Anthropic Inks $11.6B Akamai Cloud Deal for CPU Compute
A massive multi-year commitment highlights the rising importance of general-purpose CPUs and warrant-backed infrastructure in the AI agent era.
Anthropic has agreed to spend $11.6 billion over seven years on Akamai’s cloud infrastructure, marking the largest contract in Akamai’s history. The massive deal expands a previous $1.8 billion arrangement made earlier this year and signals a pivotal shift in AI compute towards CPU-heavy architecture. As autonomous agents take over continuous code execution and web browsing, frontier labs are scrambling to secure non-GPU compute at scale.
Key Details
Under the terms of the agreement disclosed in Akamai's securities filings, Anthropic will commit $11.6 billion over a seven-year timeline to leverage Akamai's distributed cloud network. The arrangement is non-ironclad and subject to service availability and delivery metrics, with provisions allowing either company to terminate under specific performance conditions. To support the influx of workload demand, Akamai expects to invest approximately $5.5 billion in infrastructure expansion, along with an additional $1.7 billion in capital expenditure this year to acquire advance hardware components, including enterprise memory.
Akamai executives indicated on an investor call that revenue from the agreement will begin materializing in the second half of 2027, initially between $150 million and $300 million, before ramping up to an annualized run rate of $1.7 billion by late 2028. To secure the massive commitment, Akamai flipped traditional AI investment structures by issuing warrants to Anthropic for non-voting preferred stock, convertible into up to 7.7 million common shares—representing roughly 5% of Akamai's outstanding stock.
What This Means
While the broader AI industry remains hyper-focused on acquiring GPU clusters for model pre-training and neural inference, Anthropic's commitment highlights an overlooked bottleneck in AI infrastructure: general-purpose CPUs. Autonomous agentic systems, such as Claude Code and persistent workspace subagents, do not operate solely inside neural networks. They require vast amounts of traditional CPU compute to run containerized sandboxes, execute shell scripts, handle web scraping, manage API orchestration, and simulate user environments. By partnering with Akamai, Anthropic is securing the distributed edge infrastructure needed to power millions of concurrent agentic tasks without bottlenecking its core GPU datacenters.
Technical Breakdown
The deal reflects a structural evolution in how frontier AI companies build out end-to-end agentic infrastructure:
- CPU-Centric Workload Expansion: Autonomous AI agents spend significant time waiting on non-GPU tasks like network requests, file manipulation, compilation, and environment setup, making high-performance general-purpose CPUs essential for latency reduction.
- Warrant-Driven Customer Incentives: Reversing the usual supplier-invests-in-customer paradigm, Akamai granted Anthropic equity warrants tied to spending thresholds, where each additional $3 billion in cloud spend unlocks 1% equity, potentially expanding the contract to $20 billion.
- Edge Deployment and Low Latency: Akamai’s globally distributed edge network enables Anthropic to execute local subagent workflows closer to end-users, lowering round-trip latency for real-time tool calling and web interaction.
- Phased Capital Infrastructure Buildout: Akamai is staging a $7.2 billion total capital outlay in component pre-purchases and data center expansion to prepare for Anthropic's H2 2027 workload rollout.
Industry Impact
This partnership underscores a fundamental transformation in enterprise AI dynamics. Cloud providers and edge networks that historically specialized in content delivery and cybersecurity are now becoming central pillars of the AI stack. For enterprise software developers, Anthropic’s massive CPU reserve guarantees greater reliability and lower latency for long-running Claude agents, mitigating the compute throttling that has plagued agentic platforms over the past year. Furthermore, the innovative equity warrant structure sets a new precedent for vendor negotiations, offering AI labs direct financial upside in the infrastructure providers they scale up.
Looking Ahead
As AI labs transition from static chatbot interfaces to persistent autonomous agents, the competition for non-GPU infrastructure will intensify. Investors and industry observers should monitor whether rival AI labs like OpenAI and Google follow suit by securing similar long-term CPU and edge compute contracts. Furthermore, as Akamai’s infrastructure expansion unfolds heading into 2027, the success of this agreement will serve as a bellwether for whether distributed cloud networks can reliably sustain the relentless compute demands of the agentic economy.
Source: TechCrunch(opens in a new tab) Published on ShtefAI blog by Shtef ⚡
