The profit rebound came from Bitcoin accounting
MARA Holdings reported third-quarter net income of $632.1 million, or $1.28 per diluted share, reversing the $611.3 million loss reported in the second quarter and exceeding the $123.1 million earned a year earlier.
The headline is not a clean measure of the mining business. MARA recorded an $885.3 million unrealized fair-value gain on its digital assets as Bitcoin rose from $58,524 at the end of June to $83,556 at the end of September. The mark was larger than total net income, just as the previous quarter's loss was heavily shaped by a Bitcoin markdown.
That accounting sensitivity is now central to reading MARA's earnings. A rising Bitcoin price can produce a large reported profit without a matching increase in cash, while a falling price can create a large loss even when the mining fleet remains operational.
Revenue fell even as MARA mined more Bitcoin
Revenue declined 35% to $164.2 million from $252.4 million in the third quarter of 2025. MARA attributes most of the decline to a 37% drop in the average price of Bitcoin mined, which reduced revenue by $94.8 million. Higher Bitcoin production offset only $11.7 million of that pressure.
The operating statistics were more resilient than the revenue line:
- MARA produced 2,247 BTC, up 5% from a year earlier but down 7% from the second quarter.
- Blocks won increased 2% year over year to 644, but fell from 700 in the prior quarter.
- Energized hashrate reached 71.2 EH/s, up 18% year over year and 1% sequentially.
- Cost per petahash per day improved 2% to $30.80.
- Purchased energy cost at owned sites remained approximately $0.04 per kWh.
- Purchased energy cost per Bitcoin was $39,674, slightly above $39,235 a year earlier as network difficulty and power costs offset fleet-efficiency gains.
The result is a miner producing more efficiently, but selling the output into a much weaker average Bitcoin price than in the comparison quarter. Scale improved; revenue did not.
MARA is no longer treating every Bitcoin as untouchable
The largest strategic change is on the balance sheet. MARA ended September with 34,012 BTC, down from 53,822 at the end of 2025. It sold 3,891 BTC during the quarter at an average price of $75,300 and disclosed 26,984 BTC sold during the first nine months of 2026.
After quarter-end, MARA sold another 13,597 BTC at approximately $81,039, generating about $1.1 billion. It used $750 million to repay its Bitcoin-collateralized term loans in full, releasing 17,477 BTC that had been pledged. The company says less than 5% of its remaining Bitcoin is now loaned or pledged.
MARA has reduced total debt by $1.4 billion, or 37%, since the start of the year. That lowers collateral and refinancing risk, but it also changes the equity story. Bitcoin is now explicitly a source of liquidity for debt repayment, infrastructure investment and potential share repurchases, not only a treasury asset intended to be held indefinitely.
The board also authorized a share-repurchase program of up to $500 million. Authorization is not execution: the program does not require MARA to buy a specific number of shares, and management says timing will depend on price, liquidity and competing uses of capital.
The AI thesis has one contract, but the larger proof is still pending
MARA describes itself as an energy-backed digital infrastructure company spanning Bitcoin mining, power, land and compute. The quarter delivered one tangible non-mining proof point: Exaion signed a four-year $79 million GPU-as-a-service contract with its first US customer on 100% take-or-pay terms.
Management says Exaion has a pipeline exceeding $3 billion in potential contract value, with about $1 billion targeted for delivery by the end of 2027. The company is clear that this pipeline is not yet contracted or committed.
The more important test is MARA's target of signing two data-center leases. Management says discussions are advanced, but acknowledges that the goal of signing both before year-end could slip into 2027. Until those agreements exist, the value of MARA's powered land and interconnection capacity remains a development thesis rather than contracted cash flow.
MARA also expects to close the $1.5 billion Long Ridge Energy acquisition in the coming weeks after receiving federal approval. The asset includes a natural-gas power plant authorized for 485 MW, with 505 MW of nameplate capacity expected in the first half of 2027, plus land that could support a data-center campus with more than 1 GW of total potential capacity.
What the quarter says about MARA now
MARA's third quarter is stronger than the net-income headline alone and weaker than the same headline suggests.
The stronger case is balance-sheet repair, higher year-over-year production, improved compute efficiency, a first contracted Exaion customer and a credible collection of power assets. The weaker case is a fourth consecutive sequential revenue decline, mining economics still tied to Bitcoin price and network difficulty, substantial reliance on Bitcoin sales for liquidity, and major AI leases that remain unsigned.
The key investor question has therefore shifted. It is no longer simply how many Bitcoin MARA can mine or hold. It is whether the company can turn powered sites into financeable, long-duration infrastructure contracts faster than its mining revenue and Bitcoin treasury are consumed by debt reduction and expansion.
Follow MARA's ongoing company coverage on the Signal21 MARA page.
This publication is market analysis, not a recommendation to buy or sell.