Digging Deeper into Losses! Bitcoin Mining Industry Undergoes Major Reshuffling: Miners Transform to AI Without Waiting for Bull Market, Electricity Becomes "Lifesaving" Chip

Hot Topics

Selected Stocks Data Center Market Center Capital Flows Simulated Trading

Client

Source: Huaxia Times

The identity of miners is shifting from Bitcoin (BTC) “workers” to AI “power landlords.”

Recently, as Bitcoin prices have fallen back, Bitcoin mining has begun to become unprofitable. Leading mining companies like Core Scientific and MARA have been liquidating their Bitcoin holdings one after another. As of now, listed mining firms have sold off over 15,000 BTC. Meanwhile, these companies are transforming mining farms into AI data centers, turning mining into a side business. A major industry migration across the crypto and AI sectors has officially begun.

“Shifting power from mining to AI hosting services signals the arrival of the ‘computing power-based’ era,” said Wang Yingbo, a digital economy scholar at Shanghai Academy of Social Sciences, to Huaxia Times. From a ‘computing power’ perspective, Bitcoin is just an early, but fundamentally flawed, token created by the surge in computing power during a specific historical period. The liquidation of BTC by top miners and their move into AI can be seen as a decoupling from the old token system and an investment in new infrastructure based on computing power. This is not just a bear market hedge but a historic industry shift.

Mining Becomes a Loss-Making Business

After the 2024 Bitcoin halving, mining profit margins dropped by about 50%, but costs did not decrease. From the second half of 2025, the continuous decline in coin prices has completely broken miners’ survival threshold. Currently, the total network mining cost of Bitcoin is severely inverted with the coin price. Cash costs, hardware depreciation, operational expenses, land, and other costs make mining a “money-consuming beast” rather than a “money-printing machine.”

Industry estimates suggest that the cost to mine one Bitcoin is about $87,000. At the time of writing, Bitcoin’s latest price is $70,343.9, meaning miners lose approximately $17,000 for each Bitcoin mined.

Generally, when Bitcoin prices fall below miners’ costs, they have two options: shut down to reduce losses or sell Bitcoin to sustain operations. Unlike traditional industries that cut production when losing money, Bitcoin mining has a unique challenge: the ongoing competition for computing power increases network difficulty. Even if prices fall, miners find it hard to exit easily, or they risk losing network share. This creates a dilemma: continue mining at a loss or cease operations and leave invested electricity, mining farms, and equipment idle.

Meanwhile, AI computing power demand is exploding, making infrastructure a core element of technological competition. Miners have already secured low-cost electricity resources, with mature grid access and high-density load operation experience. Their deployment cycle is much shorter than building new data centers. Industry reports indicate that converting a mining farm takes about 18 to 24 months, whereas new data centers often require over five years from grid connection application to operation.

Against this backdrop, most miners are making a nearly unanimous choice: sell Bitcoin and shift to AI computing services.

The key advantage enabling rapid transformation is existing infrastructure—power, land, cooling systems. Although Bitcoin mining hardware and AI servers are not interchangeable, the power supply, server space, and cooling systems of mining farms have high reuse value, matching the urgent needs of AI giants for computing centers.

“Bitcoin mining essentially involves converting electricity and chip computing power into maintaining blockchain network security. Its economic returns are highly dependent on Bitcoin prices and exhibit significant pro-cyclicality,” said Yu Jianing, Chair of the Hong Kong Digital Asset Analysts Association (HKCDAA), to Huaxia Times. Currently, power access is becoming a more scarce strategic resource than the chips themselves.

Yu further explained that GPUs can be purchased, servers deployed, but large-scale grid connection capacity, approved power capacity, and ready data center sites are difficult to replicate in the current global AI infrastructure expansion. In the AI era, halo assets (assets with heavy capital investment and low obsolescence) have relatively long-term stability. The heavy asset layout formed by miners during Bitcoin’s boom has unexpectedly gained revaluation opportunities in the new technological cycle.

