Artificial intelligence startup Anthropic has unveiled a sweeping new agreement with Google and Broadcom to secure multiple gigawatts of next-generation TPU (Tensor Processing Unit) computing capacity, set to come online in 2027. Announcing what it calls the most pivotal investment in its history, Anthropic stated that its annual revenues are on track to surge from $9 billion by the end of 2025 to a projected $30 billion, underscoring the escalating financial stakes in the AI sector.
The competition for energy infrastructure intensifies
The relentless demand for artificial intelligence technologies has triggered a direct contest over energy infrastructure, pitting AI developers against Bitcoin miners. Both industries are finding themselves vying for limited resources such as grid access, land use permits, cooling investments, and inexpensive electricity, driving up competition in the energy market.
According to data from the University of Cambridge, global Bitcoin mining operations consistently consume between 13 and 25 gigawatts of electricity. Anthropic’s move to secure several gigawatts in a single agreement highlights the voracious energy appetite of major AI firms and underscores the intensifying pressure facing crypto miners as they compete for the same infrastructure.
Beyond Anthropic, OpenAI is also making headlines by leveraging recent investments to expand its operations, distributing its workload across five cloud service providers and four chip platforms. These expansive infrastructure plans have made the AI sector one of the fastest-growing sources of electricity demand in the United States.
Miners pursue new revenue streams
Rising energy costs and competitive pressures are beginning to squeeze the profitability of Bitcoin mining. In response, companies like Core Scientific have repurposed a significant share of their capacity for artificial intelligence services through agreements such as the one with CoreWeave. Meanwhile, miners like Iris Energy and Hut 8 have increased their revenues from data centers geared toward high-performance computing and AI workloads. In a sign of shifting strategies, firms including Riot Platforms, MARA Holdings, and Genius Group collectively sold more than 19,000 Bitcoins last week alone, suggesting that maintaining operations under current price and difficulty levels is an uphill battle.
For a miner, income derived from one gigawatt of capacity can swing widely based on Bitcoin’s price and network difficulty. By contrast, leasing that same capacity to an AI company generally offers a steady, contractually-defined cash flow, lending greater financial predictability in uncertain markets.
As power prices increase and mining difficulty intensifies, leasing energy capacity to AI companies can sometimes prove more lucrative than engaging in direct mining operations, prompting miners to reevaluate their business models and seek out more stable income sources.
Anthropic revealed that, within just two months, the number of customers paying over $1 million annually for its Claude product grew from 500 to more than 1,000.
Still, Bitcoin mining has not faded into obscurity. The total computing power of the Bitcoin network continues to set new benchmarks, recently surpassing one zetahash per second, emphasizing the field’s technical endurance even as economic pressures mount.
Looking ahead, the focus for surviving miners may increasingly shift away from self-operated electricity production. Instead, leveraging their existing low-cost energy infrastructure to serve both AI companies and Bitcoin transactions could become the optimal path for adapting to an evolving digital economy.
