Anthropic locks in seven years of cloud with Akamai for up to $20B, and is paid in part with 5% of its shares

By The Boston Globe · September 27, 2026.

Akamai Technologies announced on Thursday an expansion of its agreement with Anthropic to $11.6 billion in cloud services over the next seven years. It is, according to CEO Tom Leighton, the largest contract in the history of the Cambridge, Massachusetts, company, a dot-com era veteran now cashing in on its reinvention as an infrastructure provider for AI. The deal includes additional commitments of up to $9 billion in revenue that, if they materialize, would lift the total to around $20 billion, according to figures provided by Akamai itself.

The contract did not come out of nowhere: back in May, Akamai had already sent its stock soaring by announcing a $1.8 billion agreement with an unidentified "leading frontier model provider." Neither party confirmed it at the time, but filings with the SEC (the US stock market regulator) show that the contract was signed on May 5 and corresponded to Anthropic. This week's news is, in essence, the public maturing of that relationship.

The financial terms, detailed by chief financial officer Ed McGowan on the company's earnings call, point to a long timeline with spending running ahead of revenue: Akamai will invest $5.5 billion over the next two years to support the contract and does not expect to generate revenue from it in 2026. The first earnings, of between $150 million and $300 million, would arrive in the second half of 2027, growing into 2028; from there, the company estimates roughly $1.7 billion a year on a recurring basis. Anthropic, for its part, received a warrant (a right to buy shares at a price set in advance) on 7.7 million Akamai shares, around 5% of its equity.

The market reacted more cautiously than in May: the stock rose to $135, far from the $160 it touched on the first announcement. Leighton framed the deal as a bet on turning Akamai into "the infrastructure provider for the next generation of AI-powered applications," citing uses already under way among its customers: voice agents for customer service, translation, speech recognition for subtitles, robotics and simulation rendering.

There are two readings that go beyond the financial headline. The first is strategic: Anthropic is diversifying its compute supply chain by leaning on an edge network player (edge/CDN) outside the trio of hyperscalers —Amazon, Google, Microsoft— that are also its investors. Relying exclusively on a cloud provider that is also a shareholder and potential competitor is precisely the tension that has defined the relationship between OpenAI and Microsoft; adding Akamai gives Anthropic an infrastructure partner without that conflict of interest, even if on a smaller scale than a pure hyperscaler.

The second reading is the one that really matters for understanding where this kind of contract is heading. The 5% warrant is not an accounting footnote: it is part of a pattern that keeps repeating in this cycle's big AI infrastructure deals, where the compute provider receives equity from the customer it bills, and both share prices end up intertwined with the success of the same contract. As sector context, it has already been seen in similar arrangements between chipmakers and AI labs, and it feeds a legitimate question about circular financing: how much of these companies' valuations depends on revenue that does not yet exist and on shares that rise because the market takes for granted that it will. The fact that Akamai does not forecast revenue from this contract until 2027 —and that its stock has not repeated May's enthusiasm— suggests part of the market is already applying that discount.

Our reading is that this deal is a good portrait of the industry's current moment: AI labs are buying compute capacity years in advance, with contracts that lock in future growth at the price of exposing their providers —and themselves— to whether demand for AI agents, voice and enterprise automation grows as fast as the projections promise. If it does, these contracts are literally the plumbing on which the abundance of cheap compute that underpins the long-term thesis is built: more capacity, cheaper and more widely distributed, is a precondition for AI to stop being a corporate luxury. If demand disappoints, companies like Akamai —which has already committed $5.5 billion in spending with no revenue in sight— will be the first to feel the brakes.