The role of miners is shifting from being Bitcoin (BTC) laborers to becoming electricity landlords for artificial intelligence. Recently, as Bitcoin prices have fallen, Bitcoin mining has become unprofitable. Leading mining companies such as Core Scientific and MARA Holdings have successively liquidated their Bitcoin holdings. To date, publicly listed mining firms have collectively sold over 15,000 BTC. At the same time, these companies are repurposing their mining facilities into AI data centers, reducing mining to a secondary activity and marking the beginning of a major industrial transition spanning both the cryptocurrency and AI sectors.
"Redirecting electricity from mining to AI hosting services is the most direct industrial signal ahead of the era of computing power as the standard," said Wang Yingbo, a digital economy scholar at the Shanghai Academy of Social Sciences. From the perspective of a "computing power standard," Bitcoin is merely an early, yet fundamentally flawed, token generated by computing power at a specific historical stage. Leading mining companies liquidating BTC and shifting to AI can be seen as decoupling from the old token system and investing in the infrastructure of the new standard—computing power. This is far from a bear market hedge; it is a historic shift in focus.
Mining has turned into a loss-making venture. After the 2024 Bitcoin halving, mining profitability dropped by approximately 50%, while costs remained unchanged. Starting in the second half of 2025, the continuous decline in Bitcoin prices completely breached the survival threshold for mining companies. Currently, the cost of mining Bitcoin significantly exceeds its price. When cash costs are combined with hardware depreciation, maintenance, land, and other expenses, mining has transformed from a "money printer" into a "cash burner."
Industry estimates suggest that the current cost of mining one Bitcoin is approximately $87,000. As of this writing, Bitcoin's price is $70,343.9, meaning miners lose about $17,000 for every Bitcoin they mine.
Typically, when Bitcoin prices fall below miners' costs, they have two options: shut down operations to minimize losses or sell Bitcoin to sustain operations. However, unlike traditional industries that reduce output during losses, Bitcoin mining's uniqueness lies in the continuous increase in network difficulty due to computational competition. Even when prices drop, miners cannot easily exit without losing their network share. This dilemma forces mining companies into a difficult position: continue mining and incur net losses on each Bitcoin produced, or halt operations and leave previously invested electricity and mining assets idle.
Meanwhile, AI computing power's demand for electricity is exploding, making computing infrastructure a core focus of technological competition. Mining companies have already secured low-cost electricity resources, possess mature grid connectivity, and have experience managing high-density loads—advantages that allow for much shorter deployment times compared to building new data centers from scratch. Industry reports indicate that retrofitting mining facilities takes about 18 to 24 months, whereas constructing a new data center, from grid application to operation, often requires over five years.
Against this backdrop, mining companies have made a nearly unanimous choice: sell Bitcoin and transition toward providing AI computing services.
The core advantage enabling this rapid shift lies in the ready availability of infrastructure such as electricity, land, and cooling systems. Although Bitcoin mining hardware and AI servers are not interchangeable, mining facilities' power access, server room space, and cooling systems hold significant reuse value, perfectly matching the urgent needs of AI giants for computing centers.
"Bitcoin mining essentially converts electricity and chip computing power into maintaining blockchain network security. Its economic returns heavily depend on Bitcoin's price volatility, exhibiting strong pro-cyclical characteristics," said Yu Jianing, rotating chairman of the Academic Committee of the Hong Kong Certified Digital Asset Analyst Association. Currently, access to electricity is becoming a more scarce strategic resource than the chips themselves.
Yu further explained that while GPUs can be purchased and servers deployed, large-scale grid connectivity, approved power capacity, and readily available data center sites are difficult to replicate in the current global AI infrastructure expansion. In the AI era, HALO assets—characterized by heavy investment and low obsolescence—offer relative long-term stability. The heavy asset layout previously built by mining companies for Bitcoin mining is unexpectedly being revalued in this new technological cycle.
