Marathon Digital Holdings, Inc.
MARA is paying a growth-stock multiple for a business with collapsing margins, repeated earnings misses, and a levered AI-power pivot that still lacks proof of durable returns.
MARA is still being valued like a scarce growth platform when the actual financials look like a stressed cyclical. FY2025 revenue grew 24.8% to $0.9B, but that growth came with a collapse in profitability: operating margin fell to -85.8% from 46.6% the year before, net margin was -46.7%, EPS was -$3.69, and free cash flow was -$0.6B. A company posting those numbers should not trade at 6.27x sales, especially against a peer average of 2.7x. The market is still paying for optionality that has not become earnings.
The current bull case rests on three things. First, scale in mining, with 70.3 EH/s energized, can support operating leverage if bitcoin economics improve. Second, the power portfolio, targeted at about 4.8 GW post transactions, could create a more durable infrastructure business than pure mining. Third, Long Ridge could bring $144M of annualized EBITDA with roughly 70% contracted output. Those are real positives, but they are not enough to underwrite today's valuation because each one is conditioned on execution, approvals, and financing. Long Ridge still needs FERC clearance, Matagorda's up to 2 GW is subject to ERCOT and interconnection approvals, and the company has already layered on $600M of additional borrowing at 7.56% while 54% of bitcoin holdings are pledged.
The other side of the ledger is harsher and more visible. Q2 2026 revenue fell to $175M from $239M as average bitcoin price dropped 28%, EBITDA was -$0.43B, adjusted EBITDA was -$361M, and net loss widened to $611M. This is not a one-quarter wobble from an otherwise dependable operator. MARA has missed consensus EPS in 7 of the last 8 quarters, including misses of -52.8%, -23.1%, -100.0%, and -1,060.4%. That record matters because the investment story now asks investors to trust future AI lease signings, lower G&A, hosted-cost rolloff, and acquisition synergies that have not yet shown up in reported numbers.
Price action does not fix that. The stock has rallied 47.77% in a month and sits 20.9% above its 200-day average, but it remains -43.5% below the 52-week high because the business model is still unstable. High short interest at 29.55% can fuel squeezes, not intrinsic value. The clean way to own this name would be after stable contracted EBITDA is visible and leverage is better covered by cash flow. At $13.24, that proof is missing.
FERC approval arrives on time, at least two leases are signed, Long Ridge closes and contributes toward the cited $144M annualized EBITDA, and short covering extends the rally.
Some strategic milestones land, but weak earnings quality, high leverage, and continued execution misses keep valuation anchored near the engine target.
Long Ridge slips, lease signings disappoint, bitcoin mining margins stay under pressure near industry production-cost stress, and the market compresses MARA toward peer-like sales multiples.
Quant call (SELL) diverges from analyst consensus (Buy: 4 buy / 0 hold / 2 sell) — resolve explicitly.
The single greatest edge an investor can have is a long-term orientation.
Forward earnings power is still too speculative to support the current valuation because trailing results are deeply negative and the next leg of the story depends on closing Long Ridge, signing at least 2 leases by year-end 2026, and monetizing a 4.8 GW power footprint. The moat today is mostly scale in bitcoin mining and access to power assets, but over 3-5 years that moat looks fragile because AI infrastructure attracts much larger, lower-cost competitors while bitcoin mining remains commodity-like and heavily tied to network difficulty and coin price. AI is not yet an advantage here, it is a hoped-for redeployment path for stranded or flexible power, and the failed Clarity Act vote also shows regulation can still slow crypto-adjacent equity stories. Until forward earnings become less dependent on bitcoin marks and debt-funded deals, the stock does not deserve a premium sales multiple.
- Valuation remains stretched at 6.27x P/S versus 2.7x peers, despite the engine's $6.6 price target implying -50.2% downside.
- Operating performance deteriorated sharply, FY2025 operating margin fell to -85.8% from 46.6% the prior year, and Q2 2026 EBITDA margin hit -245.8% on $175M revenue.
- Balance-sheet and execution risk are elevated, with $2.5B total debt, 1.49 debt/equity, 0.89 current ratio, $-0.6B FCF, and 7 misses in the last 8 quarters.
