Analysis
Anthropic will pay Akamai $11.6 billion over seven years under a newly disclosed cloud computing deal, according to TechCrunch.
Another Name On Anthropic's Compute Bill
The Akamai commitment joins a growing list of multibillion-dollar compute agreements Value Add VC has tracked Anthropic signing this year:
- Akamai -- $11.6 billion over 7 years (this deal)
- Meta -- $10 billion compute lease
- Riot Platforms -- $9.1 billion compute deal
- Debt facility -- $15 billion, closed Sept. 3, led by Morgan Stanley, Goldman Sachs, JPMorgan and Citigroup
Together those commitments give a sense of how much infrastructure spending Anthropic has locked in ahead of its own IPO.
Akamai has spent 2026 repositioning itself from a legacy content-delivery-network business toward cloud and edge compute, and landing a seven-year commitment from one of the best-funded AI labs is a meaningful validation of that pivot -- Akamai now competes for AI workloads against both hyperscalers (AWS, Google Cloud, Microsoft Azure) and AI-native neoclouds like Crusoe and Nscale, the British operator that itself just disclosed a $3.36 billion pre-IPO financing round this week.
For Anthropic, the Akamai contract is one more data point in a pattern Value Add VC has tracked all year: rather than concentrating compute spend with a single cloud partner, the company is deliberately spreading multibillion-dollar, multi-year commitments across several counterparties -- a hedge against any single vendor's capacity constraints, but also a structural reason its aggregate contracted-spend total keeps climbing every time a new deal surfaces.
A seven-year, $11.6 billion commitment is announced capital, not capital already spent -- Anthropic's actual draw-down pace depends on training and inference demand holding at its current trajectory, and the company's revenue, while growing fast, still trails the scale of its combined compute commitments across Akamai, Meta, Riot Platforms and its own debt facility.
Anthropic's now-public web of compute deals across at least four separate counterparties this year alone makes it one of the most heavily cross-obligated AI labs by contracted infrastructure spend, ahead of its own IPO.
