Drastic changes in valuation logic: mining companies become AI power real estate agents

According to WooFunai, the valuation logic of listed Bitcoin mining companies is undergoing fundamental restructuring. The core indicators have completely migrated from the traditional computing power scale and BTC holdings to the installed power capacity, contracted IT load, and project delivery capacity. Land, substations, and power grid access rights that were once built specifically for mining are being redefined as scarce infrastructure assets contested by AI data centers.

The macro context of this paradigm shift is that electricity has become a scarce link between the two major industries of Bitcoin mining and artificial intelligence (AI). The International Energy Agency (IEA) indicated in an updated forecast in April 2026 that global data center electricity consumption will soar from about 485 TWh in 2025 to about 950 TWh in 2030, nearly doubling. More notably, the IEA previously predicted that US data centers would contribute nearly half of the new electricity demand from 2024 to 2030.

Meanwhile, the computing power revenue of Bitcoin mining units continues to be pressured, forcing mining companies to re-examine their asset portfolios and reconfigure the power infrastructure originally serving ASIC miners as a key resource for AI high-performance computing (HPC).

The failure of traditional mining formulas stems from a sharp narrowing of profit margins and difficulties in commercializing the business. After Bitcoin was halved in 2024, the block subsidy was reduced to 3.125 BTC, compounding the long-term increase in computing power and difficulty of the entire network, and the revenue shared by a single mining machine was continuously diluted. According to VanEck data, as of mid-July 2026, the Bitcoin Hashprice had dropped to about 30.6 US dollars/ph/s/day. The average daily income of miners across the network over the past 30 days was about US$28.5 million, and inefficient mining machines were approaching or falling below the break-even line.

This means that even without a sharp drop in currency prices, increased competition will erode profits. In contrast, once AI data centers sign long-term leases with customers with strong credit, they can transform electricity access and construction and operation capabilities into cash flow with a longer period of time and higher predictability, thus eliminating real-time dependence on currency price fluctuations and the difficulty of the entire network.

CoreScientFIC's (CORZ.US) financial performance is the most direct sample of this transformation. In the second quarter of 2026, the company's total revenue reached US$164.2 million, a year-on-year surge of 109%. Among them, high-density hosting revenue reached US$136.7 million, accounting for about 83% of total revenue, compared to only US$10.6 million in the same period last year. In stark contrast to this, revenue from proprietary mining plummeted 66% from US$62.42 million in the same period last year to US$21.54 million, and recorded a gross loss of approximately US$12.17 million, with a gross margin of -56%. On the other hand, the high-density hosting business achieved gross profit of about 79.98 million US dollars, and the gross margin was as high as 59%. The same batch of assets showed very different financial results under the two business models, marking a complete shift in the company's revenue focus from mining to AI infrastructure.

The 10 billion dollar contract between giants further validates the new pricing logic for electricity capacity. In July 2026, TeraWulf (WULF.US) signed a 20-year data center lease with Anthropic, covering its JustifiedData campus in Hawesville, Kentucky, and plans to provide about 401 MW of critical IT capacity. Delivery is expected to begin in the second half of 2027 and full operation in early 2028. TeraWolf said the contract is expected to generate approximately $19 billion in revenue and receive investment-grade credit support during the initial term. In the same month, Hut8 (HUT.US) announced the signing of a second 15-year lease worth 9.8 billion US dollars at the BeaconPoint campus in Texas, adding 352 MWIT capacity, increasing the contract size of the same customer in the park to 704 MW. So far, the value of BeaconPoint's basic term contract has reached US$19.6 billion, and delivery of the first phase II data office is expected to begin in the second quarter of 2028.