- Mining scale is still growing, energized hash rate reached 70.3 EH/s in Q2 2026, up 22% Y/Y, versus the 2024 Q3 target framework that had aimed for 50 EH/s.
- Unit economics showed one real operating improvement, daily cost per petahash improved 4% Y/Y to $27.70, and hosted mining costs should ease as contracts begin expiring from Q3 2027 through Q1 2028.
- The power platform could become more valuable if execution holds, with a post-transaction portfolio of about 4.8 GW, Matagorda adding up to 2 GW potential capacity, and Long Ridge expected to contribute about $144M annualized EBITDA with roughly 70% contracted output.
- Positioning can amplify upside in squeezes, short interest is 29.55% of float with 110.1M shares short, while the stock is 20.9% above its 200-day average after a 47.77% 1-month rally.
- The income statement is broken at current bitcoin and mark-to-market conditions, FY2025 EPS was -$3.69, Q2 2026 EPS was -$1.60 versus $0.1666 expected, and adjusted EBITDA was -$361M.
- Leverage has risen into a cyclical business, including $600M of new Coinbase and Two Prime borrowings at a 7.56% weighted average cost and a $1.5B Long Ridge acquisition funded with bitcoin-backed debt and assumed debt.
- Liquidity is thin for a business pursuing expansion, current ratio is just 0.89, total debt is $2.5B, and roughly 54% of bitcoin holdings are pledged as collateral.
- The stock still carries a premium multiple without premium quality, with a 29/100 quality score, 66/100 risk score, and repeated execution misses despite the pivot toward AI infrastructure.
| Metric | Value | Context |
|---|---|---|
| Current price | $13.24 | Engine target implies -50.2% downside |
| Price target | $6.6 | Low confidence, basis sales_quality_blend |
| FY2025 revenue | $0.9B | +24.8% YoY, growth did not translate into profits |
| FY2025 EPS | -$3.69 | -149.5% YoY, deeply loss-making |
| FY2025 operating margin | -85.8% | Down from 46.6% in prior FY |
| FY2025 net margin | -46.7% | Loss profile remains severe |
| Valuation | P/S 6.27x, P/B 3.04x | P/S vs peer average 2.7x, P/E N/A vs peers 12.8x |
| Balance sheet | $2.5B debt, 1.49 debt/equity, 0.89 current ratio | Leverage is high for a volatile earnings model |
| Free cash flow | -$0.6B | Capital needs exceed internally generated cash |
| Q2 2026 performance | $175M revenue, -$0.43B EBITDA, -$1.60 EPS | EBITDA margin -245.8%, EPS miss -1060.4% |
| Quality and risk | Quality 29/100, Risk 66/100 | High risk, weak quality profile |
| 52-week position | $6.66-$23.45 | Current price is -43.5% below high, 20.9% above 200-day average |
The biggest risk is that leverage meets prolonged weak bitcoin economics, because MARA already has $2.5B debt, 54% of bitcoin collateral pledged, and a Q2 2026 net loss of $611M.
Q3 earnings on November 3, 2026, especially any update on Long Ridge FERC timing, lease signings, and whether EBITDA losses narrow from Q2's -$0.43B.
Ownership is institutionally concentrated, with the top three holders at 53.2% and insiders at only 1.7%, which limits the alignment argument. Short interest is very high at 29.55%, or 110.1M shares, but days-to-cover is only 2.4, so squeeze risk is real but can also reverse quickly once momentum fades. Options positioning is listed as a positive gamma regime, which can dampen volatility somewhat near term, but it does not change the weak fundamental setup.
Core economics remain tied to bitcoin price and network difficulty, Q2 2026 revenue fell to $175M from $239M as average bitcoin price dropped 28%.
Long Ridge is expected to add about $144M annualized EBITDA with roughly 70% contracted output if closed before year-end 2026.
Asset pipeline is concentrated in U.S. power markets including Ohio, Texas, and Long Ridge, with ERCOT and FERC approvals now material operating variables.
MARA pushed its AI infrastructure buildout, but Q2 results were weighed by large losses and weak bitcoin pricing.
