Report contents
Business model
Business model: cheap power, unproven tenants
MARA sells computing power and the bitcoin it produces. There is no traditional customer on the mining side: the Bitcoin protocol pays block rewards and fees to whoever wins blocks, and MARA converts the proceeds to dollars at market prices. In the third quarter of 2026 it mined 2,247 bitcoin with 71.2 EH/s energized, equal to 5.4% of network rewards, and reported revenue of $164.2 million, down 35% year over year. Of that decline, $94.8 million came from lower average bitcoin prices, management said on the November 9 call. Volume rose; price fell. That is the shape of the mining business.
The cost side explains the fragility. MARA pays about $0.04 per kWh at sites it owns and operates, which is the core asset. Even so, energy cost per bitcoin rose to $39,674 from $39,235 a year earlier as network difficulty increased. Revenue per coin mined works out to roughly $73,000 against that power bill, an estimate from reported figures that ignores non-mining revenue, so energy consumes more than half the top line before depreciation, G&A and capital spending. A producer with that spread does not set prices. Hashprice does.
Where MARA differs from a pure price-taker is that it owns the interconnection queue, not just the machines. Current capacity is about 1 gigawatt, with a path management puts above 4 gigawatts: the Long Ridge acquisition adds 485 megawatts of authorized capacity, expected to reach 505 megawatts in the first half of 2027 after FERC approval, and the Matagorda County site carries 2 gigawatts of potential capacity, currently classified as study load under ERCOT's Batch Zero process. That queue is the strategic asset. Whether it becomes a business depends on tenants.
One lever already pays: Vertebrai, the energy-optimization software, cut curtailment at MARA's Texas wind farm to 22% of potential generation between April and September 2026, against an estimated 46% without it. For a business whose margin is the gap between hashprice and a four-cent kilowatt-hour, recovered generation is output without new capital.
The pivot runs through contracts MARA has not signed
The intended move is from selling bitcoin at spot to leasing power and compute on contract. Exaion, the GPU-as-a-service unit, has one disclosed contract: $79 million over four years with a single US customer on 100% take-or-pay terms. Management cites a $3 billion pipeline and targets $1 billion of delivery by the end of 2027. Those are statements of intent, not backlog. On colocation, management said it is targeting two lease signings before year-end 2026 and that the timing "could move into next year." Until a lease is signed and priced, the 4-gigawatt pipeline is a capital commitment in search of a customer.
Bargaining power is the open question. Chairman and CEO Frederick Thiel said demand for the company's sites has picked up and that hyperscalers now favor developers with credible paths to power over those with delayed projects. That may be accurate and still not produce attractive terms, because the buyers are a handful of large tenants who can compare sites. The disclosure that hosting contracts covering 29% of energy capacity expire between the third quarter of 2027 and the first quarter of 2028, which management frames as a chance to cut third-party costs, also shows how much of the current footprint runs on contracts MARA does not control.
The balance sheet is part of the model, not a sidebar. MARA cut total debt by $1.4 billion, or 37%, during the year, then repaid a further $750 million after the quarter by selling 13,597 bitcoin and releasing 17,477 coins pledged as collateral. It ended the quarter with 34,012 bitcoin worth about $2.8 billion and $3.8 billion of combined cash and bitcoin. Bitcoin therefore serves as inventory, collateral and funding source at once. The $632.1 million of net income is mostly an $885.3 million unrealized fair value gain, so it says little about operating earning power.
MARA does not publish a segment P&L splitting mining, hosting, Exaion and the $2.5 million of bitcoin lending income, and it does not break out revenue by geography beyond the Texas concentration. That gap makes the pivot harder to underwrite than the headline capacity numbers suggest.
What would change the judgment. A signed lease with a creditworthy tenant at disclosed terms, or a second and third take-or-pay Exaion contract, would show MARA can earn contracted returns on its power. Another two quarters of slipping signing targets while hashrate keeps rising would instead confirm that the infrastructure story is optionality attached to a commodity miner, not a second business.
Questions worth following
- Can MARA sign anchor tenants for its power capacity?
A signed lease would turn idle megawatts into contracted take-or-pay revenue and mark a real shift from commodity mining to infrastructure. Without one, the 4-gigawatt pipeline is a capital commitment with no demonstrated tenant demand, and the company remains a price-taker on bitcoin.
- Can MARA keep gaining bitcoin network share as hashrate grows?
Reward share is the main lever management controls when bitcoin prices fall. Gains require deploying capital faster than the rest of the network, and they are paid for in dollars per petahash, so the answer decides whether mining produces cash or only consumes it.
- Will Exaion's contract pipeline convert into delivered revenue?
