additional spending milestones are reached, the overall arrangement could expand by as much as $9 billion, taking the potential value of the cloud relationship to about $20 billion.

The structure effectively links Anthropic’s growing computing requirements with an increasing potential ownership interest in its infrastructure supplier.

It is also the first time Akamai has attached a warrant to a cloud agreement.

A Different Kind of AI Infrastructure Deal

The arrangement reverses a structure that has become increasingly common across the AI industry. Typically, infrastructure companies invest in the AI laboratories that become their customers. Chipmakers and cloud providers have provided capital to AI developers while simultaneously securing demand for their hardware and computing services.

The Akamai agreement works in the opposite direction. Anthropic, the customer, receives the potential equity upside in the infrastructure provider, resulting in an unusual alignment of interests.

Anthropic has an incentive to increase its use of Akamai’s infrastructure because greater spending can unlock additional equity. Akamai, meanwhile, receives a large multiyear commitment that can help justify the capital expenditure required to build the necessary capacity.

AMD used a related structure with OpenAI last year, linking warrants to milestones for chip purchases.

The growing prevalence of these arrangements illustrates how difficult it has become to separate the financing of AI companies from the economics of the infrastructure supporting them.

The AI laboratory needs enormous amounts of computing capacity. Infrastructure providers need sufficiently large and predictable customers to justify building that capacity. Equity-linked contracts can tie the two sides together more closely than conventional supplier agreements.

CPUs Emerge As An Overlooked Bottleneck

The Akamai deal is notable for another reason. The agreement is centered on cloud infrastructure and highlights demand for CPUs, rather than focusing exclusively on the GPUs that have dominated the AI infrastructure narrative.

CPUs are general-purpose processors used for a wide range of computing tasks, including running code and managing web activity. As AI agents become capable of carrying out longer and more complicated sequences of tasks, demand for conventional computing resources can increase alongside demand for specialized AI accelerators.

Akamai did not disclose precisely how Anthropic plans to use the CPUs covered by the agreement. Still, the transaction points to an important feature of the AI buildout. Training and running advanced models requires much more than accelerators.

Data must be processed and moved. Applications need to execute code. Agents need to interact with websites and software. Systems need storage, networking, and conventional compute resources to coordinate the enormous volumes of work generated by AI applications.

As agentic AI becomes more widely deployed, those supporting workloads could become a substantial infrastructure market of their own. That gives Akamai an opportunity to participate in AI infrastructure without competing directly with the companies supplying the most prominent training accelerators.

The Economics Are Becoming Harder to Ignore

The agreement also illustrates the enormous capital requirements created by the AI boom. Akamai expects to spend roughly $5.5 billion building capacity for a contract that will generate an estimated $150 million to $300 million of revenue in 2027 and an annualized $1.7 billion by the end of 2028.

The gap between upfront investment and eventual revenue demonstrates why AI infrastructure providers increasingly need long-term commitments from large customers before they can justify massive capacity expansions.

It also raises questions about utilization.

Analysts have noted that if demand from AI laboratories continues expanding at the pace assumed in these contracts, the infrastructure investments could generate substantial recurring revenue for providers such as Akamai. If AI demand grows more slowly, however, infrastructure companies could find themselves carrying large capital commitments for capacity that is not fully utilized. That issue is becoming more significant as multiple cloud providers, chipmakers and data-center operators simultaneously expand capacity in anticipation of AI demand.

Anthropic, for its part, is effectively locking in access to infrastructure years ahead of time. That can provide greater certainty over capacity as competition for computing resources intensifies, but it also creates long-term financial obligations.

The arrangement therefore says as much about the economics of the AI infrastructure race as it does about Anthropic’s own growth.

Akamai’s stock rose as much as 17% in after-hours trading on Thursday, according to The Wall Street Journal, indicating that investors viewed the agreement as a significant commercial opportunity.

The market response is understandable given the size of the contract relative to Akamai’s existing business. The company now has a large customer commitment that can support the expansion of its cloud infrastructure while potentially increasing the scale of its AI-related operations.

But the deal also places greater importance on execution. Akamai must spend billions of dollars to deliver the capacity, meet service requirements, and ultimately convert Anthropic’s commitments into recurring revenue.

The warrant adds another layer to the arrangement. If Anthropic’s spending increases and additional portions vest, the company could become a meaningful shareholder in Akamai, aligning its financial interests with the success of the infrastructure provider. That structure is unusual enough to signal how quickly traditional relationships between technology suppliers and their customers are changing.