Marathon Digital's $611 Million Loss Hides an Accelerating Operation — and a Balance Sheet That's Starting to Worry Me

The headline from Marathon Digital's Q2 2026 earnings report is ugly: a $611 million net loss on $175 million in revenue, down from $808 million in net income a year ago. Shares dropped 5.25% to $10.65 on the day of the report. The market heard catastrophic. But if you separate the accounting marks from the operating engine, the picture is more nuanced — and one part of it is genuinely troubling.

Let me start with what actually drove that loss, because the number is almost entirely accounting fiction. Approximately $343 million of the $611 million net loss came from unrealized mark-to-market adjustments on Marathon's Bitcoin treasury holdings. Bitcoin's average price fell 28% during the quarter, dragging down revenue by roughly $66 million. The remaining loss reflects the company's strategic investments in power infrastructure and debt financing costs. This does not mean the operational business collapsed. It means the company holds a large amount of a volatile asset, and accounting rules force you to mark it down when the price drops.

Now let's talk about the operating engine, because that's where the data actually tells a story of acceleration. Marathon's energized hashrate increased 22% year-over-year to 70.3 exahash per second — the computing power that mines Bitcoin. Bitcoin production rose 3% to 2,422 BTC in the quarter, with 700 blocks won, up 1% from a year ago. The company's share of available mining rewards reached 5.9%. And on the efficiency side, daily operating cost per petahash declined 4% to $27.7, with energy costs at owned sites holding at just $0.04 per kilowatt-hour. That cost discipline is what separates surviving miners from those that get crushed when Bitcoin turns lower.

From a revenue perspective, the 27% year-over-year decline to $175 million is real, but it's a commodity-price story, not a volume story. Production volume actually increased — Marathon mined more Bitcoin than it did a year ago. The revenue drop is almost entirely explained by the lower average Bitcoin price. When you're selling a commodity, your revenue is a function of price times volume. The volume side is working; the price side is not.

Now let's talk about what's actually keeping me up at night: the balance sheet. This is where the $611 million loss is a surface symptom of something deeper. Total assets declined from $7.29 billion at year-end 2025 to $4.35 billion at June 30 — a 40% drop. Total equity fell from $3.48 billion to $1.68 billion. Long-term liabilities stand at $2.0 billion.

The leverage build is intentional but aggressive. Marathon secured $600 million in Bitcoin-backed credit facilities with Coinbase and Two Prime at a weighted average cost of debt of 7.56%. That's cheap capital when Bitcoin is stable, but it's a one-way bet when the collateral is the same asset whose price volatility just wiped $343 million off the income statement. On top of that, the pending Long Ridge acquisition — a $1.5 billion enterprise-value deal for a 505 MW natural gas power plant in Ohio — requires assuming approximately $785 million in additional debt, backstopped by a Barclays bridge loan.

Here's the part that matters most for assessing survival: 54% of Marathon's 35,577 Bitcoin holdings are now pledged as collateral under borrowings. That's roughly 19,200 BTC tied to debt obligations. After quarter-end, an additional 18,750 BTC were pledged as initial collateral for the Long Ridge acquisition financing. When I look at that number, I'm thinking about margin calls, not growth plans. If Bitcoin drops significantly from current levels near $58,500, the collateral coverage thins fast. The company would face either raising cash to post additional collateral or liquidating unrestricted Bitcoin at depressed prices — neither of which is a good outcome.

The counterargument, which management is making forcefully, is that this leverage is the price of admission to a much bigger business. Marathon's strategic pivot — what it calls the "Digital Infrastructure Triad" — targets AI computing, Bitcoin mining, and critical IT infrastructure as three revenue streams built on the same foundation: secured power, land, and compute capacity. The Long Ridge acquisition adds ~$144 million in annualized adjusted EBITDA with roughly 70% of power output covered by long-term contracts. The 2 GW Matagorda County facility in Texas, a $600 million project targeting AI and high-performance computing workloads, is positioned to begin construction late this year. The total power capacity pipeline targets 4.8 GW — a dramatic expansion from the 1.4 GW currently operational.

The re-rating thesis is explicit in management's math. Bitcoin miners historically trade at 6x–12x EV/EBITDA. AI infrastructure companies trade at 20x–25x. If Marathon can credibly become an AI power-and-compute platform rather than a Bitcoin miner that happens to own some data centers, the multiple expansion alone could justify the current valuation even if earnings don't grow much. Management expects to sign at least two AI infrastructure leases before year-end and says tenant demand has exceeded initial expectations.

But expectations are not leases, and leases are not EBITDA. None of these AI projects are generating revenue yet. The European subsidiary Exaion is expected to produce low-eight-digit revenues this year. The Hashrate Under Management blockchain platform is projected as an eight-digit annualized business — not material to total revenues in the near term. The hosting agreements with third-party operators are expiring between Q3 2027 and Q1 2028, which will eliminate hosting costs but also remove those revenue lines. Marathon is burning through its Bitcoin treasury to fund an infrastructure build-out that hasn't proven it can attract paying tenants at scale.

From a valuation perspective, shares at $10.65 reflect deep skepticism. Comparing Marathon to peers, IREN trades at $37.93 with a $13.6 billion market cap and an 86x P/E, Cipher Digital sits at $18.21 with a $7.6 billion market cap, Hut 8 is at $90.65 with an $11.2 billion market cap, and CleanSpark is at $12.75. The peer group itself is wildly bifurcated — some miners are commanding premium AI-infrastructure multiples while others remain cheap. Marathon's position at the bottom of the pack reflects the market's judgment that its AI story is the least credible among the names actively pursuing it. Whether that's right or wrong is the $10.65 question.

Management's guidance adds some structure to the recovery narrative. Q3 2026 EPS is projected at $0.43, and Q4 is expected to be break-even. Full-year 2026 revenue is forecast at $810 million, growing to $947 million in 2027. Those numbers assume Bitcoin price stability, successful closing of Long Ridge with FERC approval, and the execution of planned AI leases. That's three big assumptions stacked on top of each other. If any one of them breaks — and Bitcoin price stability is the hardest one to control — the guidance gets harder to hit.

While it's true that Marathon's operating metrics are improving and its strategic direction is intelligently designed, I would argue that the leverage risk has crossed a threshold where the math is no longer in the company's favor. A company that has pledged more than half its Bitcoin treasury, assumed nearly $800 million in acquisition debt, and carries $2 billion in long-term liabilities against $4.35 billion in total assets is playing a high-stakes game where the collateral itself is the biggest risk factor.

All things considered, the operating business is fine. Hashrate is growing, efficiency is improving, and the strategic pivot toward contracted AI power infrastructure is at least directionally correct for a sector where demand is outstripping supply. But value investing is not just about buying cheap stocks — it's about buying stocks trading below intrinsic value with a reasonable margin of safety. At this point, the intrinsic value of Marathon is highly uncertain. The AI revenue story is unproven, the leverage is elevated, and the collateral concentration in a single volatile asset creates a tail risk that could accelerate quickly.

I am rating MARA a Hold. The stock is cheap, but cheap isn't the same as safe. If Bitcoin stabilizes, the Long Ridge deal closes, and those AI leases materialize in the second half of the year, this could be the entry point that looks great in retrospect. But even if all of that happens, there are other names in the AI-infrastructure and mining space with cleaner balance sheets that deserve consideration first. I'll be watching the FERC decision on Long Ridge, the actual lease signings, and Bitcoin's price action with particular attention. Any one of those three moving against Marathon's case, and I'll be reconsidering this position on the downside.