The significantly higher and more stable gross margins of AI hosting and cloud services compared to mining are also driving the transition. Data shows that AI workloads generate over three times more revenue per megawatt than traditional mining, with operating margins reaching 80% to 90%, far exceeding those of mining operations. For instance, Bit Digital Ltd.'s WhiteFiber cloud service boasts a gross margin of approximately 65%, while mining firm IREN's AI cloud service margin (after operating costs) reaches as high as 86%.
Analysts at CoinShares note that the value of Bitcoin miners transitioning to AI lies in the stable income derived from electricity resources and future computing power contracts. This type of revenue is less correlated with Bitcoin's price and is therefore more favored by public market investors.
Mining companies are collectively changing their business focus. Faced with mining losses and attractive AI profits, leading firms are acting decisively, sparking an industry-wide transformation wave. In January of this year, U.S. miner Core Scientific, Inc. sold approximately 1,900 Bitcoin, raising $175 million. In 2025, the company's mining revenue shrank from $400 million to $230 million, while its AI hosting revenue surged 168% to $65.4 million. Core Scientific has now entered a 12-year partnership with CoreWeave, with total revenue expected to reach $10.2 billion. The company recently secured a substantial $1 billion credit facility, which will be fully invested in its AI hosting business.
Well-known miner Hut 8 Mining Corp laid a solid foundation for its AI sector development by signing a $7 billion AI infrastructure agreement with tech giant Google as early as December last year. In filings with the U.S. Securities and Exchange Commission (SEC), U.S.-listed miner MARA Holdings indicated its intention to sell part of its Bitcoin holdings in 2026. Affected by falling Bitcoin prices, MARA's Q4 2025 revenue was $202.3 million, down approximately 6% year-over-year. In late February this year, MARA announced a collaboration with investment firm Starwood Capital to develop large-scale computing data centers for AI and cloud computing clients, leveraging its existing mining infrastructure.
Riot Platforms sold 1,080 Bitcoin in January, raising about $96 million specifically for acquiring the Rockdale site and developing an AI computing data center project. The company also signed a data center leasing and services agreement with AMD.
Beyond the major players, small and medium-sized mining enterprises are also embarking on differentiated transition paths. In February, Bitfarms Ltd. announced plans to rename itself and accelerate its transformation into a digital infrastructure service provider. The company had already converted $300 million in debt financing into project financing in October 2025, specifically for data center construction, and sold its PasoPe mining facility for $30 million in January.
Wang Yingbo believes that in the future, the Bitcoin mining industry will evolve into a "computing power resource management industry." Bitcoin mining will become just one business line among others, with the core competency shifting to securing low-cost energy and efficiently, flexibly converting it into standardized, marketable computing power products.
What impact will this collective transition have on Bitcoin? Yu Jianing suggests that in the short term, large-scale sell-offs by major miners will increase selling pressure on the market. However, the volume of these sales is digestible relative to Bitcoin's daily trading volume of tens of billions of dollars, and may not necessarily constitute a sustained negative factor in the medium to long term.
"Once some miners derive more income from long-term hosting contracts and infrastructure leasing, their cash flow dependence on selling Bitcoin will decrease. This could potentially weaken the cyclical supply-side shocks from miners. Subsequently, the dominant forces driving Bitcoin's price may shift more towards ETF flows, institutional allocation patterns, and the macro liquidity environment, with the marginal impact of miners' balance sheets on price tending to decline," Yu stated.
It is important to note that partnerships with tech giants do not eliminate all risks. Yu pointed out that factors such as high-leverage financing, data center retrofit timelines, client delivery constraints, GPU maintenance capabilities, and grid and environmental regulations will ultimately determine the success of this transition.
"The collective shift of Bitcoin miners to AI hosting is essentially a repricing of computing power assets," said Gao Chengyuan, Dean of the Influence Academy. The long-term power purchase agreements and compliant data centers held by mining companies恰好 match the massive demand from AI firms for stable computing power.
Gao believes that future Bitcoin mining will exhibit a "bifurcation": on one end, large-scale, intelligent computing centers integrated with AI, relying on clean energy; on the other end, distributed, modular edge mining facilities serving as grid frequency regulation tools and redundant backups for decentralized networks. Mining will not disappear but will degenerate into a functional component of the energy system rather than an independent industry.