Bitcoin miners divert billions to AI, may regret within year

Bitcoin miners are redirecting billions of dollars and constrained grid capacity into data centers for artificial intelligence workloads as mining revenue weakens, industry figures indicate.

Bitcoin has traded near $64,000 in recent weeks, about 50% below its October peak. The April 2024 halving reduced the block subsidy to 3.125 BTC. An industry tracker reports miners’ daily revenue is roughly 40% lower year over year and averaged about $28.5 million over a 30-day period. Hashprice, a measure of miner earnings per unit of computing power, has fallen to about $30 per petahash per second per day, putting older and less-efficient machines at or below breakeven depending on electricity costs.

Because application-specific mining machines cannot run AI workloads, operators are selling space, power and cooling to customers that run GPU clusters. Several large miners have signed multi-year compute contracts, shifted investment away from new mining equipment and focused on hosting and operating high-performance computing infrastructure, according to company filings.

Core Scientific reported colocation revenue of $136.7 million in its most recent quarter, up from $10.6 million a year earlier, while its self-mining revenue fell 66% to $21.5 million. Colocation represented about 83% of the company’s revenue and produced a 59% gross margin; the self-mining business recorded a negative gross margin.

CoinShares has tallied more than $70 billion of announced AI and high-performance computing contracts by public miners and estimates listed miners could derive as much as 70% of revenue from AI by the end of 2026, up from roughly 30% today. Fred Thiel, CEO of MARA Holdings, noted that AI customers pay higher rates for reliable electricity and long-term capacity, increasing revenue per unit of power.

Industry estimates place Bitcoin mining infrastructure costs at about $700,000 to $1 million per megawatt, while facilities built for AI workloads can cost around $8 million to $15 million per megawatt. The Bank for International Settlements estimates the five largest cloud providers could spend more than $1 trillion on AI-related capital expenditure across 2025 and 2026 and warned that intense competition could produce excessive investment.

André Dragosch, head of research at Bitwise Europe, warned that demand for AI compute, including from advanced applications, may take longer to develop and that Bitcoin could recover sooner than some expect. If AI-ready capacity expands faster than commercial demand, colocation pricing could compress. If Bitcoin prices or hashprice improve, the return gap between mining and hosting would narrow. Analysts identify a central risk in miners committing land, long-term power contracts and large capital sums to AI projects that may take time to reach full commercial scale.