Anthropic has struck one of the largest infrastructure commitments yet seen in the frontier AI sector, agreeing to pay Akamai $11.6 billion over seven years for cloud capacity in a deal that could expand to roughly $20 billion, according to the terms disclosed in the arrangement. The pact highlights how the economics of advanced AI are increasingly being shaped not only by model performance and product demand, but by long-term access to compute, power and specialized hosting capacity.
The deal is notable for both its scale and its structure. Anthropic, one of the leading developers of large language models, is making a major bet on CPU-based infrastructure rather than relying solely on the GPU-heavy supply chain that has dominated the AI boom. That choice suggests a strategic push to diversify compute sources, reduce concentration risk and secure capacity for workloads that may not require the most expensive accelerators. For Akamai, a company better known for content delivery and edge services than for being a central AI infrastructure player, the agreement offers a high-profile foothold in the race to serve frontier model developers.
Compute Beyond GPUs
The AI industry's infrastructure race has largely centered on Nvidia-powered systems, with hyperscalers and model developers competing for scarce accelerators. Anthropic's commitment to Akamai signals a broader view of what AI infrastructure can look like as models scale and operational costs rise. CPUs remain essential for a range of tasks, including inference, orchestration and certain distributed workloads, and the deal indicates that large AI companies are willing to lock in non-GPU capacity if it offers cost, availability or architectural advantages.
The seven-year horizon also matters. Multi-year infrastructure contracts are becoming a defining feature of the sector because they provide predictability in a market where demand can surge faster than supply can be built. For Anthropic, the agreement may help secure a stable base for future growth, especially as competition intensifies among model developers seeking to commercialize enterprise products and expand usage. For Akamai, the contract could help justify further investment in cloud and edge infrastructure tailored to AI customers.
Equity Tied To Spend
What sets the arrangement apart is the equity component. Akamai is giving Anthropic the potential to obtain up to 5% of its stock, with the stake increasing as Anthropic spends more under the deal. That structure is unusual in the cloud market and suggests a deeper strategic alignment than a standard vendor-customer relationship. It effectively ties infrastructure consumption to a possible ownership interest, creating incentives for both sides to sustain and expand the partnership.
The stock-linked feature may also reflect the bargaining power of major AI buyers in a capital-intensive market. As frontier AI companies commit billions to compute, they are increasingly able to negotiate terms that go beyond pricing and capacity alone. In this case, Akamai appears to be using equity potential as a way to secure a long-duration customer, while Anthropic gains a path to participate in the upside of a supplier it is helping to scale.
Strategic Signal For AI
The agreement arrives as AI infrastructure spending continues to accelerate across the sector, with companies racing to build the physical and financial foundations needed to support increasingly capable models. Deals of this kind are becoming strategic signals: they reveal which firms have the balance-sheet strength to commit for years, which suppliers can adapt to AI demand, and how the market is fragmenting across different compute architectures.
For Anthropic, the Akamai pact reinforces its position as one of the best-capitalized and most ambitious players in frontier AI. For Akamai, it marks a potential redefinition of its role in the cloud ecosystem, moving from a traditional internet infrastructure provider toward a more direct participant in the AI compute economy. The long-term outcome will depend on execution, demand and the pace at which AI workloads evolve, but the message is immediate: the next phase of AI competition is being written in infrastructure contracts as much as in model releases.