Higher stable gross margins of AI hosting and cloud services compared to mining are also driving miners’ transformation. Data shows that revenue per megawatt of AI workloads is over three times that of traditional mining, with operating profit margins reaching 80-90%, far exceeding mining. WhiteFiber, a subsidiary of Bit Digital, has a cloud service gross margin of about 65%; IREN’s AI cloud service (after operational costs) reaches 86%.

CoinShares analysts note that the value of Bitcoin miners shifting to AI lies in stable income from power resources and future computing power contracts, which are less correlated with Bitcoin prices and thus more attractive to public market investors.

Mining Companies Are “Changing Careers” Collectively

Amid losses in mining and lucrative AI profits, leading companies are acting decisively, sparking a wave of industry transformation. In January this year, US miner Core Scientific sold about 1,900 BTC, cashing out $175 million. Although its mining revenue in 2025 is expected to shrink from $400 million to $230 million, AI hosting revenue has surged by 168% to $65.4 million. Core Scientific has also entered a 12-year partnership with CoreWeave, with total revenue reaching $10.2 billion. Recently, the company secured a $1 billion credit line, which will be fully invested in AI hosting.

Famous miner Hut 8 signed a $7 billion AI infrastructure agreement with tech giant Google last December, laying a solid foundation for its AI development. US-listed miner MARA disclosed in SEC filings plans to sell some of its Bitcoin holdings in 2026. Due to Bitcoin price declines, MARA’s revenue in Q4 2025 was $202.3 million, down about 6% year-over-year. In late February, MARA announced a partnership with investment firm Starwood Capital to build large-scale computing centers for AI and cloud computing clients, leveraging existing mining infrastructure.

Riot Platforms (RIOT) sold 1,080 BTC in January, raising about $96 million, to fund the acquisition of Rockdale land and develop AI computing centers. The company also signed data center leasing and service agreements with AMD.

Beyond top-tier companies, small and medium miners are also pursuing differentiated transformation. In February, Bitfarms (BITF) announced plans to rebrand and accelerate as a digital infrastructure service provider. Previously, the company converted $300 million of debt financing into project funding for data center construction, and in January sold the PasoPe mining farm for $30 million.

Wang Yingbo believes that in the future, the Bitcoin mining industry will evolve into a “computing resource management industry,” with Bitcoin mining being just one of its business lines. The core capability will shift to acquiring low-cost energy and efficiently converting it into standardized, sellable computing power.

What impact will the collective transformation of miners have on Bitcoin? Yu Jianing said that in the short term, the liquidation of holdings by top miners will increase market selling pressure, but the scale of their sales relative to Bitcoin’s daily trading volume of billions of dollars is manageable. In the medium to long term, it may not be a sustained negative factor.

“If some miners generate more revenue from long-term hosting contracts and infrastructure leasing, their cash flow dependence on selling coins will decrease. The cyclical shocks from supply side may weaken. The dominant factors influencing Bitcoin prices could shift more toward ETF flows, institutional allocations, and macro liquidity conditions, with miners’ asset-balance impacts diminishing,” Yu explained.

It is worth noting that cooperation with industry giants does not mean all risks are eliminated. Yu pointed out that high leverage financing, data center transformation cycles, customer delivery constraints, GPU operations, grid and environmental regulations will all influence the final outcome of this transition.

“Bitcoin miners’ collective shift to AI hosting is essentially a re-pricing of computing power assets,” said Gao Chengyuan, director of Influence Academy, to Huaxia Times. “Long-term power purchase agreements and compliant data centers match well with AI companies’ huge demand for stable computing resources.”

Gao believes that future Bitcoin mining will show a “polarized” trend: on one end, large-scale, intelligent, clean energy-powered computing centers integrated with AI; on the other, distributed, modular edge farms serving as grid frequency regulation tools and decentralized network backups. Mining will not disappear but will evolve into a functional component of energy systems rather than an independent industry.

Editor: Xu Yunqian Chief Editor: Gong Peijia