Anthropic Commits $11.6 Billion to Akamai in 7-Year Cloud Infrastructure Mega Deal

TL;DR

• Anthropic has committed to spend a minimum of $11.6 billion over seven years on Akamai's cloud infrastructure, with total spending potentially climbing to $20 billion with expansions and renewals.
• The deal is centered on massive CPU capacity for AI inference rather than GPU training, leveraging Akamai's globally distributed edge network for low-latency, cost-efficient agentic workloads.
• In an unusual structure, Akamai is granting Anthropic warrants for up to 5% of its stock that vest as Anthropic hits spending milestones, a move that sent Akamai shares soaring.

The Deal at a Glance

Anthropic is making one of the biggest infrastructure bets in its history — and it's not with AWS, Google, or Microsoft.

The AI lab confirmed this week that it has agreed to spend at least $11.6 billion over seven years to use Akamai Technologies' cloud platform, in a partnership that could reach $20 billion in total value if expanded. The agreement, disclosed alongside Akamai's financial filings on September 25, 2026, instantly ranks among the largest cloud contracts ever signed by an AI model developer.

For context, Anthropic already has multi-billion-dollar compute partnerships with Amazon, Google, and Microsoft Azure, plus newer neocloud deals with CoreWeave and others. The Akamai commitment adds a massive new pillar focused almost entirely on CPU-based cloud capacity.

Akamai stock surged more than 15% on the news, adding billions in market value as investors repriced the former CDN company as a serious contender in AI infrastructure.

Why CPUs? Why Now?

At first glance, an $11.6 billion AI deal without GPUs sounds strange. But Anthropic says that's exactly the point.

While training frontier models like Claude still requires tens of thousands of NVIDIA GPUs, running those models for hundreds of millions of users — inference — is increasingly CPU-intensive, especially for agentic AI.

Modern AI agents that browse the web, run code, call tools, orchestrate workflows, and manage long context windows need huge fleets of CPUs to handle orchestration, data preprocessing, retrieval, memory management, and networking alongside GPU accelerators. As Anthropic scales Claude for enterprise automation, coding, and computer use, its CPU needs are exploding.

CPUs are also cheaper, more available, and more power-efficient for many inference tasks than scarce GPUs. By locking in seven years of CPU supply now, Anthropic is hedging against both GPU shortages and rising inference costs.

Industry analysts say the move signals a broader shift: the next phase of the AI cloud race won't just be about who has the most H100s and B200s, but who can deliver high-throughput, low-latency inference at global scale.

Why Akamai?

Anthropic didn't pick Akamai by accident. The Cambridge, Massachusetts-based company operates one of the world's largest distributed networks, with points of presence in more than 4,000 locations across 130+ countries, supercharged by its $900 million acquisition of Linode in 2022.

That edge footprint is ideal for inference. Instead of routing every Claude query back to a centralized mega data center in Virginia or Iowa, Akamai can run workloads closer to users, cutting latency for real-time agents, voice, video, and enterprise applications.

Three factors reportedly sealed the deal:

1. Scale and price: Akamai has been aggressively building out cloud regions and offering compute at significantly lower prices than the Big Three hyperscalers.

2. Distribution: No other provider can match Akamai's edge density, which is critical as Anthropic pushes for faster, more responsive AI experiences worldwide.

3. Availability: With AWS, Azure, and Google Cloud capacity constrained by AI demand, Akamai offered Anthropic dedicated, long-term capacity with room to grow to $20 billion.

Anthropic's infrastructure chief framed it as diversification: GPUs for training in centralized superclusters, CPUs for inference everywhere.

Inside the 5% Equity Kicker

The most unusual part of the agreement isn't the dollar figure — it's the stock.

As part of the deal, Akamai will issue warrants to Anthropic allowing it to purchase up to roughly 5% of Akamai's outstanding shares. The warrants vest in tranches tied directly to how much Anthropic actually spends under the cloud contract.

In other words, the more Anthropic uses Akamai's cloud, the larger an owner it becomes.

The structure is rare in cloud deals, but increasingly popular in the AI boom. Similar equity-linked arrangements have appeared in deals involving CoreWeave, Nebius, and OpenAI's infrastructure partners. For Akamai, it locks in a marquee anchor customer and aligns incentives for seven years. For Anthropic, it offers massive upside if Akamai's transformation into an AI cloud pays off — effectively letting Anthropic profit from the infrastructure buildout it is funding.

If fully vested at current prices, the stake would be worth well over $800 million, with far more upside if Akamai continues to climb.

Filings show the warrants have an exercise price based on recent trading averages and vest over the life of the agreement, protecting Akamai shareholders if Anthropic doesn't meet its $11.6 billion minimum.

What It Means for the AI Cloud Race

The Anthropic-Akamai mega deal redraws the map for AI infrastructure in three ways.

First, it validates Akamai as a neocloud power player. Long seen as a content delivery network facing disruption from cloud giants, Akamai has now landed the ultimate proof point: a frontier lab willing to bet $11.6 billion on its platform.

Second, it confirms that inference is the new battleground. As AI shifts from training giant models to serving billions of agentic requests daily, distributed CPU clouds with global reach may matter as much as GPU supercomputers.

Third, it intensifies pressure on AWS, Microsoft, Google, and Oracle. Anthropic is a close partner to both Amazon and Google, which have invested billions in the startup. Choosing Akamai for this tranche shows Anthropic is willing to go outside its investor ecosystem for better economics and architecture — a warning shot to incumbents.

For Akamai, execution risk remains enormous. It must now build and operate CPU capacity at unprecedented scale while maintaining reliability and margins. For Anthropic, it must actually use — and pay for — $11.6 billion in compute even as models and hardware evolve rapidly.

But if the bet works, both companies win big: Anthropic gets cheap, ubiquitous inference to power the next generation of Claude agents, and Akamai completes its reinvention from CDN to core AI cloud.

AndroGuider Team

Articles written by the AndroGuider team. We try to make them thorough and informational while being easy to read.