The more you mine, the more you lose! Major reshuffle in Bitcoin mining: miners are not waiting for the bull market to shift to AI, with electricity becoming the "lifeline" chip.

Ask AI · How can power resources become the core bargaining chip for AI competition among mining companies?

This paper (chinatimes.net.cn) reporter Zhao Yi, Shanghai reports

Miners’ identities are shifting from Bitcoin (BTC) “working people” to AI “power landlords.”

Recently, with the pullback in Bitcoin prices, Bitcoin mining has fallen into an unprofitable situation. Top mining companies such as Core Scientific and MARA have successively liquidated their holdings of Bitcoin. As of now, listed mining companies have cumulatively sold more than 15,000 BTC. At the same time, mining companies have begun converting mining sites into AI data centers—pushing mining down to a side business. A full-fledged industrial migration spanning both the crypto and AI sectors is now underway.

“Mining companies shifting electricity from mining to AI hosting is the most direct industry signal on the eve of the arrival of the ‘compute-power-first’ era,” Wang Yingbo, a scholar of digital economics at the Shanghai Academy of Social Sciences, told The China Times. In a “compute-power-first” perspective, Bitcoin is merely a Token generated by compute power during a specific early historical stage—yet one that has fundamental flaws. The top mining companies selling off BTC and moving toward AI can be seen as “de-coupling” from the old Token system and instead investing in infrastructure for the new base value (compute power). This is far from risk-avoidance in a bear market; it is a historic shift in tracks.

Mining becomes a money-losing business

After the Bitcoin halving in 2024, mining profit margins fell by about 50%, but costs did not decrease. Then, starting in the second half of 2025, the sustained decline in coin prices completely pierced miners’ survival bottom line. As of now, the cost of mining Bitcoin across the entire network is seriously out of sync with the coin price. With cash costs layered with hardware depreciation, operations and maintenance, land, and other expenses, mining has changed from a “money-printing machine” into a “cash-eating beast.”

According to industry estimates, the cost of mining one Bitcoin is currently about $87,000. As of the time of this report, Bitcoin’s latest price is $70,343.9—meaning that for every Bitcoin a miner mines, they lose about $17,000.

Typically, when Bitcoin prices fall below miners’ cost basis. Miners usually have two options: shut down to reduce losses, or sell Bitcoin to keep operations running. But unlike loss-driven production cuts in traditional industries, the special nature of Bitcoin mining is that compute power competition continues to raise network difficulty across the entire network. Even if the coin price drops, miners cannot easily exit without losing network share. This model leaves mining companies stuck in a dilemma: if they keep mining, every coin produced is a net loss; if they stop mining, the electricity, mining-site assets invested earlier will be completely idle.

Meanwhile, AI compute power is driving explosive growth in demand for electricity. Compute-power infrastructure has become the core of technological competition. Mining companies already have low-cost power resources and mature grid interconnection capabilities, along with experience in high-density load operations and maintenance. Their deployment timelines are far shorter than those of new data centers. Industry reports show that the construction timeline for converting mining sites is about 18 to 24 months, while building new data centers—from applying for grid connection to going into operation—often takes more than five years.

Against this backdrop, mining companies made almost identical choices: sell Bitcoin and transition to AI compute-power services.

The reason mining companies can pivot quickly lies in the existing infrastructure for electricity, land, cooling, and more. While Bitcoin mining rigs and AI server hardware cannot be used interchangeably, the mining sites’ power interconnection, data room space, and cooling systems have extremely high reuse value. They align perfectly with the urgent demand from AI giants for compute-power centers.

“Bitcoin mining is essentially converting electricity and chip compute power into maintaining the security of a blockchain network. Its economic returns depend highly on coin price fluctuations and show a clear pro-cyclical characteristic,” said Yu Jialing, rotating chair of the academic committee at the Hong Kong Registered Digital Asset Analysts Association (HKCDAA), in an interview with The China Times. “At present, power entry is becoming a strategic resource that is even scarcer than the chip itself.”

Yu Jialing further explained: “GPUs can be purchased, and servers can be deployed. But large-scale grid interconnection capability, approved electricity capacity, and data-center sites that can be delivered quickly are difficult to replicate in the current global expansion of AI infrastructure. In the AI era, HALO assets—an asset category characterized by heavy assets and low obsolescence—have relatively long-term stability. The heavy-asset layout that mining companies formed in the past due to mining has unexpectedly gained an opportunity for revaluation in the new technology cycle.”

