Skip to content

AI

Anthropic signs $11.6B check to Akamai, and it has exit doors

Anthropic will spend $11.6 billion over seven years on Akamai's cloud, the biggest deal in Akamai's history. The fine print lets either side walk, so read the asterisks, sapiens.

TL;DR

  1. Anthropic commits $11.6 billion to Akamai's cloud over seven years, Akamai's largest deal ever.
  2. That is more than six times the $1.8 billion deal Bloomberg reported in May.
  3. It is not ironclad: delivery and availability targets apply, and either side can exit.

What happened

Akamai announced on Thursday that Anthropic will spend $11.6 billion on its cloud infrastructure over seven years. It is the largest deal Akamai has ever signed.

For scale, Bloomberg reported a $1.8 billion agreement between the two companies back in May. The new commitment is more than six times that. Anthropic's compute appetite, it turns out, does not do portion control.

Now the asterisk. According to Akamai's securities filing, the money depends on Akamai hitting certain delivery and service-availability requirements. Either company can also end the agreement under specific conditions. So this is a big number with trapdoors, not a vault.

The money will not show up quickly either. Akamai expects no revenue from the deal this year.

For 2027, executives guided investors to between $150 million and $300 million, with the money kicking in during the back half of the year. 7 billion.

Building all that capacity is not free. Akamai expects to spend about $5.5 billion to do it.

Why it matters

Most AI infrastructure headlines are about GPUs. This one is about CPUs, the general-purpose chips that run code and browse the web. Demand for them has grown as AI agents take on more tasks, per the reporting.

That is the interesting bet here. If agents are going to click, fetch and execute all day, somebody has to supply the unglamorous silicon underneath. Anthropic is lining up that supply now.

Akamai did not say what Anthropic will use the capacity for, so treat any specific theory as a guess. What the filing does show is a company locking in compute years ahead and spreading that risk across more than one kind of chip.

It also shows how the money flows in this business.

A customer commits billions, the supplier commits billions of its own to build the capacity, and everyone agrees to review the arrangement if the servers do not show up on time. Romantic, in a procurement sort of way.

The headline number is huge, but the exit clauses are part of the deal too.

For the rest of us, the signal is simple. Frontier AI labs are still buying capacity faster than most companies can plan a budget, and the supply side is scrambling to build it.

What to do with it

  • Read the conditions, not just the headline.Big cloud numbers often carry service requirements and exit rights, so check them before you quote the total.
  • Watch the CPU angle.If your stack leans on agents, plan for general-purpose compute costs too, not only GPU bills.
  • Track Akamai's 2027 numbers.The $150 million to $300 million guidance is the first real test of whether this ramps on schedule.
  • Diversify your own compute.If a frontier lab spreads its bets across providers, your AI workloads probably should too.

Ciao, sapiens. The machines are signing seven-year leases. Check your own rent.