(P1) Anthropic’s staggering $200 billion, five-year commitment to Google Cloud reshapes the AI infrastructure landscape, locking in a key customer for Google’s custom chips and challenging Nvidia's dominance in the data center. The deal, which begins in 2027, is for five gigawatts of server capacity and accounts for over 40% of Google's disclosed cloud revenue backlog.
(P2) "This reflects the AI boom's reliance on a few heavily-funded startups," a source familiar with the deals told The Information. The massive outlays from Anthropic and OpenAI now constitute nearly half of the estimated $2 trillion in revenue backlogs across Amazon, Google, Microsoft, and Oracle.
(P3) The agreement provides Anthropic with access to Google's tensor processing units (TPUs), which Google co-develops with Broadcom. This arrangement offers Google higher profit margins compared to reselling Nvidia's in-demand GPUs. For Anthropic, whose server costs are projected to hit $20 billion by 2026, the deal secures a massive amount of computing power for its Claude AI models.
(P4) For investors, this deal solidifies the AI industry into a two-camp battle: Google and its $40 billion investment in Anthropic versus Microsoft's deep partnership with OpenAI. The move sent Alphabet's stock up 2% in after-hours trading, as the market digests a deal that both de-risks Google's massive TPU investment and concentrates a significant portion of its future cloud revenue on a single, high-growth AI firm.
The financial commitments underpinning the current AI development boom are reaching astronomical scales. Anthropic's $200 billion pledge to Google is just one piece of a larger capital puzzle. The AI startup has also secured a multi-billion dollar cloud deal with Amazon and further capacity from CoreWeave.
These circular, mega-deals—where cloud providers invest in AI startups who then spend that money on the provider's infrastructure—are becoming a defining feature of the industry. OpenAI is projected to spend $45 billion on servers in 2026, with some estimates putting its 2029 server bill at a stunning $180 billion for that year alone.
A key factor in Google's ability to secure the massive deal is its custom silicon. By providing its own TPUs instead of Nvidia's GPUs, Google can offer a more cost-effective solution for Anthropic while capturing more of the value chain. This strategy is a direct challenge to Nvidia, whose H100 and other AI-focused GPUs have become the industry standard, creating a multi-billion dollar business.
The deal also highlights the immense capital required to compete at the highest level of AI. Anthropic's own projections, based on optimistic growth scenarios from late last year, saw its total payments to Amazon, Google, and Microsoft reaching $200 billion by 2029. Both Anthropic and OpenAI are banking on revenue growth of 20 to 30 times 2025 levels to justify these expenditures, a prospect some investors view with skepticism.
This article is for informational purposes only and does not constitute investment advice.
