Akamai Anthropic Deal: What the $11.6B CPU Pact Means

Published September 25, 2026. Akamai announced the agreement on September 24, 2026, at 4:01 p.m. U.S. Eastern Time, after the U.S. stock market closed (early on September 25 in Pakistan and much of Asia).

The Akamai Anthropic deal announced this week is one of the clearest signs yet that the AI compute race is no longer only about graphics processors. Akamai Technologies, a company best known for delivering web content and blocking cyberattacks, says Anthropic has committed $11.6 billion over seven years to its cloud platform, specifically to handle the Claude maker’s growing demand for CPU capacity.

The agreement also comes with an unusual twist: Akamai has handed Anthropic a warrant that could turn the AI lab into a shareholder owning up to about 5% of the company. Below, we break down what was actually agreed, what the numbers imply, and which questions neither company has answered yet.

Akamai Technologies headquarters building at 145 Broadway in Cambridge, Massachusetts — Akamai’s headquarters at 145 Broadway in Cambridge, Massachusetts. Photo: Daderot via Wikimedia Commons, released into the public domain under CC0 1.0.

What Happened?

In an official press release, also filed with the U.S. Securities and Exchange Commission as an exhibit to a Form 8-K, Akamai described a “significantly expanded relationship” with Anthropic. The core commitment is $11.6 billion of contracted spending across seven years on Akamai Cloud’s distributed infrastructure and software.

The word “expanded” matters. The two companies were already working together. In May, Akamai disclosed a $1.8 billion, seven-year commitment from an unnamed frontier AI company, and Bloomberg later reported that the customer was Anthropic. This week’s contract is more than six times larger.

Investors reacted quickly. Akamai’s shares rose sharply in after-hours trading on September 24, with Bloomberg’s report noting a 17% gain to $129.60 shortly after the announcement and other outlets reporting moves of about 20% later in the session.

The Key Details

The following terms come directly from Akamai’s filing unless otherwise noted.

Term	What was disclosed
Initial commitment	$11.6 billion over seven years
Workload type	CPU workloads on Akamai Cloud
Optional expansion	Up to $9 billion more, for roughly $20 billion in total
Warrant	Non-voting convertible Series B Preferred Stock, equal to 7.7 million common shares (about 5% of Akamai’s stock outstanding)
Exercise price	$111.33 per common share
Vesting now	About 2%, tied to the $11.6 billion commitment
Vesting later	About 1% for each additional $3 billion of services purchased
Akamai’s capital spending for the deal	About $5.5 billion in total
Extra 2026 capital spending	About $1.7 billion to secure components, including memory
2026 revenue guidance	No change expected

In an interview reported by Bloomberg, Akamai CEO Tom Leighton added several details that were not in the press release. He said the service is expected to start in the second half of 2027, that revenue from the Anthropic business should be roughly $150 million to $300 million next year, and that it should reach an annual run rate of about $1.7 billion by 2028.

What Is New Here?

Three things make this agreement stand out from the steady stream of AI infrastructure announcements.

1. It is explicitly about CPUs

Most headline-grabbing AI compute deals involve GPUs or other accelerators used to train and run large models. This contract is described as supporting Anthropic’s “accelerating CPU workload demands.” That is a notable choice of words from a company whose customer builds some of the world’s most advanced models.

2. The customer gets equity

Leighton told Bloomberg this is the first time Akamai has granted a warrant to a customer as part of a cloud deal. “It’s a serious step, but I think in this case it made sense to do,” he said. The structure rewards Anthropic with more of the warrant only if it keeps buying more services.

3. The scale is transformative for Akamai

Akamai earlier reported more than $2.8 billion in multi-year cloud infrastructure commitments across all customers this year. A single $11.6 billion contract dwarfs that. According to Bloomberg, the related capital spending alone is more than six times what Akamai spent in all of 2025.

Why the Akamai Anthropic Deal Matters for AI

For years, the public conversation about AI hardware has focused almost entirely on accelerators. This deal is a reminder that serving AI to millions of people depends on a much wider stack: general-purpose processors, memory, storage, networking and security.

Our analysis: neither company has said exactly which Anthropic workloads will run on these CPUs. However, modern AI products increasingly do more than generate text. AI agents run code, call external tools, search files, and manage long multi-step tasks. Much of that surrounding work is ordinary computing that runs on CPUs rather than GPUs. If agent-style products keep growing, demand for this type of capacity is likely to rise alongside, not instead of, demand for accelerators. This is an informed interpretation, not a company statement.