With the same power capacity, service AI can get a higher premium. The underlying reason is the fundamental difference in risk structure and capacity requirements. Bitcoin mining can withstand high interruption rates, and mining companies can actively shut down or migrate equipment when electricity prices rise; while AI training and inference loads require extremely high power supply stability, network bandwidth, cooling capacity, and system redundancy. Therefore, what actually earns a premium is not “electricity” itself, but a combination of four capabilities: power capacity that has been electrified or has clear grid connection arrangements, engineering capacity to complete construction on schedule, financing ability to withstand huge upfront investment, and customer credit with long-term leases. Long-term leases transform mining companies from volatile income similar to commodity production to stable cash flows close to data center developers or infrastructure asset operators. A lease supported by an investment-grade hyperscale cloud service provider can help mining companies finance at a lower cost. Similarly, if the 100MW capacity has poor credit, the valuation will be significantly different.

Wall Street has begun pricing according to MW rather than BTC. Vaneck used GrossEnergizedPower as the main comparison standard in the AI infrastructure valuation framework for mining companies released in June 2026. Based on data as of June 4, 2026, companies that have signed AI or HPC leases generally have valuation multiples of more than 10 times; while companies that have almost no contracted capacity and only show forward power reserves, the valuation multiples are only 2 to 6 times.

The multiples here aren't traditional price-earnings ratio, EV/revenue, or EV/Ebitda. The market is strictly distinguishing four capacity states: the planned capacity is in the development phase; the locked power capacity has been agreed upon but is not electrified; the energized capacity has the conditions for power supply; and the delivered billing capacity has generated revenue. As more projects run, companies such as CoreScientFIC, TeraWolf, Hut8 (, Cipher (CIFR.US)) are getting closer to data center development and operation platforms; while MARA, CleanSpark (CLSK.US), etc. still maintain high Bitcoin mining exposure, the mining stock sector is facing logical differentiation.

However, the AI transformation has also brought about execution thresholds far higher than traditional mining and huge funding gaps. According to data compiled by WooFunai, based on VanEck's estimates as of June 4, 2026, the relevant companies only delivered about 25% of the leased capacity at the time, and the capital gap between recent capital expenditure requirements and existing cash is about 50 billion US dollars.

Furthermore, the long-term capital expenditure requirements of related companies are close to US$221 billion. GW and the total contract amount that appeared in numerous press releases still correspond to construction plans for 2027 or even after 2028. Projects may be affected by grid upgrades, equipment delivery, construction costs, financing terms, regulatory permits, and community resistance. For mining companies that lack experience in building high-density data centers, any delays or overspending could hurt cash flow. There is also a risk of customer concentration. A single tenant cuts capital expenses or next-generation chip design changes, which may make it difficult to convert heavy assets into use.

Furthermore, raising capital through the issuance of shares, convertible bonds, project loans, and customer advances may result in equity dilution, higher leverage, and complex financing constraints.

The end of the industry points to the evolution of mining companies to power infrastructure companies. Bitcoin mining has not lost value; it can still quickly monetize electricity and provide highly flexible returns when currency prices rise. Unlike traditional data centers, mines can actively reduce loads, participate in grid demand response, and provide transitional revenue for uncontracted capacity. However, for a number of listed mining companies, Bitcoin is changing from being the only main business to a way to monetize electricity infrastructure. What is really difficult to replicate is the acquired land, grid connection qualifications, transmission facilities, and large-scale power arrangements. ' The most valuable thing is that electricity 'requires additional restrictions: it is not the planned GW, but electricity that can be connected to the grid on schedule, complete financing, build a high-density data center, and be leased by reliable customers for a long time. Mining companies are moving from finding cheap electricity to produce BTC to monetizing the time value of scarce grid-connected resources and infrastructure through leases.

As a result, Bitcoin mining entered a new fork. Some companies will continue to bet on the currency price and computing power cycle to maintain the traditional mining model; others may completely remove the “mining enterprise” label and become a new type of power real estate company in the AI era. This shift is not only an adjustment in business focus, but also a fundamental reshaping of asset attributes and valuation systems. As the 2027-2028 delivery period approaches, the market will test whether these companies can turn their power blueprint into actual revenue to determine whether they ultimately become highly leveraged losers or establish themselves as next-generation infrastructure giants. Following the rise of Internet infrastructure, this is another major paradigm shift in the field of energy and computing integration.