Execution has been mixed operationally and poor financially. Management said in 2024 Q3, "So, we have hit more than 40 exahash already and we are driving towards the 50 exahash target," and the company has since exceeded that level with 70.3 EH/s energized in Q2 2026, so the hash-rate buildout was delivered. But the economic result missed badly: 2024 Q3 EBITDA margin was 135.1% on $0.25B revenue, versus -245.8% EBITDA margin on $0.17B revenue in Q2 2026, and the company has missed EPS estimates in 7 of the last 8 quarters, including -1,060.4% in August 2026. Management also said in 2024 Q3, "These three facilities have a combined interconnect-approved capacity of 372 megawatts, which MARA intends to fully energize by the end of 2025," but the provided data does not confirm that milestone directly, while strategy has shifted materially from dismissing AI hosting as unattractive to actively pushing AI infrastructure and lease discussions in 2026. That pivot may be rational, but the abrupt strategic change and weak forecasting record argue for low credibility until stable cash flow is actually delivered.
Working alongside Starwood, we are progressing lease discussions across multiple sites. And we remain confident in our ability to sign at least 2 leases before year end.
Once completed, the Long Ridge acquisition is expected to contribute approximately $144 million in annualized EBITDA, and durable free cash flow with roughly 70% of its output secured under long term contracts.
Revenues during the second quarter of 26 were $175 million compared to $239 million in the prior year period.
We expect our quarterly G&A run rate excluding stock based compensation and acquisition and integration costs, to continue to trend lower as these savings are realized over time.
We expect FERC to respond to us, you know, definitely before year end. But, you know, much sooner than that.
MARA's edge is scale and power access, not technology. The company has built a large energized hash rate base at 70.3 EH/s and is assembling a power portfolio of about 4.8 GW, which can matter in a capital-intensive industry. The problem is durability: bitcoin mining is still close to a commodity business, and the AI infrastructure angle faces much stronger competitors with lower capital costs and established hyperscale relationships, so excess returns are hard to defend over time.
Capital allocation has turned aggressive. MARA is pursuing a $1.5B Long Ridge acquisition and added $600M of new borrowings at 7.56%, while FY2025 FCF was -$0.6B and there were no buybacks or dividends. The strategy may make sense if power assets are monetized into stable contracted cash flow, but today it looks debt-funded and execution-dependent rather than disciplined by current cash generation.
MARA rates a SELL. The company has real assets and real optionality, but the stock still prices too much future success into a business with negative free cash flow, worsening margins, and a weak forecasting record. The AI and power pivot may eventually work, yet right now it is an expensive promise funded with leverage, not a proven earnings stream. A more positive view would require evidence that Long Ridge closes on time, lease signings materialize, and contracted EBITDA begins to offset bitcoin volatility. Until then, downside toward the $6.6 base case remains the cleaner call.
≣ Full research note tap to expand
Executive Summary
SELL, Medium conviction. MARA rates a SELL at Medium conviction. The stock trades at $13.24, still -43.5% below its 52-week high, but that drawdown does not make it cheap when FY2025 showed $0.9B revenue, -85.8% operating margin, -46.7% net margin, and $-0.6B FCF while the company trades at 6.27x sales versus a 2.7x peer average. The forward story now depends on acquisitions, leases, and AI infrastructure monetization landing on time, yet management just posted a Q2 2026 net loss of $611M, pledged 54% of bitcoin holdings as collateral, and has missed earnings in 7 of the last 8 quarters.
Investment Thesis
MARA is still being valued like a scarce growth platform when the actual financials look like a stressed cyclical. FY2025 revenue grew 24.8% to $0.9B, but that growth came with a collapse in profitability: operating margin fell to -85.8% from 46.6% the year before, net margin was -46.7%, EPS was -$3.69, and free cash flow was -$0.6B. A company posting those numbers should not trade at 6.27x sales, especially against a peer average of 2.7x. The market is still paying for optionality that has not become earnings.
