Bitcoin miners have signed over $100 billion in AI contracts, despite only generating $1.1 billion in annualized revenue. The transition to AI has led to a surge in investor interest, with companies with contracted AI or HPC capacity trading at a premium compared to those without such agreements. This premium is largely due to the valuable asset miners have accumulated for Bitcoin mining – grid-connected power.
The scarcity of power has turned mining sites into premium AI assets, as new data-center projects face longer permitting processes and congested power grids across the US. CoinShares reported numerous moratoriums and restrictions on data-center development, with some states introducing statewide pauses on environmental permits for large facilities. The US grid interconnection queue of 2,600 gigawatts further complicates the situation, with projects waiting years to become operational.
The potential revaluation of existing mining campuses is significant, with some data centers in Northern Virginia valued at $27 million per megawatt. Despite the high cost of retrofitting BTC mining sites for AI use, the economics justify the expense. AI infrastructure generates approximately $1.5 million in annualized profit per megawatt, three times more than Bitcoin mining under current conditions.
Some operators are absorbing losses and abandoning equipment to accelerate the transition to AI. Core Scientific paid $41.9 million to terminate a mining equipment agreement, redirecting infrastructure towards AI and HPC customers. Keel Infrastructure shut down its Bitcoin mining operations and is expected to report no mining revenue in the third quarter.
The shift away from Bitcoin mining is evident across the industry, with companies like Cipher Digital and TeraWulf focusing on AI and HPC leases. Approximately 35 EH/s of installed capacity is scheduled to leave publicly listed miners as conversions continue. The shift could be challenging to reverse, even as Bitcoin’s recovery improves mining economics.
The industry now faces the challenge of turning signed AI contracts into functioning data centers. Only 550 MW of the contracted 4 GW capacity is currently billing, leaving a significant backlog dependent on construction, financing, and deployment. Companies like Core Scientific, Cipher, and IREN are making progress in converting their facilities for AI use.
As the industry’s AI and HPC revenue run rate is expected to double, the gap between signed contracts and revenue generation is set to narrow. However, the risk remains that construction, financing, or power infrastructure may not arrive quickly enough. Companies that complete projects on schedule will begin putting cash flow behind their valuations, while others may remain priced for an AI business that exists mostly in backlog.