Mining company stocks are drifting further away from cryptocurrency
Author: Zhou, ChainCatcher
According to RootData market data, in the past year, BTC has fallen by a cumulative 46.12%, but Bitcoin mining stocks have not declined in tandem. Among them, HUT has risen by 363.26%, WULF by 268.95%, IREN by 121.14%, RIOT by 59.90%, and CLSK by 12.41%.

This round of increase has not been built on improvements in mining fundamentals. June operational data shows that despite the continuous reduction in mining difficulty, the output of CleanSpark, BitFuFu, and Canaan has still decreased month-on-month by 9% to 29%.
It is not difficult to see that the focus of the market has changed. Since July, CleanSpark has signed an initial 20-year infrastructure lease worth approximately $6.6 billion, TeraWulf plans to raise $3.5 billion to expand its data center campus, and MARA has acquired a Texas project company with a planned power capacity of up to 2 GW for up to $600 million.
Mining company stock prices are no longer revolving solely around coin prices, output, and hash rate; the market has begun to value them according to a different logic.
The Source of Mining Stock Volatility is No Longer On-Chain
At the beginning of this month, the market experienced a typical dislocation, with mining stocks collectively retreating by about 20%, while BTC remained stable around $64,000.
On the output side, in June, CleanSpark produced 614 BTC, down from 671 in May, a month-on-month decline of 9%, with a nominal hash rate of 50 EH/s and an average operational hash rate of only 42.6 EH/s, widening the gap from 3.8 EH/s in May to 7.4 EH/s, indicating downtime or reduced load.
BitFuFu produced 125 BTC, a month-on-month decline of 29.4%, with total hash rate dropping from 19.5 EH/s to 15.3 EH/s, mainly due to a reduction in third-party hosted hash rate from 16.3 EH/s to 11.8 EH/s.
Canaan produced 64 BTC, a month-on-month decline of 29%, with the company attributing part of the reason to grid maintenance at the mining site.
This round of production cuts occurred after continuous reductions in difficulty. On June 14, the Bitcoin network difficulty was reduced by 10.09%, the second-largest negative adjustment in 2026, and on July 11, it was further reduced by 5% to 127.17 T, a cumulative decline of about 18% from the peak of approximately 155 T in November 2025.
The decrease in difficulty should have allowed miners remaining on the network to mine more coins per unit of hash rate, yet output continues to decline.

On the other hand, against the backdrop of a sluggish market and pressured profitability, some miners are continuously exiting the network or shutting down equipment. Galaxy Research states that miners are entering a surrender phase, marking the largest withdrawal since China's comprehensive crackdown on Bitcoin mining in 2021.

The reasons for the clearance are also straightforward. According to CoinShares' Q1 2026 mining report, the average cash production cost of listed mining companies had risen to about $79,995 in Q4 2025, while JPMorgan estimates the current production cost to be around $78,000, with BTC's current price hovering around $64,000, resulting in a price gap that has persisted for five months, with about 20% of miners in a loss-making state.
According to Hashrate Index data, around March 2026, hashprice briefly fell to a post-halving low of $28 to $30 per PH/s per day, currently around $32, still in the historically lowest range.

Classified Under AI Infrastructure Valuation System
The new logic is not complicated; AI data centers currently lack integrated power capacity, contiguous land, cooling, and building frameworks, and mining companies happen to possess these resources.
They have large-scale power connection capabilities, sites that can be transformed, established operational systems, and are more familiar with the construction rhythm of high-load facilities.
PJM data shows that AI infrastructure projects that began operations in 2025 took an average of over seven years, with about three years spent obtaining interconnection service agreements and another four years waiting for grid connection. An already connected mining site effectively skips these seven years, and the value of mining companies comes from this.
Take CleanSpark as an example; on July 14, the company announced it had signed a 20-year tripartite lease with an unnamed high-investment-grade technology company located in the Sandersville campus in Georgia, with initial contract revenue of approximately $6.6 billion corresponding to a critical IT load of 175 megawatts, starting delivery in Q4 2027. The market response was also strong, with CLSK rising by 22% during the day.
Also in July, MARA spent up to $600 million to acquire a Texas project company with a planned power capacity of up to 2 GW. However, this company holds a letter of intent signed with the power company. The gap between the letter of intent and actual power connection is precisely those seven years.
In addition, the credit market is also pricing them according to new standards. According to Bloomberg, TeraWulf plans to raise $3.5 billion led by Morgan Stanley, including leveraged loans and high-yield bonds, to expand the Justified Data campus in Hawesville, Kentucky, marking its first entry into the leveraged loan market. Lenders are also beginning to assess miners' balance sheets based on infrastructure.
According to Guosheng Securities research, as of early May 2026, the signed contracts for site hosting, bare metal, and cloud within the sector totaled approximately 3,201 megawatts of critical IT load, with a total contract value exceeding $91.4 billion. The institution also found a significant positive correlation between the market capitalization of companies in the sector and their AI power reserves and contracted AI power in North America.

