humanity and google This is a deal that gives the artificial intelligence company access to up to 1 million of Google’s custom-designed Tensor Processing Units (TPUs).
The deal, worth tens of billions of dollars, is the company’s largest TPU commitment to date and is expected to bring well over gigawatts of AI computing power online by 2026.
Industry estimates put the cost of a 1 gigawatt data center at about $50 billion, of which about $35 billion is typically allocated to chips.
While competitors have touted loftier plans, including OpenAI’s 33-gigawatt “Stargate” head, Anthropic’s move is a quiet power play rooted in execution rather than spectacle.
Founded by former OpenAI researchers, the company intentionally adopts a slower, more stable ethos that is efficient, diversified, and focused on the enterprise market.
Key to Anthropic’s infrastructure strategy is a multicloud architecture.
The company’s Claude family of language models runs across Google’s TPUs. Amazon’s Custom Trainium Chip, and Nvidia’s GPU. Each platform is assigned a specialized workload such as training, inference, or research.
Google said the TPU provides Anthropic with “strong cost performance and efficiency.”
“Anthropic and Google have a long-standing partnership, and this latest expansion helps us continue to scale the compute needed to define the frontiers of AI,” Anthropic CFO Krishna Rao said in the release.
Anthropic can distribute workloads across vendors, allowing you to fine-tune it to meet price, performance, and power constraints.
According to people familiar with the company’s infrastructure strategy, this model extends compute per dollar more than what is locked into a single vendor’s architecture.
Google is leaning toward this partnership.
“Anthropic’s choice to significantly expand its use of TPUs reflects the excellent cost performance and efficiency that our team has seen with TPUs over the past few years,” Google Cloud CEO Thomas Kurian said in a release, highlighting that the company’s 7th generation ‘Ironwood’ accelerators are part of a mature portfolio.
Claude’s incredible revenue growth
Anthropic’s growing demand for computing reflects its explosive business growth.
The company’s annual revenue is now approaching $7 billion, and Claude has helped more than 300,000 businesses. This is a staggering 300x increase over the past two years. The number of large customers, each with more than $100,000 in run-rate revenue, has increased nearly seven times over the past year.
The company’s agent coding assistant, Claude Code, generated $500 million in annual revenue within just two months of launch, which Anthropic claims makes it the “fastest growing product” in history.
Google is driving the next phase of Anthropic’s computing expansion. Amazon We will continue to be your most deeply embedded partner.
The retail and cloud giant has invested $8 billion in Anthropic to date, more than double Google’s confirmed capital of $3 billion.
Still, AWS is considered Anthropic’s primary cloud provider, and its influence is structural as well as financial.
The supercomputer custom-built for Claude, known as Project Rainier, runs on Amazon’s Trainium 2 chips. This change is important not only for speed, but also for cost. Trainium avoids the premium margins of other chips and enables more compute per dollar spent.
Wall Street is already seeing results.
Rothschild & Company Redburn analyst Alex Heisl estimated that Anthropic added 1 to 2 percentage points to AWS growth in the fourth quarter of last year and the first quarter of this year, and expects its contribution to exceed 5 percentage points in the second half of 2025.
Wedbush’s Scott Devitt previously told CNBC that if Claude becomes the default tool for enterprise developers, usage will be directly reflected in AWS’s revenue, and he believes this dynamic will drive AWS growth “for many years.”
Meanwhile, Google continues to play an important role. In January, the company agreed to a new $1 billion investment in Anthropic, adding to its previous $2 billion and 10% stake.
Importantly, Anthropic’s multi-cloud approach proved resilient during Monday’s AWS outage, which did not impact Claude thanks to its diverse architecture.
Still, Anthropic is not a favorite to win. The company maintains control over model weights, pricing, and customer data, and has no exclusivity over any cloud provider. This neutral position could be key as competition among hyperscalers intensifies.
WATCH: Anthropic’s Mike Krieger talks about new model releases and the race to build real-world AI agents
