The latest sign of the A.I. industry’s unrelenting hunt for computing power comes from an expanded agreement between Anthropic and Google—a deal that, like several others struck in recent months, underscores the rise of circular arrangements across Silicon Valley. Under the new agreement, Google will provide Anthropic with well over one gigawatt of computing capacity by 2026, the companies announced yesterday (Oct. 23).
Anthropic noted that the deal is worth “tens of billions of dollars” but didn’t provide an exact figure. The partnership further deepens the startup’s ties with Google, which has already invested about $3 billion in Anthropic and is expected to supply the company with up to 1 million of its custom A.I. chips, called tensor processing units (TPUs).
Such partnerships are increasingly essential as leading A.I. startups scale at a breakneck pace. Anthropic, which now serves over 300,000 business customers, said the number of clients generating more than $100,000 in annual revenue has grown nearly sevenfold in the past year. “Anthropic and Google have a longstanding partnership, and this latest expansion will help us continue to grow the compute we need to define the frontier of A.I.,” said Krishna Rao, Anthropic’s chief financial officer, in a statement.
Founded in 2021 by CEO Dario Amodei and several former OpenAI employees, Anthropic positioned itself as a safety-focused alternative to early A.I. players. Best known for its chatbot Claude, the company recently hit a $183 billion valuation and is reportedly on track to generate $9 billion in annual revenue.
Despite its closer ties with Google, Anthropic emphasized that it remains committed to its “primary training partner,” Amazon, which has invested $8 billion in exchange for providing compute through its chips and A.I. cluster Project Rainier. The company also continues to rely on Nvidia’s GPUs as part of what it calls a “multi-platform approach.” Anthropic said it will keep investing in additional compute capacity as demand grows.
Anthropic’s mutually beneficial partnerships with Google and Amazon reflect a broader trend: a broader industry trend: a growing web of interconnected A.I. partnerships between model developers and compute providers, each investing in and purchasing one another’s technology. OpenAI has been at the forefront of this shift, announcing a flurry of major deals in recent months, including an agreement with AMD to access six gigawatts of computing power, a deal with Nvidia to access 10 gigawatts of compute, and a $300 billion, five-year partnership with Oracle.
The growing prevalence of such circular arrangements has raised some eyebrows in Silicon Valley, recalling the speculative interdependencies of the dot-com bubble and its eventual crash. But unlike that era, today’s A.I. spending is bolstered by stronger capitalization and clearer monetization potential, said Stephanie Aliaga, global market strategist for JPMorgan Chase, in a blog post earlier this month.
Still, Aliaga cautioned that the concern isn’t misplaced. “The scale of spending is enormous, the pace unprecedented, and some assumptions around ROI, like the useful lives of assets, remain open questions,” she said. “History reminds us that enthusiasm can run ahead of reality,” she wrote.
