TechFlow Logo
Login/ Sign up
ETH Gas
Gwei
Fear
gas
The Real Report Card of Bitcoin Miners Pivoting to AI: Revenue Decline, Widening Losses, But Power Assets Are Being Revalued

The Real Report Card of Bitcoin Miners Pivoting to AI: Revenue Decline, Widening Losses, But Power Assets Are Being Revalued

2026.08.07
Share
For investors, they only need to care about one thing: "Which company can truly convert electricity into AI revenue before running out of cash flow."

By: Xiao Bing

The mining companies' earnings season has just concluded. MARA's quarterly revenue fell 27% year-over-year, CleanSpark dropped 30.5%, and both recorded significant net losses simultaneously. Meanwhile, both keep talking about "digital infrastructure platforms" and continue to double down on AI/HPC.

The narrative has been told for nearly two years, and the market is starting to develop antibodies.

MARA: 4.8 GW Power Pipeline, Near-Zero AI Revenue

MARA released its Q2 earnings report on August 6. Revenue was $174.9 million, down 26.7% year-over-year, below analysts' expectations of $209.4 million, a miss of 16.1%. Net loss was $611.3 million, compared to a profit of $808.2 million in the same period last year. Loss per share was $1.60, while Wall Street expected a profit of $0.35.

The main cause of the loss was BTC fair value write-downs. During the quarter, 2,422 BTC were mined, 2,213 were sold at an average price of approximately $73,078, and approximately 35,577 were held. Total assets shrank from $7.3 billion at the end of 2025 to $4.3 billion, with digital assets dropping from $4.7 billion to $2.1 billion. Operationally, activated hashrate reached 70.3 EH/s, up 22% year-over-year, and network share rose to 5.9%.

In terms of AI/HPC transformation, MARA is pursuing three directions simultaneously: co-building data centers with Starwood Capital (near-term target 1 GW, long-term 2.5 GW, currently in permitting and tenant discussion stages, no signed customers yet); acquiring French HPC operator Exaion (completed, but full-year expected revenue is only "low eight figures," i.e., several million to ten or twenty million dollars); acquiring Long Ridge Energy & Power (enterprise value $1.5 billion, including a 505 MW gas power plant, annualized EBITDA $144 million, initial AI load target 200 MW, planned delivery mid-2028, acquisition awaiting approval).

Combined, these three paths give MARA a so-called 4.8 GW power pipeline. However, to date, AI/HPC contribution to revenue is near zero. Management was candid on the conference call: the first half was about "expansion and transformation," while the second half focuses on "signing customers, launching assets, and demonstrating profitability."

No money earned yet, just the framework built.

CleanSpark: 100% Mining Revenue, but a $6.6 Billion Lease Changes Valuation Logic

CleanSpark released its fiscal 2026 Q3 earnings report on the same day (ended June 30). Revenue was $138 million, down 30.5% year-over-year. Net loss was $239.8 million, compared to a profit of $257.4 million in the same period last year. Adjusted EBITDA turned from positive $377.7 million to -$113 million.

A key fact: 100% of this quarter's revenue comes from Bitcoin mining, zero AI/HPC revenue.

But on July 14, CleanSpark signed a contract that could rewrite the company's valuation: a 20-year triple net lease with a confidential "high investment-grade global tech company" to deploy 175 MW of critical IT load at the Sandersville campus in Georgia, expected to start delivery in Q4 2027. Total contract revenue is approximately $6.6 billion, reaching up to $11.6 billion with extension options. Management stated the NOI contribution rate is near 100%, with average annual NOI of approximately $330 million.

$6.6 billion in contract revenue is a quantum leap for a company with annual revenue under $600 million. However, construction costs are also high, $10 million to $12 million per MW, with total investment for 175 MW approximately $1.75 billion to $2.1 billion, while the company's current long-term debt is already $1.8 billion. CleanSpark has completed equity funding and prepaid for long-lead equipment, pushing forward with real money, which is qualitatively different from transformation declarations at the PPT level.

The tenant also signed an exclusive letter of intent for the Texas portfolio (718 acres, 885 MW). If Sandersville is the prologue, Texas will be the main act.

Control Group: Core Scientific Has Passed the Inflection Point

In the same earnings season, Core Scientific provided a template for completed transformation.

Q2 revenue was $164.2 million, doubling year-over-year. Of this, AI/HPC colocation hosting revenue was $136.7 million, up over 1190% year-over-year, accounting for 83% of total revenue. Proprietary mining shrank to $21.5 million, accounting for only 17%. By mid-July, billing was already underway for 437 MW of customer capacity, with annualized hosting revenue of approximately $635 million. Plus the 529 MW large contract signed with AMD, the total lease portfolio reaches 1.1 GW, with potential contract revenue exceeding $24 billion.

Core Scientific Q2 net loss was $1.16 billion, but the source of the loss is completely different from MARA, mainly driven by changes in the fair value of warrants ($1.05 billion non-cash measurement change). Adjusted EBITDA was positive $41.1 million, and operating cash flow for the first half was positive $230.9 million.

From signing the first 200 MW contract with CoreWeave in June 2024 to Q2 hosting revenue of $136.7 million, Core Scientific's conversion cycle was approximately two years.

Three Stages, Three Valuations

Putting the three companies together, the progress grading of mining companies transforming to AI is very clear.

Core Scientific is in the "Revenue Realization Period".

83% of revenue comes from AI/HPC, operating cash flow is positive, and the market is already valuing it as a data center operator.

CleanSpark is in the "Contract Lock-in Period". The $6.6 billion lease is signed, funds are in place, but AI revenue is zero. Investors are buying future cash flows starting to realize in Q4 2027, with valuation based on discounted contract value.

MARA is in the "Platform Building Period". The 4.8 GW pipeline is the largest in scale, but there are no signed AI customers and no billable capacity; valuation is still mainly driven by BTC holdings.

To judge the real progress of mining companies' AI transformation, do not look at the GW number of the power pipeline, look at the billable MW number. MARA's 4.8 GW and Core Scientific's 437 MW being billed are worlds apart in value.

The underlying logic of mining companies' transformation holds. AI computing power demand growth speed is far faster than power infrastructure construction speed, with new substation scheduling in the US taking up to 5 to 7 years. Mining companies hold large-scale grid-connected power capacity, land, and cooling infrastructure, which are exactly the scarcest resources for hyperscale cloud providers. This is why Starwood is willing to cooperate with MARA and tech giants are willing to sign 20-year contracts with CleanSpark.

For the same kilowatt-hour of electricity, after accessing AI workloads, the revenue created can reach several times that of mining; power assets are switching from "pricing based on BTC price" to "pricing based on computing power rent".

But all revenue for MARA and CleanSpark still relies on mining. BTC price is hovering in the $64,000-$65,000 range, and production costs continue to rise after the halving. MARA sold 2,213 BTC in Q2 to maintain operations, and Long Ridge used BTC collateral for debt financing, equivalent to consuming Bitcoin assets to purchase computing power assets.

For investors, there is only one thing to care about: "Which company can truly convert electricity into AI revenue before cash flow runs out."