The current bull case rests on three things. First, scale in mining, with 70.3 EH/s energized, can support operating leverage if bitcoin economics improve. Second, the power portfolio, targeted at about 4.8 GW post transactions, could create a more durable infrastructure business than pure mining. Third, Long Ridge could bring $144M of annualized EBITDA with roughly 70% contracted output. Those are real positives, but they are not enough to underwrite today's valuation because each one is conditioned on execution, approvals, and financing. Long Ridge still needs FERC clearance, Matagorda's up to 2 GW is subject to ERCOT and interconnection approvals, and the company has already layered on $600M of additional borrowing at 7.56% while 54% of bitcoin holdings are pledged.
The other side of the ledger is harsher and more visible. Q2 2026 revenue fell to $175M from $239M as average bitcoin price dropped 28%, EBITDA was -$0.43B, adjusted EBITDA was -$361M, and net loss widened to $611M. This is not a one-quarter wobble from an otherwise dependable operator. MARA has missed consensus EPS in 7 of the last 8 quarters, including misses of -52.8%, -23.1%, -100.0%, and -1,060.4%. That record matters because the investment story now asks investors to trust future AI lease signings, lower G&A, hosted-cost rolloff, and acquisition synergies that have not yet shown up in reported numbers.
Price action does not fix that. The stock has rallied 47.77% in a month and sits 20.9% above its 200-day average, but it remains -43.5% below the 52-week high because the business model is still unstable. High short interest at 29.55% can fuel squeezes, not intrinsic value. The clean way to own this name would be after stable contracted EBITDA is visible and leverage is better covered by cash flow. At $13.24, that proof is missing.
Key Metrics
| Metric | Value | Context |
|---|---|---|
| Current price | $13.24 | Engine target implies -50.2% downside |
| Price target | $6.6 | Low confidence, basis sales_quality_blend |
| FY2025 revenue | $0.9B | +24.8% YoY, growth did not translate into profits |
| FY2025 EPS | -$3.69 | -149.5% YoY, deeply loss-making |
| FY2025 operating margin | -85.8% | Down from 46.6% in prior FY |
| FY2025 net margin | -46.7% | Loss profile remains severe |
| Valuation | P/S 6.27x, P/B 3.04x | P/S vs peer average 2.7x, P/E N/A vs peers 12.8x |
| Balance sheet | $2.5B debt, 1.49 debt/equity, 0.89 current ratio | Leverage is high for a volatile earnings model |
| Free cash flow | -$0.6B | Capital needs exceed internally generated cash |
| Q2 2026 performance | $175M revenue, -$0.43B EBITDA, -$1.60 EPS | EBITDA margin -245.8%, EPS miss -1060.4% |
| Quality and risk | Quality 29/100, Risk 66/100 | High risk, weak quality profile |
| 52-week position | $6.66-$23.45 | Current price is -43.5% below high, 20.9% above 200-day average |
Financial Strength
The balance sheet does not leave much room for mistakes. A 0.89 current ratio and $2.5B of debt would be manageable in a stable contracted infrastructure business, but MARA still generates volatile, bitcoin-linked earnings and posted $-0.6B of free cash flow. The company is trying to bridge into a more durable power-and-AI model, yet it is doing so with higher leverage, bitcoin-backed borrowings, and collateralized treasury exposure. Margin trajectory is moving the wrong way, from 46.6% operating margin in the prior FY to -85.8% in FY2025, which makes the capital structure look more aggressive than strategic.
Competitive Position
MARA's edge is scale and power access, not technology. The company has built a large energized hash rate base at 70.3 EH/s and is assembling a power portfolio of about 4.8 GW, which can matter in a capital-intensive industry. The problem is durability: bitcoin mining is still close to a commodity business, and the AI infrastructure angle faces much stronger competitors with lower capital costs and established hyperscale relationships, so excess returns are hard to defend over time.
Management & Guidance
Management is guiding to a larger and more diversified platform, not near-term earnings recovery. The key points are a roughly 4.8 GW power portfolio post pending transactions, at least 2 leases signed by year-end 2026, Long Ridge closing before year-end 2026, Long Ridge contributing about $144M annualized EBITDA, and quarterly G&A trending lower over time. Those targets could improve the mix of cash flows, but credibility is weak because the company has missed EPS expectations in 7 of 8 reported quarters and just delivered a quarter with -$0.43B EBITDA and -$611M net loss. Guidance is plausible, but not yet bankable.