Higher stable gross margins of AI hosting and cloud services—well above those of mining—have also become a key driver of the mining companies’ transition. Data show that revenue per megawatt for AI workloads is more than three times that of traditional mining, and operating profit margins can reach 80% to 90%, far exceeding the mining business. Bit Digital’s WhiteFiber cloud service gross margin is about 65%; and IREN’s AI cloud service gross margin (after deducting operating costs) is as high as 86%.

CoinShares analysts pointed out that the value of Bitcoin mining companies shifting to AI lies in the stable income brought by power resources and future compute-power contracts. Such income is less correlated with the price of Bitcoin, so it is more favored by public-market investors.

Mining companies are collectively “switching lines”

With mining stuck in losses and AI profits looking attractive, top mining companies moved decisively and triggered a wave of industry transformation. In January this year, the U.S. mining company Core Scientific sold about 1,900 BTC in one go, cashing out $175 million. In 2025, its mining revenue shrank from $400 million to $230 million, but its AI hosting revenue surged 168% to $65.4 million. Currently, Core Scientific has already reached a 12-year cooperation agreement with CoreWeave, with total revenue as high as $10.2 billion. The company also recently successfully obtained a credit facility of up to $1 billion, which will be fully投入 into the AI hosting business.

Well-known mining company Hut 8 signed—back in December of last year—an AI infrastructure agreement worth up to $7 billion with tech giant Google, laying a solid foundation for its development in the AI space. U.S.-listed mining company MARA disclosed in a filing submitted to the U.S. Securities and Exchange Commission (SEC) that it intends to sell some of its held Bitcoin in 2026. Affected by the decline in Bitcoin’s price, MARA’s revenue in the fourth quarter of 2025 was $202.3 million, down about 6% year over year. At the end of February this year, MARA announced that it had reached a cooperation with investment firm Starwood Capital. Relying on existing mining-site infrastructure, it will build large-scale compute-power data centers for artificial intelligence and cloud computing customers.

RiotPlatforms (RIOT) sold 1,080 BTC in January this year, raising about $96 million, earmarked for the acquisition of the Rockdale tract and the development of an AI compute-power data center project. At the same time, the company signed a data center leasing and services agreement with AMD, providing services to it.

In addition to top-tier companies, smaller and mid-sized mining companies are also launching differentiated transformation paths. In February this year, Bitfarms (BITF) announced a plan to change its name and accelerate its transition to becoming a digital infrastructure services provider. Previously, the company had in October 2025 converted $300 million in debt financing into project financing, earmarked for data center construction. And in January this year, it sold the PasoPe mining site for $30 million.

Wang Yingbo believes that in the future, the Bitcoin mining industry will evolve into a “compute-power resource management industry.” Bitcoin mining will be only one of its business lines; its core capability will become obtaining low-cost energy and transforming it efficiently and flexibly into standardized compute-power products that can be sold.

What impact will the collective transformation of mining companies have on Bitcoin? In response, Yu Jialing said: In terms of the impact on Bitcoin’s supply-demand and price, in the short term, top mining companies liquidating their holdings will increase selling pressure on the market. However, relative to Bitcoin’s daily trading depth of tens of billions of dollars, the scale of their sell-off is digestible, and in the medium to long term it may not constitute a sustained bearish factor.

“Once some mining companies’ revenue comes more from long-term hosting contracts and infrastructure leasing, their dependence on selling coins for cash flow will decrease, and the periodic shock on the miners’ supply side may actually weaken. After that, Bitcoin’s price may be driven more by ETF capital flows, institutional allocation timing, and the macro liquidity environment, and the marginal impact of miners’ assets and liabilities on price will tend to decline,” Yu Jialing said.

It is worth noting that partnering with giants does not mean that risk has already cleared. Yu Jialing said that highly leveraged financing, the data center conversion cycle, customer delivery constraints, GPU operations and maintenance capabilities, and grid and environmental protection regulatory requirements will all determine the final outcome of this transformation.

“The collective shift of Bitcoin mining companies toward AI hosting is, in essence, the revaluation of compute-power assets,” said Gao Chengyuan, dean of the Influence Academy research institute, in an interview with The China Times. “The long-term power purchase agreements and compliant data centers held by mining companies perfectly match AI companies’ huge demand for stable compute power.”

Gao Chengyuan believes that in the future, Bitcoin mining will show “polarization.” On one side are large-scale, intelligent compute-power centers built on clean energy, deeply integrated with AI compute power. On the other side are distributed, modular edge mining sites, serving as redundant backup for grid frequency regulation tools and decentralized networks. Mining will not disappear, but will degrade into a functional component of the energy system rather than an independent industry.

Editor: Xu Yunci Chief Editor: Gong Peijia