Anthropic Signs $11.6B Deal with Akamai, Betting Big on CPU Computing Power

A Mega-Deal Reshaping the Cloud Computing Landscape

AI startup Anthropic has struck a remarkable cloud infrastructure agreement with content delivery network giant Akamai. Under the terms of the deal, Anthropic has committed to paying Akamai $11.6 billion over the next seven years — with the potential for that figure to grow to roughly $20 billion.

The numbers alone are staggering. A seven-year, $11.6 billion commitment translates to an average of more than $1.6 billion per year in computing expenditure. For an AI company still in the midst of rapid expansion, such a long-term capital commitment speaks to the seemingly insatiable demand for underlying compute power that comes with training and running large-scale AI models.

Why Bet on CPU Instead of GPU?

Perhaps the most intriguing aspect of this deal is that it represents a major bet on CPU computing power — not the GPU resources that the AI industry has been racing to secure.

At a time when NVIDIA GPUs are widely considered the backbone of AI compute, Anthropic's decision to invest heavily in CPU infrastructure stands out. This likely reflects Anthropic's view that CPUs offer unique value in inference serving, data preprocessing, model orchestration, and similar workloads. Not every AI task requires expensive GPUs — a large share of auxiliary computation, concurrent request handling, and cost-sensitive operations can actually be served more efficiently and cost-effectively by CPUs.

For Akamai, this deal also signals an important strategic pivot. As a legacy CDN and edge computing provider, Akamai is using this agreement to embed itself more deeply into the AI infrastructure supply chain.

The fundamental difference between GPUs (Graphics Processing Units) and CPUs (Central Processing Units) lies in parallel computing capability. GPUs feature thousands of small cores optimized for executing large numbers of identical matrix operations simultaneously — exactly what neural network training's matrix-multiplication-heavy workloads demand. CPUs, by contrast, have fewer but more powerful cores, better suited for tasks with complex logic branching and serial dependencies. Across a large model's full lifecycle, the training phase is highly GPU-dependent, but inference — where the deployed model responds to user requests — isn't always exclusively GPU territory. When concurrent request volumes are high but individual computation loads are relatively modest, CPU clusters can have the edge in both latency and cost. Beyond that, a huge volume of "glue layer" workloads — data cleaning, feature engineering, API gateways, load balancing, log processing — are CPU-centric. These needs are substantial in day-to-day AI company operations, yet often overlooked.

An Unusual Equity-Linked Arrangement

The most unconventional element of this deal is a consumption-linked equity incentive that Akamai is offering Anthropic.

Under the agreement, Akamai will grant Anthropic up to 5% of the company's equity, with that stake growing incrementally as Anthropic's spending on Akamai's platform increases. In other words, the more Anthropic spends, the larger its potential ownership stake in Akamai becomes.

This kind of arrangement is extremely rare in traditional cloud service contracts. It creates a deep alignment of interests between customer and vendor: Akamai not only locks in a massive long-term client, but also ties Anthropic's growth to its own stock performance through equity. For Anthropic, it secures guaranteed compute capacity while gaining the opportunity to share in its supplier's upside.

Linking spending behavior to equity rewards has precedent in venture capital and M&A, but is genuinely rare in cloud service contracts. These arrangements are sometimes called "Strategic Equity Incentives," and the core logic is to align a vendor's stock-price incentives with a customer's platform stickiness. For Akamai, giving up a potential 5% stake may seem costly — but if it secures an anchor customer spending more than $1.6 billion per year on average, along with the brand validation that comes from Anthropic's growth story, the long-term return could far outweigh the dilution. It's also worth noting that this structure places a certain constraint on Anthropic's financial flexibility: since more spending means more equity, aggressively cutting expenditures would simultaneously forfeit potential equity gains — effectively creating a "the more you spend, the more locked in you are" dynamic.

The Industry Logic Behind the Deal

Viewed from a broader perspective, this transaction reflects several key trends shaping the AI industry today.

The compute race is intensifying. Leading AI companies are locking up compute supply through ultra-long-term, mega-scale contracts. A commitment in the range of $11.6 billion to $20 billion makes clear that compute has become the most critical strategic resource in the AI race. Whoever can secure stable, cost-controlled computing capacity gains a decisive advantage.

Supplier diversification is a priority. Anthropic already has deep partnerships with cloud giants like Amazon and Google. Bringing Akamai into the mix signals a deliberate effort to diversify its compute sources and reduce dependence on any single supplier.

Legacy infrastructure players are pivoting to AI. Through this deal, Akamai demonstrates a viable path for traditional infrastructure companies to reinvent themselves for the AI era — moving beyond content delivery to stake a claim in the high-growth AI compute supply chain.

As a traditional CDN (Content Delivery Network) provider, Akamai's core business has been caching and distributing web content, video, and other assets from edge nodes around the world to reduce latency and bandwidth costs. The essence of a CDN is "putting data closer to users" — and that distributed infrastructure maps naturally onto the deployment requirements of AI inference services. Pushing model inference capabilities down to edge nodes can dramatically reduce response latency for end users. This means Akamai's transition isn't a simple leap into an unrelated field — it's layering AI compute service capabilities on top of its existing edge network assets. The company's globally distributed data center footprint is likely a key factor in attracting Anthropic. For the cloud computing industry as a whole, this signals that the convergence of edge computing and AI inference services could become a defining battleground in the next wave of infrastructure competition.

Questions Worth Watching

This deal also leaves some open questions. How will Anthropic actually deploy $11.6 billion — or potentially $20 billion — in spending over seven years? Does such an aggressive compute commitment align with its revenue growth trajectory? And how significant a role can the CPU route realistically play in AI infrastructure — is it a complementary option or a long-term strategic priority?

Regardless, this agreement has already become a landmark event in the AI infrastructure space. It captures both AI companies' voracious appetite for compute and signals that the competitive landscape of cloud computing and infrastructure is being rewritten by the rise of AI.