Exaion is the clearest test of whether MARA can sell compute on contract instead of selling bitcoin at market prices. The $3 billion pipeline and $1 billion end-2027 target are management statements; only signed, priced contracts would show the unit is a business.
Financial quality
Financial quality: profits are Bitcoin marks, mining cash is still negative
MARA's reported earnings stopped describing its mining business in 2025, when fair-value accounting began running every Bitcoin price change through net income. The September 2026 quarter shows the gap. Net income of $632.1 million rested on an $885.3 million unrealized gain on digital assets, while revenue fell 35% to $164.2 million and operating cash flow was -$170 million. Adjusted EBITDA of $857.7 million, up 117% year over year, is the same mark seen through a different lens. On the figures the company reported on November 9, the mining operation has not funded itself from operations in any year since 2021. The treasury does the earning; the rigs do the spending.
| USD bn; ratios in % or x | FY2021 | FY2022 | FY2023 | FY2024 | FY2025 | FY2026 Q3 (3 mo) | Assessment |
|---|---|---|---|---|---|---|---|
| Revenue | 0.16 | 0.12 | 0.39 | 0.66 | 0.91 | 0.16 | Weak: price-driven, falling |
| Operating income | -0.06 | -0.67 | 0.22 | 0.31 | -0.82 | 0.18 | Very weak |
| Net income | -0.04 | -0.69 | 0.26 | 0.54 | -1.31 | 0.63 | Very weak: mark-to-market |
| Operating cash flow | -0.02 | -0.18 | -0.32 | -0.68 | -0.80 | -0.17 | Very weak |
| Capex (derived) | 0.71 | 0.04 | 0.02 | 0.25 | 0.41 | 0.09 | Heavy |
| Free cash flow | -0.73 | -0.22 | -0.34 | -0.93 | -1.21 | -0.26 | Very weak |
| ROE (%) | -7.5 | -129.8 | 26.1 | 18.8 | -34.5 | 31.6 | Very weak |
| ROA (%) | -4.2 | -52.5 | 16.4 | 12.3 | -18.6 | 12.7 | Very weak |
| Net margin (%) | -23.3 | -589.4 | 67.4 | 82.5 | -144.6 | 386.3 | Very weak |
| Liabilities / assets (%) | 53 | 68 | 19 | 39 | 52 | 57 | Weak, improving |
| Current ratio (x) | 49.9 | 10.6 | 30.5 | 4.9 | 1.3 | 2.3 | Strong on paper; excludes Bitcoin |
| Total assets | 1.45 | 1.19 | 1.99 | 6.80 | 7.29 | 5.68 | Volatile |
| Total debt | 0.73 | 0.78 | 0.33 | 2.45 | 3.60 | 3.02 | Weak |
| Cash | 0.27 | 0.10 | 0.36 | 0.39 | 0.55 | 0.95 | Adequate |
The income statement tracks the coin, not the rigs
Return on equity ran from -129.8% in FY2022 to 26.1% in FY2023 and back to -34.5% in FY2025; net margin spans -589% to 386%. Neither measures an operating business. Both track the Bitcoin price applied to a marked treasury. Gross margin is no cleaner, reading -119% in the December 2025 quarter and +83% a quarter earlier, because impairment, depreciation and remeasurement land in different lines each period. In the latest quarter gross profit was $23 million while operating income reached $180 million, so gains booked below the gross line did more work than mining.
Unit economics are the only stable read. Energized hashrate was 71.2 EH/s, up 18% year over year, and production was 2,247 Bitcoin, up 5%, lifting the company's share of network rewards to 5.4%. Energy cost was $39,674 per Bitcoin at owned sites and $0.04 per kWh. Management attributed the higher cost per coin to rising network difficulty. Against roughly $73,000 of revenue per coin mined, that leaves a site-level spread before hosting fees, staff, depreciation and interest. Difficulty decides whether the spread survives.
Cash: five negative years, then a change in who pays
Operating cash flow was negative in each year from FY2021 through FY2025 and in all three 2026 quarters. Free cash flow was negative in every period, about -$3.4 billion cumulative over FY2021–FY2025. Capital spending absorbed $410 million in FY2025, 45% of revenue, and $80–100 million a quarter in 2026, exceeding operating cash flow every time. Financing covered the gap: $3.95 billion of net inflow in FY2024 and $1.63 billion in FY2025. In 2026 investing cash flow turned positive, $1.19 billion and $1.47 billion in the first two quarters, as Bitcoin was sold and debt retired. The funding mix improved, but the cash now comes from the coin stash instead of lenders and shareholders.
