Wall Street re-prices mining companies: Bitcoin miners are becoming power real estate developers in the AI era

Editor | Wu Says Blockchain

TL;DR

· A group of publicly listed Bitcoin mining companies shifting towards AI and high-performance computing (HPC) are changing their valuation logic from hash rate, Bitcoin output, and BTC holdings to powered capacity, contracted IT load, and project delivery capability. Land, substations, and grid access previously prepared for mining are becoming scarce assets contested by AI data centers.

· Core Scientific's high-density hosting revenue reached $136.7 million in Q2 2026, accounting for about 83% of total revenue; self-mining revenue, on the other hand, fell approximately 66% year-over-year to $21.54 million, recording a gross loss of about $12.17 million, indicating a shift in revenue focus from mining to AI infrastructure.

· TeraWulf signed a data center lease with Anthropic covering approximately 401 MW of critical IT load, expected to be worth $19 billion; Hut 8 also signed a new lease for 352 MW, valued at $9.8 billion. However, most of the related capacity will not be delivered until 2027 to 2028.

· "The most valuable thing is electricity" does not refer to planned GW, but rather to infrastructure that can be powered on time, complete financing, build high-density data centers, and be leased long-term by reliable customers. VanEck estimated that as of June 4, 2026, related companies had delivered only about 25% of the leased capacity, with a recent funding gap of about $50 billion.

In the past, the market assessed a Bitcoin mining company primarily based on three indicators: how large its hash rate is, how many Bitcoins it can mine daily, and how many Bitcoins it holds on its balance sheet. Today, for a group of publicly listed mining companies transitioning to AI and HPC, this formula is becoming ineffective.

The new valuation units are no longer just EH/s and BTC, but MW and GW. Investors are beginning to ask: how much powered or clearly grid-connected electrical capacity does a mining company control, how much has been contracted to AI clients, and how much has been delivered and started billing.

Behind this change are two entirely different industry situations. The unit revenue from Bitcoin mining continues to be under pressure, while the demand for electricity from AI data centers is still rapidly increasing. The International Energy Agency (IEA) updated its forecast in April 2026, stating that global data center electricity consumption will rise from about 485 TWh in 2025 to about 950 TWh in 2030, nearly doubling. The IEA previously estimated that U.S. data centers would contribute nearly half of the new electricity demand from 2024 to 2030.

Electricity has thus become a scarce asset connecting the two industries. Land, substations, transmission access, and long-term power purchase agreements previously prepared for Bitcoin mining are being reconfigured as AI infrastructure. Mining companies are no longer just producing Bitcoin through computation; they are also monetizing electrical capacity and corresponding construction and operational capabilities through long-term data center leases.

The Old Formula of Mining is Becoming Ineffective

The shift of mining companies towards AI is not primarily because the AI narrative is more compelling, but because the profit margins of traditional mining operations are narrowing.

After the Bitcoin halving in 2024, the block subsidy will drop to 3.125 BTC. Meanwhile, the overall network hash rate and mining difficulty have been rising for a long time, continuously diluting the revenue that a single mining machine can earn. VanEck data shows that as of mid-July 2026, the Bitcoin Hashprice had fallen to about $30.6 per PH/s/day, with the average daily revenue for miners across the network being about $28.5 million, and inefficient mining machines are hovering around or below the breakeven point.

For mining companies, this means that even if the price of Bitcoin does not continue to fall sharply, they may earn less due to increased competition. Mining machines must be continuously updated, electricity prices must be low enough, and any downtime, difficulty adjustments, or rising financing costs can quickly erode profits.

More importantly, Bitcoin mining is still fundamentally a commoditized business. There is no significant difference in the hash rate produced by the same generation of mining machines, making it difficult for mining companies to achieve long-term premiums based solely on brand; revenues fluctuate in real-time with coin prices, transaction fees, and overall network difficulty. In contrast, once an AI data center signs a long-term lease with a creditworthy client, it has the opportunity to convert electricity access, data center construction, and operational capabilities into longer-term, more predictable cash flows.

This is precisely the starting point for mining companies to reassess their assets: the most valuable may not be the ASIC mining machines in the data center, nor the unsold Bitcoins, but the electricity access rights already obtained behind the mining sites.

Core Scientific's Financial Report Has Changed Its Face

Core Scientific is the most direct example of this change.

