Anthropic's $11.6 Billion Akamai Deal Is a Bet on CPUs, Not GPUs. Here's Why That Matters for AI Stocks.
- On September 24, 2026, Akamai Technologies announced a $11.6 billion, seven-year commitment from Anthropic for cloud computing, mainly CPU capacity. Anthropic can expand it to about $20 billion.
- Akamai issued Anthropic warrants for up to about 5% of its stock. About 2% vests with the initial commitment, plus roughly 1% for each additional $3 billion of spending.
- Akamai expects $150-$300 million of revenue from the deal in 2027, starting in the second half, rising to about a $1.7 billion annual pace by the end of 2028.
- Akamai plans about $5.5 billion of capital spending to build the capacity, plus about $1.7 billion for components this year.
- Akamai shares rose as much as 17% in after-hours trading after the announcement.
A different kind of AI deal
For three years, AI infrastructure has meant one thing to most investors: GPUs, the graphics chips made mostly by Nvidia that train and run AI models.
On September 24, Akamai Technologies announced a deal about something else. Anthropic, the AI company behind the Claude models, committed $11.6 billion over seven years to Akamai's cloud, mainly for CPU capacity. The commitment can grow to about $20 billion.
Akamai shares jumped as much as 17% in after-hours trading.
Why AI agents need CPUs
Think of a GPU as a specialist and a CPU as a generalist.
- GPUs do massive amounts of parallel math. That's what training and running a large language model requires.
- CPUs handle everything else: running code, calling other software, browsing web pages, moving data and managing tasks.
A chatbot answering one question mostly uses GPUs. An AI agent that writes and runs code, searches the web and completes multi-step tasks also needs a lot of CPU time. As agents take on more work, CPU demand grows too. TechCrunch described the deal as a bet on "a less-hyped corner of AI infrastructure."
What's in the deal
| Term | Detail |
|---|---|
| Commitment | $11.6 billion over 7 years |
| Expansion option | Up to about $20 billion |
| Warrants to Anthropic | Up to ~5% of Akamai; ~2% vests on the initial commitment, ~1% per extra $3 billion |
| Revenue timing | $150-$300 million in 2027 (second half), ~$1.7 billion annual pace by end-2028 |
| Akamai capex | ~$5.5 billion to build capacity, plus ~$1.7 billion for components this year |
The underlying master services agreement dates to May 2026. The two project plans behind the $11.6 billion were signed September 18.
The warrant: a pattern worth understanding
Akamai is giving its customer the right to own up to 5% of the company. The more Anthropic spends, the more stock it can get.
This structure is becoming common in AI. Suppliers hand equity or financing to big AI customers to win long contracts. It locks in demand, but it also means:
- Existing shareholders get diluted if the warrants are exercised.
- The customer's success and the supplier's stock become linked. If Anthropic grows, both win. If AI spending slows, both feel it.
We explain the broader pattern, and its risks, in Nvidia, OpenAI and Circular Financing.
The risks for Akamai shareholders
- Big spending first, revenue later. About $5.5 billion of capex comes before most of the revenue. Akamai's annual revenue has been roughly $4 billion, so this is a very large bet relative to its size.
- Customer concentration. One client could become a large share of revenue by 2028.
- Execution. Akamai has to secure chips, power and data-centre space on schedule.
- Higher interest rates. Funding billions in capex costs more with the 10-year Treasury above 5%.
What it signals for AI investors
- AI spending is broadening. Money is flowing beyond GPUs to CPUs, networking, power and edge cloud providers.
- Server CPU makers such as AMD, Intel and Arm-based chip designers could benefit if agent workloads keep growing.
- Anthropic is spreading its compute across many partners, including Amazon, Google, Microsoft and AMD. Anthropic CEO Dario Amodei said the company doesn't take part in these deals at the "same scale as some other players."
If you'd rather not pick winners, see How to Invest in AI Without Picking Winners.
The honest uncertainty
Multi-year AI contracts are only as strong as the customer's growth. AI demand is booming today, but these deals commit suppliers to spend now based on demand years out. The Akamai deal shows where AI is heading, toward agents that do real work. Whether the economics hold up will show in Akamai's results from 2027 on.
Frequently asked questions
What is the Akamai-Anthropic deal?
Announced September 24, 2026, it's a seven-year, $11.6 billion commitment by AI company Anthropic to buy cloud computing capacity from Akamai, mostly CPU-based, with an option to expand to about $20 billion. Akamai also issued Anthropic warrants for up to about 5% of its shares.
Why do AI companies need CPUs if GPUs train the models?
GPUs do the heavy math of training and running AI models. But AI agents also run code, call tools, browse websites and manage workflows, and those general-purpose tasks run on CPUs. As agents do more work, demand for CPU capacity grows alongside GPU demand.
What does it mean that Akamai gave Anthropic warrants?
A warrant is a right to buy shares at a set price. Akamai giving warrants to its customer is an incentive: the more Anthropic spends, the more of Akamai it can own. It sweetens the deal for Anthropic, but it dilutes existing Akamai shareholders if exercised.
Is Akamai stock a good AI investment?
The deal gives Akamai a large, long-term customer, but it also requires billions in upfront spending and concentrates revenue in one client. The payoff depends on Anthropic's continued growth and on Akamai delivering the capacity on time. RiskStock doesn't give personal investment advice.
Primary sources
Data & disclaimer: This article is for educational purposes only and is not financial or investment advice. Figures reflect data available as of Sep 26, 2026, and conditions change — always confirm current pricing, rates and rules with the provider before you act. Written by Elizabeta Dimoska. See our editorial standards and disclosure.
Comments