Listed Bitcoin miners are no longer easy to evaluate as a simple combination of coin price, hashrate and electricity cost. Some are retiring mining equipment, selling Bitcoin and converting energized sites into facilities for artificial intelligence and high-performance computing. The shift can improve revenue stability, but a power connection alone does not make a site ready for an AI tenant.
Mining economics created the pressure
CoinShares estimates that the listed mining sector fell below aggregate cash breakeven in the second quarter of 2026. Its weighted average ex-tax cash cost was about $75,500 per bitcoin, while Bitcoin ended the quarter near $58,400. The report also places June’s average hash price at a record low of $27.70 per petahash per second per day.
Those figures are sector estimates, not a cost basis for every operator. Power contracts, fleet efficiency, curtailment income and financing differ by company. They do explain why management teams facing weak mining margins may assign capital to another use for the same electrical capacity.
Grid access has become the scarce input
The report argues that permitted, energized land is increasingly valuable because new data centers face long interconnection queues and local restrictions. It counts 225 moratoriums or development restrictions across 30 US states, with 151 still in force. Existing sites may therefore offer something that a new entrant cannot reproduce quickly: access to large amounts of power.
That advantage comes with an expensive qualification. CoinShares estimates that converting capacity to AI-grade infrastructure can cost roughly $8 million to $15 million per megawatt, compared with about $700,000 to $1 million per megawatt for mining infrastructure. Cooling, redundancy, network connectivity and service-level commitments also determine whether a tenant can use a site. An energized mine is a candidate for conversion, not a finished data center.
Contracts and billings are different milestones
Investors have already placed higher valuations on miners with contracted AI or HPC capacity. CoinShares reports an average enterprise-value-to-next-twelve-month-sales multiple of 12.9 for that group, against 3.7 for miners without such contracts. Yet it also identifies more than $100 billion of disclosed backlog supporting only about $1.1 billion of annualized AI and HPC revenue at the time of the report. Roughly 550 megawatts were billing, compared with more than four gigawatts contracted.
The gap matters because a signed agreement may depend on construction, financing, equipment delivery and phased acceptance before it generates revenue. Backlog can indicate demand, but it should not be read as cash already earned.
A better checklist for the transition
Review each operator at the site level. Check whether power is energized or merely proposed, whether permits and interconnection rights transfer, who finances the conversion, when capacity is scheduled to bill, and whether a tenant has committed creditworthy payments. Separate mining revenue, hosting revenue and contracted backlog rather than adding them into one growth figure.
The strongest operators may use mining as a flexible load while building a broader compute business. Others may absorb large conversion costs without securing durable tenants. The useful question is not whether miners are moving into AI, but how much usable capacity has reached the billing stage and what shareholders must spend to get it there.
Source: BTC-Pulse.