In Q2 2026, the company's total revenue was $164.2 million, a year-over-year increase of about 109%. Among this, high-density hosting business revenue reached $136.7 million, accounting for about 83% of total revenue, compared to only $10.6 million in the same period last year.

In contrast, self-mining revenue fell from $62.42 million in the same period last year to $21.54 million, a year-over-year decline of about 66%. The cost of this business for the quarter was $33.7 million, corresponding to a gross loss of about $12.17 million, with a gross margin of -56%; the high-density hosting business achieved a gross profit of about $79.98 million, with a gross margin of 59%.

The same set of data center assets presents completely different financial results under the two business models. As of mid-July 2026, Core Scientific had 437 MW of customer electrical capacity beginning to bill, corresponding to an annualized GAAP revenue of about $635 million; its contracted customer electrical capacity was about 1.1 GW, with potential contract revenue exceeding $24 billion.

This does not mean that the transformation is complete. The company's capital expenditures in the second quarter reached $797.5 million, far exceeding revenue for the same period, and a considerable portion of the contracted capacity has yet to start billing. However, it at least indicates that the financial focus of mining companies can change in a very short time: a decline in Bitcoin output no longer necessarily means a simultaneous decline in company revenue, as long as the capacity originally used for mining can be converted into billable customer electrical capacity.

A Data Center Lease Can Be Worth Billions

If Core Scientific demonstrates a change in revenue structure, then TeraWulf and Hut 8 show how electrical capacity and data center construction and operational capabilities can be repriced through long-term contracts.

In July 2026, TeraWulf signed a 20-year data center lease with Anthropic. The agreement covers its Justified Data campus in Hawesville, Kentucky, planning to provide about 401 MW of critical IT capacity for AI loads, expected to start delivery in the second half of 2027 and be fully operational by early 2028. TeraWulf stated that this contract is expected to bring about $19 billion in contract revenue during the initial term and is anticipated to receive investment-grade credit support.

In the same month, Hut 8 announced a second 15-year lease worth $9.8 billion at the Beacon Point campus in Texas, adding 352 MW of IT capacity, bringing the total contracted scale for the same client at this campus to 704 MW. Thus, the base term contract value for Beacon Point reached $19.6 billion, while the first second-phase data hall is not expected to start delivery until the second quarter of 2028.

These transactions reflect a new industry reality: in the eyes of AI clients, the hardest thing to obtain may not be GPUs, but large-scale electricity that can be brought online within a specified timeframe. Chips can be procured, servers can be deployed, but transmission lines, substations, land permits, and grid access often take years to secure.

Bitcoin mining companies have already completed part of this work in advance. In pursuit of low-cost electricity, they have spent the past decade seeking locations close to power generation, with large load capacities and allowing for rapid construction. Previously, these assets were used to keep tens of thousands of ASIC mining machines running continuously; when AI data centers face electricity bottlenecks, the same site selection and energy capabilities gain new monetization channels.

Why the Same Electrical Capacity Can Command a Higher Premium for AI Services

The essence of the transformation of mining companies is not simply unplugging mining machines and plugging in GPUs.

Bitcoin mining can tolerate a higher interruption rate. Mining companies can proactively shut down during rising electricity prices or grid stress, and they can move equipment to another mining site. AI training and inference loads, however, require higher power stability, network bandwidth, cooling capacity, and system redundancy; high-density GPU cabinets impose far greater requirements on data center design than traditional mining sites.

The premium here does not come from reselling the same kilowatt-hour at a markup, but from converting electrical capacity into highly reliable critical IT loads. Therefore, the true premium is not "having electricity" itself, but rather a combination of four capabilities: electrical capacity that is already powered or has clear grid connection arrangements, construction capabilities that can complete projects on schedule, financing capabilities that can bear significant upfront investments, and customer credit that can secure long-term leases.

Long-term leases also change the risk structure for mining companies. Mining revenue depends on Bitcoin prices, overall network difficulty, and transaction fees, which fluctuate almost daily; AI hosting revenue, on the other hand, depends more on contract terms, delivery progress, and tenant performance capabilities. The former resembles commodity production, while the latter gradually approaches the role of data center developers or infrastructure asset operators.

This also explains why the market is beginning to care about "who is renting these data center capacities." A long-term lease supported by investment-grade hyperscale cloud service providers can help mining companies secure project financing at lower costs; similarly, for the same 100 MW capacity, if the tenant is smaller, has a shorter lease term, or weaker credit, its valuation will differ significantly.

