Anthropic PBC has officially signed an $11.6 billion, seven-year contract with Akamai Technologies Inc. to secure massive amounts of computing power. This deal marks another significant expansion in the AI developer’s growing list of data center agreements, signaling a long-term commitment to infrastructure scale that goes beyond typical cloud rental models. The magnitude of the contract highlights how critical stable, high-volume compute access has become for frontier model developers.

The Shift to Generalist Chips

Unlike many high-profile AI deals that focus exclusively on graphical processing units, Akamai will supply Anthropic access to central processing units, or CPUs. These generalist chips are seeing renewed demand in data centers as they play a crucial supporting role in AI services. This move suggests Anthropic is optimizing for the broader orchestration and data handling layers required to keep massive models running efficiently, rather than just raw inference speed.

Equity Ties Deepen the Partnership

The financial structure of the deal includes more than just cash flow; Akamai has issued a warrant giving Anthropic the right to buy Series B shares. These shares are priced at $111.33 each and are convertible into 7.7 million shares of common stock. This equity component aligns the incentives of both companies, effectively making Anthropic a stakeholder in the infrastructure provider that powers its models.

Key Takeaways

  • Anthropic has secured $11.6 billion in compute resources over a seven-year period with Akamai Technologies.
  • The deal focuses on CPUs, indicating a strategic emphasis on generalist chip support for AI infrastructure.
  • Akamai issued a warrant for 7.7 million shares at $111.33 per share, tying Anthropic's success to Akamai's equity performance.

The Bottom Line

This isn't just a rental agreement; it's a strategic entrenchment. By locking in CPU capacity and taking equity in Akamai, Anthropic is betting that the bottleneck for the next generation of AI isn't just raw GPU flops, but the holistic efficiency of the data center stack. The market should watch how this CPU-heavy approach impacts latency and cost-per-token for their upcoming models.