Anthropic announced a partnership with Google and Broadcom for “several gigawatts” of next-generation TPU computing capacity expected to come online starting in 2027, a commitment the company called its largest yet as revenue growth accelerated to an annual rate of $30 billion, up from $9 billion at the end of 2025.
The scale of demand for AI computing now competes directly with bitcoin mining for the same scarce resources: network connections, land permits, cooling infrastructure, and cheap electricity.
We have signed an agreement with Google and Broadcom for several gigawatts of next-generation TPU capacity, coming online from 2027, to train and serve the pioneering Claude models.
– Anthropic (@AnthropicAI) April 6, 2026
A Cambridge tracker estimates that Bitcoin mining consumes around 13 to 25 gigawatts of continuous power worldwide, depending on hardware efficiency assumptions.
Anthropic securing multiple gigawatts from a single deal, on top of existing capacity on AWS Trainium, Google TPUs, and Nvidia GPUs, shows how quickly AI is becoming a leading contender for the same energy infrastructure that miners depend on.
And Anthropic is one company. OpenAI, which raised $122 billion last week and describes computing as a “strategic moat,” draws on an even broader infrastructure portfolio spanning five cloud providers and four chip platforms.
The overall development of AI computing now represents one of the largest sources of new electricity demand in the United States, arriving at the same time that Bitcoin miners are deciding whether to mine Bitcoin or lease their infrastructure to AI companies.
This decision is increasingly going in one direction. Core Scientific has converted a significant portion of its mining capacity into AI hosting through an agreement with CoreWeave. Iris Energy and Hut 8 have increased their AI and high-performance computing revenues. Riot Platforms, MARA Holdings, and Genius Group revealed that they sold more than 19,000 BTC from their treasuries last week, a sign that the mining economy alone is not sufficient to sustain operations at current prices and difficulty levels.
A Bitcoin miner running at one gigawatt capacity earns revenue that fluctuates depending on the price of Bitcoin and network difficulties. The same gigawatt leased to an AI company earns a contract rate with predictable cash flows.
At $69,000 worth of bitcoin, with difficulty at unprecedented levels and energy costs rising alongside all the other industrial consumers competing for the same network capacity, AI rental often pays better.
The revenue numbers behind the expansion tell their own story. Anthropic said the number of business customers spending more than $1 million a year on Claude doubled from 500 to more than 1,000 in less than two months.
However, none of this means that Bitcoin mining is dying. The network’s hashrate continues to reach record levels above 1 zetahash per second.
But the miners surviving the current cycle may look less like energy companies producing Bitcoin and more like infrastructure companies mining Bitcoin in parallel while leasing their real asset, cheap energy at scale, to an AI industry that can’t build data centers fast enough.
