MARA's $611 Million Loss Is A Bitcoin Balance Sheet Shuffle — The Mining Business Is Running Better

The headlines say Marathon Digital lost $611 million in the second quarter. They also say its Bitcoin stash shrank 29%. Both are technically true, and both miss the story entirely.

$343 million of that loss — 56% of it — was an unrealized mark-to-market writedown on Bitcoin that never changed hands. The rest was a mix of depreciation, interest costs on new infrastructure debt, and $10.2 million in litigation settlements. Strip out the accounting noise and the mining operation underneath improved for the third straight quarter.

The production story the headline hides

MARA mined 2,422 Bitcoin in Q2. That's up 3% from the same quarter last year, even after the April 2024 halving cut block rewards in half. Energized hashrate climbed 22% to 70.3 exahash per second. Daily operating cost per petahash fell 4% to $27.70. The company's own-site energy cost sits at 4 cents per kilowatt-hour.

What changed is what matters. The mining business is more productive, more efficient, and cheaper to run than it was a year ago. The revenue decline — 27% to $174.9 million — came from Bitcoin's average price falling 28% during the quarter, not from worse operations. When the tape falls but the machine gets better, the setup starts getting interesting.

Where the Bitcoin actually went

The 29% decline in holdings — from 49,951 to 35,577 BTC — deserves closer scrutiny because it is the single most misread number from this quarter. The drop wasn't an operational bleed. It was a deliberate capital-allocation shift.

In March, MARA sold 15,133 BTC for roughly $1.1 billion. The proceeds funded a $1 billion repurchase of convertible notes at an approximate 9% discount, capturing about $88 million in immediate value. That alone accounts for 30% of the year-over-year decline.

The remaining gap is collateral. At quarter-end, 4,528 BTC were pledged and 4,742 BTC were loaned. After quarter-end, another 18,750 BTC were pledged to back a $600 million credit facility from Coinbase and Two Prime — the non-dilutive financing for the Long Ridge acquisition. That's roughly 54% of holdings now encumbered.

So the 29% headline is really a composite of one large strategic sale, collateral pledges for infrastructure debt, and lower Bitcoin prices. The actual exposure — the total BTC on the books whether free or pledged — still represents a substantial reserve. At the quarter-end spot price of $58,524, 35,577 BTC is worth approximately $2.1 billion.

The Long Ridge inflection

Here is where the next 12 months actually move. MARA's $1.5 billion acquisition of Long Ridge Energy adds a 505-megawatt combined-cycle gas plant in Hannibal, Ohio, 1,600 contiguous acres, and a line of sight to over 1 GW of total potential capacity. More than 70% of Long Ridge's existing power output is already under long-term contracts, which means stable cash flow on day one — unrelated to crypto.

Management projects approximately $144 million in annualized adjusted EBITDA from the asset. On a $1.5 billion enterprise value, that's roughly an 8.4x earnings multiple for a contracted utility-grade asset in the PJM electricity market. The deal awaits FERC approval, expected before year-end. The risk is execution timing, not deal quality. And if PJM imposes conditions limiting exclusive use — a live concern the market monitor has flagged — the worst case is still access to a high-margin gas plant with sub-$15/MWh operating costs.

When Long Ridge closes, MARA's power portfolio jumps from 1.9 GW to roughly 2.2 GW today, with a pipeline targeting 4.8 GW once the 2 GW Matagorda County site in Texas comes online. Management expects to sign at least two AI or high-performance computing leases by year-end. The Matagorda construction timeline targets 1 GW by October 2027 and full capacity by 2028.

Why the market still sees the old MARA

The stock is trading around $10.70 after the earnings report, well off its 52-week high of $23.45. The implied market capitalization is roughly $3.2 billion against $1.68 billion in shareholders' equity. At the quarter-end BTC price, the company's Bitcoin alone is worth about $2.1 billion, or roughly $8.80 per share. That means the market is assigning roughly $1.90 per share to every other asset, business line, and optionality the company owns — the mining operation, Long Ridge, Matagorda, Exaion, and the AI lease pipeline.

The market is still pricing MARA as a Bitcoin miner with a balance sheet problem. The numbers say it is increasingly a power and infrastructure company that happens to mine Bitcoin while it waits for leases to sign. That mismatch between what the market sees and what the balance sheet and pipeline are becoming is where inflections tend to happen.

What needs to go right over the next 12 months

The financial bridge is straightforward. Management guides to $0.43 EPS in Q3 and break-even in Q4. Full-year 2026 revenue is projected at $810 million, rising to $946.5 million in 2027. Third-party hosting contracts expire between Q3 2027 and Q1 2028, removing a layer of cost that has inflated unit economics. Long Ridge contributes $144 million in annualized EBITDA once it closes.

If the Long Ridge deal closes on schedule and at least one AI lease lands in 2026, the infrastructure narrative shifts from "plans" to "contracted revenue." That single data point — a signed lease at a named price — would force the re-rating because it converts optionality into earnings visibility.

What breaks the thesis

FERC denies or materially conditions the Long Ridge deal. This is the single most important risk. The deal is not closed, and regulatory pushback in PJM over exclusive data center use is a real concern. No Long Ridge means no $144 million EBITDA bridge and the whole infrastructure thesis gets delayed by quarters, possibly years.

Bitcoin collapses below $40,000 and stays there. MARA's collateral positions would trigger margin calls on the $600 million credit facility. The non-dilutive financing structure becomes dilutive if BTC reserves are forced into liquidation.

No leases sign by year-end and management's confidence exceeds reality. If "at least two" becomes zero, the infrastructure pivot remains a presentation deck.

The setup

The market bar has cleared. A $611 million loss and a 29% BTC decline have done the heavy lifting for skepticism. The mining operation is running better. The balance sheet was cleaned with the convertible note buyback. A contracted utility asset worth $144 million in annualized EBITDA is sitting in regulatory limbo. The company's total prospective power capacity targets 4.8 GW — nearly a 2.5x increase from the start of the year.

The old story is "Bitcoin miner with a volatile balance sheet." The numbers already point to something else: a power platform with contracted revenue, growing hashrate, declining costs, and a crypto reserve that functions as both collateral and optionality. The gap between those two frames is where the move happens — if Long Ridge closes and at least one lease follows.

Discipline over ego. The tripwire is a denied Long Ridge deal. Everything else is noise around a bridge that may already be under construction.