Anthropic has agreed to spend $11.6 billion over seven years on Akamai’s cloud infrastructure. Akamai announced the deal Thursday. According to TechCrunch AI, that’s more than six times the $1.8 billion agreement between the two companies that Bloomberg reported in May. It’s also the largest contract in Akamai’s history.
⚡ Key Takeaways
- The size: $11.6 billion over seven years. It could grow to about $20 billion.
- The chips: The deal is about CPUs, not GPUs. That’s a less-hyped part of AI infrastructure that’s getting more important as AI agents take on more work.
- The twist: Akamai, the supplier, is giving Anthropic, the customer, a warrant for up to about 5% of its stock. Usually the money flows the other way.
- The fine print: The deal isn’t locked in. It depends on Akamai hitting delivery and service-availability targets, and either company can walk away under certain conditions.
- The timeline: Akamai won’t book any revenue from the deal this year. The money starts in the second half of 2027.
🧠 Why CPUs Matter Here
Most AI infrastructure headlines are about GPUs, the chips used to train and run large models. CPUs are the general-purpose chips that run code, browse the web, and handle the everyday computing around a task.
TechCrunch AI notes that CPU demand has grown as AI agents take on more tasks. An agent that writes and runs code, calls tools, or moves around a website needs plenty of ordinary computing alongside the model itself. Akamai didn’t say what Anthropic will use the capacity for. Still, it’s a good guess this is about agent workloads, since agentic coding and computer use are where Anthropic has been pushing hardest.
This is significant because it points to where AI spending goes next. Training clusters get the attention, but running agents at scale needs a very different, more spread-out kind of infrastructure.
💰 The Money Breakdown
Here’s what Akamai executives told investors on Thursday’s call:
- 2027 revenue: $150 million to $300 million, starting in the second half of the year
- End of 2028: Revenue running at about $1.7 billion a year
- Buildout cost: About $5.5 billion to add the capacity
- Extra capex this year: About $1.7 billion to buy components like memory ahead of time
That last number stands out. Memory prices and availability have become a real bottleneck across the AI supply chain, so Akamai is buying early rather than risk shortages later.
🔄 How the Warrant Works
As part of the deal, Akamai gave Anthropic a warrant, basically the right to buy shares at a set price. It covers nonvoting preferred stock convertible into 7.7 million common shares, or up to about 5% of the company, at $111.33 a share.
Here’s how it vests:
- About 2% becomes available once Anthropic makes its first payment.
- Each additional $3 billion Anthropic commits unlocks roughly another 1%.
- Full vesting would push the total deal up by as much as $9 billion, to about $20 billion.
Bloomberg reported this is the first time Akamai has attached a warrant to a cloud deal. The structure flips the usual pattern in circular AI deals, where chipmakers and cloud providers invest in the AI labs that buy their products. Here, the customer gets a stake that grows as it spends more. AMD used a similar setup with OpenAI last year, tying warrants to chip-purchase milestones.
🏗️ The Bigger Picture
This continues Anthropic’s run of big compute deals. Amazon, Google, Microsoft, and AMD have all invested or agreed to invest in Anthropic while also selling it chips or cloud capacity. CEO Dario Amodei told The New York Times last December that Anthropic doesn’t take part in these arrangements at the “same scale as some other players.” An $11.6 billion commitment that could reach $20 billion makes that line harder to hold.
For Akamai, a company best known for content delivery and edge networking, the deal is a coming-of-age moment as an AI cloud provider. Investors noticed. Its shares rose as much as 17% in after-hours trading, The Wall Street Journal reported.
🔭 What to Watch
- Execution risk: Akamai has to hit its delivery and uptime targets, or the deal can unravel.
- More CPU deals: If agent workloads keep growing, expect other labs to lock up CPU-heavy capacity too.
- New deal structures: Supplier-to-customer warrants could become a common way for smaller cloud players to win big AI customers.
The AI compute race is no longer just about who has the most GPUs. Full details are available in the original TechCrunch AI report.