Anthropic is buying compute at a scale that would have been unthinkable for a cloud contract just a few years ago. On Thursday, September 24, 2026, the AI lab announced an $11.6 billion cloud services agreement with Akamai Technologies — a seven-year commitment that can expand by a further $9 billion, bringing the potential total to roughly $20 billion.
The deal also gives Anthropic a warrant to acquire up to 5% of Akamai's outstanding common stock at $111.33 a share, with about 2% expected to vest under the initial commitment. Akamai shares jumped 15% in extended trading on the news.
For Anthropic, the agreement secures the computing backbone it needs as demand for its Claude models keeps climbing across enterprise customers, developers, and its consumer products. For Akamai, long known as a content-delivery and edge-computing company, it is the largest single customer commitment in the company's history — and a loud signal that it wants a seat at the AI infrastructure table.
A $11.6 billion bet on compute
Cloud deals of this size used to be the preserve of the hyperscalers — Amazon, Microsoft, and Google. Anthropic itself has historically relied on those giants for training and serving its models. Committing $11.6 billion to Akamai over seven years diversifies its compute supply at a moment when data-center capacity remains the industry's most contested resource.
The structure of the deal matters as much as its size. The extra $9 billion expansion option means Anthropic can scale its commitment without renegotiating from scratch, while Akamai gets a revenue pipeline it can plan data-center buildouts around. It is the kind of long-dated arrangement that has become the template for AI infrastructure in the capacity-crunch era.
The surprising detail: it's CPUs, not GPUs
The part of the announcement that raised eyebrows among infrastructure watchers is what Anthropic is actually buying: this is CPU capacity, not GPU capacity. While the industry's attention has been fixed on GPU clusters for model training, a huge share of real-world AI workloads — serving models to users, running agents, preprocessing data, and handling the software around the models — runs on conventional processors.
That detail reframes the deal. Anthropic is not just securing training firepower; it is securing the everyday inference and platform capacity that turns a model into a product millions of people use daily. As AI companies shift from training bigger models to serving more users, the CPU layer of the stack is becoming a strategic asset in its own right.
The trillion-dollar infrastructure race
The Akamai agreement lands in the middle of an AI infrastructure buildout of staggering proportions. The Brookings Institution has estimated the AI infrastructure buildout could reach $10.3 trillion through 2032, a figure that captures data centers, power, networking, and the chips inside them.
Deals are stacking up across the industry. Microsoft, Amazon, and Google have all announced multi-billion-dollar capacity expansions this year, and AI labs are increasingly signing direct agreements with infrastructure providers rather than routing everything through traditional cloud channels. Anthropic's move fits the pattern: lock in capacity early, diversify suppliers, and treat compute the way airlines treat fuel — as the single largest input cost worth hedging years ahead.
The equity warrant adds another layer. By tying itself to Akamai's stock, Anthropic aligns incentives with its infrastructure partner and gets upside if the market keeps rewarding the AI buildout story. Akamai's 15% after-hours pop suggests investors liked the arrangement.
Why it matters for Pakistan
Pakistan sits on the demand side of this story — but the effects are direct. Every dollar of infrastructure investment that lowers the cost of serving AI models eventually shows up as cheaper, faster AI products for Pakistani developers, startups, and freelancers who build on Claude and its competitors.
There is a local opportunity too. As global data-center capacity strains, edge computing and regional cloud presence become more valuable — and Pakistan's IT sector has been pushing to attract exactly this kind of infrastructure investment. The lesson of the Anthropic–Akamai deal for policymakers and investors here is that compute is now treated as strategic national infrastructure, not just IT spending. Countries and companies that plan capacity early will pay less later.
For the thousands of Pakistani developers building AI products on tight budgets, the near-term takeaway is simpler: the infrastructure race is driving inference costs down, and the tools are getting cheaper even as the deals get bigger.