| USD bn; ratio in x | Q1 FY25 | Q2 FY25 | Q3 FY25 | Q4 FY25 | Q1 FY26 | Q2 FY26 | Q3 FY26 | Assessment | Trend |
|---|---|---|---|---|---|---|---|---|---|
| Revenue | 0.21 | 0.24 | 0.25 | 0.20 | 0.17 | 0.17 | 0.16 | Weak | Declining |
| Operating income | -0.54 | 0.68 | 0.05 | -0.33 | -0.25 | -0.27 | 0.18 | Very weak | Erratic |
| Net income | -0.53 | 0.81 | 0.12 | -1.71 | -1.26 | -0.61 | 0.63 | Very weak | Mark-driven |
| Operating cash flow | -0.21 | -0.16 | -0.20 | -0.23 | -0.25 | -0.47 | -0.17 | Very weak | Negative throughout |
| Capex (derived) | 0.04 | 0.12 | 0.08 | 0.16 | 0.08 | 0.10 | 0.09 | Heavy | Steady |
| Total assets | 6.44 | 7.72 | 9.15 | 7.29 | 4.95 | 4.35 | 5.68 | Volatile | Shrank, then rebuilt |
| Total liabilities | 2.72 | 2.92 | 3.98 | 3.81 | 2.62 | 2.59 | 3.23 | Weak | Peaked a year ago |
| Cash | 0.20 | 0.11 | 0.83 | 0.55 | 0.53 | 0.42 | 0.95 | Adequate | Rebuilding |
| Long-term debt | 2.30 | 2.25 | 3.25 | 3.20 | 2.22 | 1.93 | 2.68 | Weak | Lower year over year |
| Current ratio (x) | 0.79 | 0.54 | 2.09 | 1.27 | 1.84 | 0.89 | 2.31 | Thin | Volatile |
The balance sheet carries the same message. Total assets fell from $9.15 billion at September 30, 2025 to $4.35 billion nine months later, then recovered to $5.68 billion, mostly from marking the Bitcoin position down and selling part of it. Assets that can halve without any change in the plants describe a leveraged bet on one price.
Deleveraging is real; the collateral is still the coin
Liabilities were about 57% of assets at September 30, 2026, with $3.02 billion of debt against $0.95 billion of cash. Management put the year-to-date debt reduction at $1.4 billion, or 37%, after a further $750 million repaid in October by selling 13,597 Bitcoin, which released 17,477 coins from collateral. That sale would leave roughly 20,400 coins unless MARA buys more. Interest expense is about $10 million a quarter, so coverage is not the binding constraint; collateral value and refinancing are. The reported current ratio of 2.31 also overstates strain relief in one direction and understates liquidity in another: the 34,012 Bitcoin held at quarter end, about $2.8 billion, are not counted in current assets, and that same holding stands behind the loans.
What would change the read
Two consecutive quarters of positive operating cash flow, with hashrate near current levels, would show the mines can carry corporate costs. Net income cannot answer that question. Also worth tracking: disclosed pledged Bitcoin and the secured share of debt, since a price drawdown could force sales at the worst moment; revenue per EH/s, because hashrate grew 18% while revenue fell 35%, meaning capital spending is buying share of a shrinking payout; and whether the new $500 million buyback is funded from Bitcoin sales while free cash flow is negative. The AI colocation buildout, including the pending Long Ridge purchase and a $79 million GPU-as-a-Service contract, adds capital demands on top of mining. How that is financed will decide whether the pivot dilutes shareholders or adds leverage.
Questions worth following
- Can MARA's mining operation turn operating cash flow positive?
Reported profit reflects Bitcoin marks, so only cash flow tests whether 71 EH/s of capacity earns its keep. Two consecutive quarters of positive operating cash flow, without leaning on treasury sales, would show the mining business can carry corporate costs and debt service.
- Can MARA cut debt faster than its Bitcoin collateral loses value?
Much of MARA's borrowing is secured by Bitcoin and mining equipment. If the treasury's value falls faster than debt is repaid, lenders can demand more collateral or force asset sales at low prices, turning a manageable leverage ratio into a liquidity problem.
- Can MARA fund its AI colocation buildout without new leverage?
The pivot toward data-center and GPU hosting needs capital for Long Ridge and new sites while mining cash flow is negative. Equity or Bitcoin sales would dilute the mining recovery; new secured debt would raise collateral risk. Financing may matter more than the contracts themselves.
Generated by DeepVal AI. Data and models may contain errors or be outdated. For research only; not personalized investment advice.
Continue reading the full report
- Key risksWhat could break the investment case
- Valuation and margin of safetyWhat the current price assumes, and how much room for error it leaves
- SummaryOur overall judgement and what to watch next
Free to sign up · 20 research points · unlocking this report uses 10
20 points cover 2 full reports. No card needed.
Research reflects its generation date. Preview the first two modules.
More on MARA Holdings, Inc.
MARA Holdings, Inc.
Business, moat and key questions
Open company page