Wall Street is Starting to Price in MW, Not BTC

In June 2026, VanEck released a valuation framework for mining companies' AI infrastructure, using Gross Energized Power as the primary metric for comparing the enterprise value of current mining companies. Based on data as of June 4, 2026, it pointed out that companies that have signed AI or HPC leases generally have valuation multiples exceeding 10 times; those with almost no contracted capacity, still primarily showcasing future power reserves, are around 2 to 6 times. These multiples are not based on price-to-earnings ratios, EV/Revenue, or EV/EBITDA.

This means the market is distinguishing between different states of capacity: planned capacity is still in project reserve or development conceptual stages; locked-in electrical capacity has secured power agreements or grid connection arrangements but may not yet be powered; powered capacity has actual supply conditions; and delivered billing capacity has been handed over to clients and started generating revenue. The four cannot be equivalently calculated.

As more projects come online, valuation standards will continue to evolve. The market first rewards mining companies with power resources, then those that sign contracts, and ultimately returns to more traditional questions: whether projects are delivered on time and within budget, how much cash flow each MW can generate, and whether capital returns can cover financing costs.

As a result, Bitcoin mining companies are beginning to differentiate. Companies like Core Scientific, TeraWulf, Hut 8, and Cipher are increasingly resembling data center development and operation platforms; others like MARA and CleanSpark still maintain high exposure to Bitcoin mining. In the future, so-called "mining stocks" may no longer be a sector that can be valued using the same logic.

The Biggest Risk is Treating Power Blueprints as Realized Revenue

The AI transformation has raised the valuation ceiling for mining companies but also brought execution thresholds far higher than traditional mining.

VanEck estimated that as of June 4, 2026, related companies had delivered only about 25% of the leased capacity, with a recent funding gap of about $50 billion between capital expenditure needs and existing cash. This estimate does not account for future operating cash flows or funds that may be obtained from selling or mortgaging BTC. VanEck also estimated that the long-term capital expenditure needs of related companies are close to $221 billion but did not define this as a "long-term funding gap." The GW and total contract amounts frequently mentioned in press releases still correspond to construction plans beyond 2027 or even 2028.

This means that a contract worth billions of dollars cannot be directly equated to current revenue of the same scale. Projects may still be affected by grid upgrades, equipment delivery, construction costs, financing conditions, regulatory permits, and community resistance. For mining companies lacking experience in high-density data center construction, any delays or cost overruns could simultaneously impact cash flow and market valuation.

Customer concentration is another risk. Long-term leases can enhance revenue visibility but may allow a single client to determine the future of a campus or even a company. If AI infrastructure investments cool, tenants cut capital expenditures, or new generations of chips change data center designs, mining companies may face the challenge of heavy assets built for specific clients that are difficult to repurpose quickly.

Additionally, the transformation itself requires massive capital. Mining companies can raise funds through issuing stocks, convertible bonds, project loans, and customer prepayments, but these methods may also lead to equity dilution, higher leverage, and complex financing constraints. Controlling power resources is just the entry ticket; whether they can build resources into billable assets at reasonable costs ultimately determines the returns.

Mining Companies are Becoming Power Infrastructure Companies

Bitcoin mining has not lost its value. It can still quickly monetize electricity and provide highly elastic returns when coin prices rise. Unlike traditional data centers, mining sites can also proactively reduce loads and participate in grid demand response, providing transitional income for uncontracted electrical capacity.

However, for a group of publicly listed mining companies, Bitcoin is transitioning from the sole core business to a monetization method for power infrastructure. Mining machines can be replaced, mined Bitcoins can be sold, but what is truly difficult to replicate are the land, grid connection qualifications, transmission facilities, and large-scale power arrangements that have already been secured.

Therefore, "the most valuable thing is electricity" still requires a qualification: it is not the GW written in the plans, nor simply low-cost electricity, but the electricity that can be connected to the grid on time, complete financing, build high-density data centers, and be leased long-term by creditworthy customers.

Mining companies previously sought cheap electricity to produce more Bitcoins; now, they are beginning to monetize scarce grid resources and the time value of infrastructure through data center leases. Bitcoin mining thus enters a new fork: some companies will continue to bet on coin prices and hash rate cycles, while others may completely shed the "mining company" label and become new types of power real estate developers in the AI era.