Dossier · Ongoing

Anthropic's hunt for compute

Billion-dollar contracts, credit structures and an in-house chip team: how Anthropic has been assembling compute for Claude since July 2026.

Anthropic needs more compute for Claude than its own capital can cover – and since the summer of 2026 it has been sourcing it through a chain of ever larger constructions: expanding its compute team, negotiating a ten-billion-dollar arrangement with Meta, a roughly 200-billion-dollar credit package tied together by Google and Broadcom, a six-year contract with the Norwegian start-up Volta, and most recently an in-house chip design team.

This dossier tracks the individual agreements, their financing models and the partners involved. A recurring pattern: Anthropic has no credit rating, so special purpose vehicles, letters of credit and infrastructure investors carry the contracts – and several of the deals remain officially unconfirmed by the company to this day.

Timeline

  1. Google Ties $200 Billion in Loans to Anthropic Chip Deal

    A web of special purpose vehicles, loans, and guarantees gives Anthropic access to Google's AI chips – despite lacking a credit rating.

  2. Anthropic secures computing power for 10 billion dollars

    A seven-month-old start-up secures billions for a data center in Norway – according to Bloomberg, Anthropic is behind it.

  3. Anthropic builds its own chip design team for Claude

    The AI provider is specifically hiring semiconductor engineers to design chips and language models together and reduce the operating costs of Claude.

  4. Anthropic builds data centers with Macquarie and GIC

    The AI company and the investors Macquarie and GIC are founding the platform Theseus Infrastructure for new data centers in the USA.