Anthropic has unveiled a partnership with Google and Broadcom to provide 'multiple gigawatts' of next-generation computing capacity, set to come online starting in 2027, marking the company's most substantial commitment to date as it achieves a $30 billion annual revenue run rate from $9 billion at the end of 2025. The scale of AI computing demand now directly competes with bitcoin mining for the same limited resources, including grid connections, land permits, cooling infrastructure, and affordable electricity. A deal has been signed with Google and Broadcom for multiple gigawatts of next-generation computing capacity to train and serve frontier Claude models, starting in 2027. According to a Cambridge tracker, bitcoin mining globally consumes approximately 13 to 25 gigawatts of continuous power, depending on hardware efficiency assumptions.
Anthropic's acquisition of multiple gigawatts from a single deal, in addition to existing capacity across AWS Trainium, Google TPUs, and Nvidia GPUs, demonstrates the rapid growth of AI as a peer-level competitor for the same energy infrastructure that miners rely on. Moreover, Anthropic is just one company, while OpenAI, which raised $122 billion last week and views compute as a 'strategic moat,' is building across an even broader infrastructure portfolio spanning five cloud providers and four chip platforms. The aggregate AI computing buildout is now one of the largest sources of new electricity demand in the United States, coinciding with bitcoin miners' decisions on whether to mine bitcoin or rent their infrastructure to AI companies. This decision is increasingly leaning towards the latter, as Core Scientific converted a significant portion of its mining capacity to AI hosting through a deal with CoreWeave, and Iris Energy and Hut 8 have expanded their AI and high-performance computing revenue.
Riot Platforms, MARA Holdings, and Genius Group disclosed selling over 19,000 BTC from their treasuries last week, indicating that mining economics alone are not sustaining operations at current prices and difficulty levels. A bitcoin miner operating a gigawatt of capacity earns revenue that fluctuates with bitcoin's price and network difficulty, whereas renting the same gigawatt to an AI company earns a contracted rate with predictable cash flows.
At $69,000 bitcoin with difficulty at all-time highs and energy costs rising alongside every other industrial consumer competing for the same grid capacity, the AI rental often pays better. The revenue numbers behind the expansion tell their own story, with Anthropic stating that the number of business customers spending over $1 million annually on Claude has doubled from 500 to over 1,000 in less than two months. While this does not signify the demise of bitcoin mining, the network's hashrate continues to reach record levels above 1 zetahash per second.
However, the miners that survive the current cycle may evolve to resemble infrastructure companies that happen to mine bitcoin on the side while renting their primary asset, affordable power at scale, to an AI industry that is struggling to build data centers quickly enough.