Anthropic signs $1.8b computing deal with Akamai

Anthropic signed a US$1.8 billion computing deal with Akamai Technologies, a US cloud and cybersecurity firm, to add capacity as demand for its AI software rises.

Akamai had disclosed a long-term cloud agreement with an unnamed frontier model provider in its earnings statement, and its shares climbed 25% on May 7 before rising about 28% to US$149.05 in the next day’s trading.

The company forecast second-quarter revenue of US$1.08 billion to US$1.10 billion, while Anthropic also said this week it would use computing resources from SpaceX.

🔗 Source: Bloomberg

🧠 Food for thought

Implications, context, and why it matters.

Akamai is betting heavily on AI infrastructure

The agreement runs for seven years and locks in capacity commitments. Akamai expects it to add about US$20 million to US$25 million in fourth-quarter revenue in 2026 1.
The company is moving away from its older content delivery business, which fell 7%. It wants AI infrastructure to make up more of the business 1.
Cloud infrastructure is already Akamai’s fastest-growing unit. It rose 40% year over year to US$95 million in the first quarter 1.
The shift comes with a high price tag. Akamai expects capital spending to reach 40% to 42% of revenue in 2026, which helps explain its careful profit outlook even after the contract win 2.

Anthropic’s pick adds pressure to the AI cloud market

The partnership adds momentum to AI infrastructure built for inference, the real-time running of AI models. Training still relies on huge centralized data centers 3.
Akamai uses more than 4,200 distributed global locations for this work. That setup aims for lower latency, or faster response times. It also targets higher throughput, or more data processed, where centralized cloud models can struggle 3, 4.
Akamai says its distributed setup can cut inference costs by up to 86% against traditional hyperscale infrastructure, meaning the large data center networks run by the biggest cloud providers 3.
As AI apps spread, companies may need more than one kind of infrastructure. That opens room beyond the standard model used by today’s cloud giants 4.