CoinShares expects that by the end of 2026, up to 70% of the revenue of listed mining companies will come from AI and HPC, compared to about 30% at the beginning of the year. TeraWulf has already arrived, with its HPC leasing revenue of $21 million in Q1 surpassing its mining business revenue of less than $13 million for the first time.
The Cost of Revaluation: Three Layers of Risk
The first layer of risk comes from valuation.
Mining companies are being revalued based on AI infrastructure, which means they must endure the overall volatility of the AI narrative.
According to a report from 10x Research, Bitcoin mining stocks have largely decoupled from the price of coins, with RIOT's stock price becoming more synchronized with the Philadelphia Semiconductor ETF since April 2026.
Bitcoin mining companies are now deeply tied to the AI theme, which currently revolves more around global supply chains and competition rather than crypto adoption or financial digitization. Additionally, the performance of Chinese LLM concept stocks and the outlook for the South Korean semiconductor supply chain are directly affecting the trends of Bitcoin mining stocks.

After experiencing a surge, risk appetite is contracting. The Philadelphia Semiconductor Index fell by 10.8% over ten trading days, with Reuters estimating that the entire industry has evaporated about $1.3 trillion in market value, attributing the root cause to doubts about the return on investment in AI infrastructure, bubble-level valuations, and a more hawkish Federal Reserve.

The second layer of risk comes from return rates.
According to a Bernstein report, Core Scientific's five-year average asset return rate in collaboration with CoreWeave reached 75%, but the driving factor is the capital expenditure structure rather than transaction terms, with tenants bearing $750 million of the total cost of $855 million through revenue prepayments. Riot relies on transforming existing mining sites, achieving a return rate of 23%.
However, these two are not industry benchmarks; the report points out that the actual baseline return rates in the industry fall at TeraWulf 5%, Cipher 4%, CleanSpark 4%.
A report on July 1 stated that Meta plans to launch Meta Compute to sell surplus AI training and inference computing power to enterprise customers, causing the Philadelphia Semiconductor Index to drop by 6.3% that day. The next day, SK Hynix CEO announced that SK Group would invest 100 trillion won in South Korea to build AI data centers in phases, starting with 5 GW and eventually expanding to 15 GW.
Meta, as the largest buyer, claims to have surplus, while chip manufacturers say they will build their own, whereas mining companies are signing long-term contracts of 15 to 20 years, not already realized revenue. This is how the recent 20% retreat in mining stocks came about.
The third layer of risk comes from execution.
Mining companies are now pricing the future rather than realized revenue. For example, CleanSpark has just signed a $6.6 billion long-term contract, but its revenue currently still comes entirely from Bitcoin mining, with AI business not yet generating substantial income, and the first deliveries are not expected until Q4 2027.
Valuation has already moved ahead, but realization still needs to pass three major hurdles:
The first hurdle is financing capability. According to CleanSpark's submitted 8-K, the construction cost for the campus is $10 million to $12 million per megawatt, corresponding to a capital expenditure of $1.75 billion to $2.1 billion for 175 megawatts, which has not yet been raised. The document also states that failure to achieve any of the milestones in financing, construction, or delivery will trigger rent reductions or even lease termination.
The second hurdle is regulatory permits. On July 14, New York Governor Hochul signed an executive order suspending the issuance of state-level permits for large data centers, with a threshold of over 50 megawatts of grid demand. The New York State Environmental Protection Agency has shelved all discretionary permits not deemed complete before July 14, with the suspension period tied to the completion of a general environmental impact statement rather than a fixed date, lasting up to one year.
The third hurdle is tenant quality. Bernstein points out that tenant quality directly affects the valuation level of mining companies; ultra-large cloud providers can bring more stable cash flow and lower financing costs, while small GPU cloud service providers correspond to higher operational risks and capital costs.
Miners' Selling Logic Decoupled from Coin Prices
The valuation logic has changed, and miners' behaviors have changed accordingly. However, this change has a more direct impact on the crypto space, reflected in how miners sell coins.
According to industry reports, listed mining companies sold a total of about 32,000 BTC in Q1 2026, exceeding the total for the entire year of 2025. Among them, Riot produced 1,473 BTC in Q1 and sold 3,778 BTC during the same period, more than double its output, reducing its holdings to 15,680 BTC, an 18% year-on-year decrease.
In the past, miners sold coins primarily based on cash flow logic, selling coins to pay electricity bills, repay loans, and maintain daily operations, being reluctant to sell at low prices, waiting for rebounds to sell. Now, there is an additional layer of transformation financing logic, where selling coins also needs to create space for site repairs, land acquisition, capex, and longer-term AI construction plans.
Thus, even if coin prices do not experience extreme fluctuations, miners may continue to sell coins.
The same logic also determines whether the exiting hash rate will return.
In the past, the market assumed that hash rate exiting the network would return after coin prices rose and difficulty decreased. After China's comprehensive crackdown on mining in 2021, difficulty dropped by 46%, and it was recovered in half a year. However, what is leaving now may not just be mining machines, but also the underlying power and capital expenditures.

Currently, mainstream AI contracts are mostly long-term contracts of over 10 years. Once mining companies lock in their sites, power, and financing structures into such contracts, resources will be much less flexible to flow back into BTC mining as they did in the past.
Therefore, mining companies are moving further away from crypto; more accurately, the capital market has begun to value them based on their departure from the pure mining framework.
They will still influence the Bitcoin network and continue to earn income from mining, but their pursuit of power, land, and long-term leases is transforming them into a different type of company.