If you are new to the topic, our explainer on how AI agents are changing business covers why agents place different demands on infrastructure than a simple chatbot does.

How the Technology Fits Together

Akamai started as a content delivery network, a system of servers spread around the world that stores copies of websites close to users so pages load faster. Over time it added security services and, more recently, a cloud computing platform.

Akamai says its cloud spans thousands of points of presence and uses “diversified hardware.” In plain terms, that means computing capacity located in many places rather than in a few giant data centers. For AI services, that geographic spread can help with speed and resilience, although the filing does not say how much of Anthropic’s capacity will sit in large core sites versus smaller edge locations.

The warrant is worth understanding too. A warrant is a right to buy shares later at a set price. Here, the price is $111.33 per share. At that price, exercising the full 7.7 million-share warrant would cost roughly $857 million. Anthropic benefits only if Akamai’s stock rises above the exercise price, which ties part of its reward to Akamai’s success.

What It Means for AI Users

For people who use Claude, nothing changes immediately. The service is not expected to start until the second half of 2027.

Over the longer term, more capacity generally means fewer rate limits, faster responses and room for heavier features such as longer-running agents. That is a reasonable expectation, not a promise: Anthropic has not published any product plans tied to this contract.

What It Means for Businesses

Cloud buyers: Hyperscalers such as Amazon, Google and Microsoft are no longer the only realistic homes for large AI workloads. Leighton said Akamai is discussing more business with “all the major players,” including hyperscalers and large enterprises.
Akamai investors: The company is taking on a large, front-loaded spending commitment. Leighton said most of the capital spending would happen next year and would go toward servers, chips and networking gear.
Hardware suppliers: Akamai’s plan to pre-purchase memory adds to already heavy demand for memory chips from AI buildouts.
AI startups: Large labs are locking up capacity years in advance, which can make it harder and more expensive for smaller players to secure the same resources.

Industry Context

This contract is one piece of a much larger capacity push by Anthropic. According to SiliconANGLE, Anthropic signed a 20-year lease in July for about 401 megawatts at a TeraWulf data center campus in Kentucky, and a memory and storage supply agreement with Micron in June. Bloomberg reports that Anthropic has also sourced chips from companies including Google and SpaceX.

The timing is also notable. Anthropic is widely reported to be preparing for a public listing, which we covered in our report on Anthropic’s profitable quarters and IPO plans. Large, long-term infrastructure commitments will be closely examined by future investors.

Finally, the deal lands in the middle of a broader surge in AI infrastructure spending, from chips to data centers. Our recent piece on Meta’s Muse agent and the chip stock rally shows how quickly markets are pricing in agent-driven demand.

Timeline

May 2026: Akamai discloses a $1.8 billion, seven-year commitment from an unnamed frontier AI firm, later reported to be Anthropic.
September 24, 2026: Akamai announces the $11.6 billion expanded agreement and the warrant.
Second half of 2027: Service expected to begin, according to Leighton.
2027: Akamai expects $150 million to $300 million in revenue from the relationship.
By 2028: Leighton expects an annual run rate of about $1.7 billion.

What Remains Unclear

Anthropic’s view: The announcement came from Akamai. We did not find a separate public statement from Anthropic explaining what the capacity will be used for.
Workload specifics: “CPU workloads” is broad. It could cover agent tooling, data processing, inference support or internal systems.
Circular-deal concerns: Some investors worry about arrangements where suppliers give equity to customers, because it can blur how much real demand exists. Bloomberg noted this debate. Here, the warrant only fully vests if Anthropic keeps spending, but the concern is still part of the picture.
Execution risk: Akamai must buy and deploy billions of dollars of hardware while memory and other components are in high demand. Its own filing lists supply chain disruption among the risks.

What Happens Next?

The key milestones are straightforward to track. Watch Akamai’s next quarterly results for updated capital spending and cloud revenue figures. Look for any statement from Anthropic, especially in its IPO paperwork, describing how it plans to use this capacity. And watch whether Anthropic exercises its expansion option, since each additional $3 billion of purchases vests another slice of the warrant and would be a public signal of growing demand.

Conclusion

The Akamai Anthropic deal is significant less for its headline figure than for what it reveals about the shape of AI demand. A leading model developer is committing billions to general-purpose processors spread across a distributed network, and paying partly in a closer financial relationship with its supplier. For Akamai, it is a company-defining bet that its path beyond content delivery runs through AI. For the wider industry, it is evidence that the next phase of AI infrastructure will be broader than GPUs alone. Whether that bet pays off will become clearer once the service goes live in 2027.