Segment Analysis
- Bitcoin mining and digital asset revenue (Majority of revenue, exact split not disclosed in provided data): Core economics remain tied to bitcoin price and network difficulty, Q2 2026 revenue fell to $175M from $239M as average bitcoin price dropped 28%.
- Energy and infrastructure platform (Early-stage, not separately quantified in provided data): Long Ridge is expected to add about $144M annualized EBITDA with roughly 70% contracted output if closed before year-end 2026.
- Geography (Primarily U.S., no revenue geography split provided): Asset pipeline is concentrated in U.S. power markets including Ohio, Texas, and Long Ridge, with ERCOT and FERC approvals now material operating variables.
Capital Allocation
Capital allocation has turned aggressive. MARA is pursuing a $1.5B Long Ridge acquisition and added $600M of new borrowings at 7.56%, while FY2025 FCF was -$0.6B and there were no buybacks or dividends. The strategy may make sense if power assets are monetized into stable contracted cash flow, but today it looks debt-funded and execution-dependent rather than disciplined by current cash generation.
Management Execution & Track Record
Execution has been mixed operationally and poor financially. Management said in 2024 Q3, "So, we have hit more than 40 exahash already and we are driving towards the 50 exahash target," and the company has since exceeded that level with 70.3 EH/s energized in Q2 2026, so the hash-rate buildout was delivered. But the economic result missed badly: 2024 Q3 EBITDA margin was 135.1% on $0.25B revenue, versus -245.8% EBITDA margin on $0.17B revenue in Q2 2026, and the company has missed EPS estimates in 7 of the last 8 quarters, including -1,060.4% in August 2026. Management also said in 2024 Q3, "These three facilities have a combined interconnect-approved capacity of 372 megawatts, which MARA intends to fully energize by the end of 2025," but the provided data does not confirm that milestone directly, while strategy has shifted materially from dismissing AI hosting as unattractive to actively pushing AI infrastructure and lease discussions in 2026. That pivot may be rational, but the abrupt strategic change and weak forecasting record argue for low credibility until stable cash flow is actually delivered.
Positioning & Flows
Ownership is institutionally concentrated, with the top three holders at 53.2% and insiders at only 1.7%, which limits the alignment argument. Short interest is very high at 29.55%, or 110.1M shares, but days-to-cover is only 2.4, so squeeze risk is real but can also reverse quickly once momentum fades. Options positioning is listed as a positive gamma regime, which can dampen volatility somewhat near term, but it does not change the weak fundamental setup.
Bull Case
- Mining scale is still growing, energized hash rate reached 70.3 EH/s in Q2 2026, up 22% Y/Y, versus the 2024 Q3 target framework that had aimed for 50 EH/s.
- Unit economics showed one real operating improvement, daily cost per petahash improved 4% Y/Y to $27.70, and hosted mining costs should ease as contracts begin expiring from Q3 2027 through Q1 2028.
- The power platform could become more valuable if execution holds, with a post-transaction portfolio of about 4.8 GW, Matagorda adding up to 2 GW potential capacity, and Long Ridge expected to contribute about $144M annualized EBITDA with roughly 70% contracted output.
- Positioning can amplify upside in squeezes, short interest is 29.55% of float with 110.1M shares short, while the stock is 20.9% above its 200-day average after a 47.77% 1-month rally.
Bear Case
- The income statement is broken at current bitcoin and mark-to-market conditions, FY2025 EPS was -$3.69, Q2 2026 EPS was -$1.60 versus $0.1666 expected, and adjusted EBITDA was -$361M.
- Leverage has risen into a cyclical business, including $600M of new Coinbase and Two Prime borrowings at a 7.56% weighted average cost and a $1.5B Long Ridge acquisition funded with bitcoin-backed debt and assumed debt.
- Liquidity is thin for a business pursuing expansion, current ratio is just 0.89, total debt is $2.5B, and roughly 54% of bitcoin holdings are pledged as collateral.
- The stock still carries a premium multiple without premium quality, with a 29/100 quality score, 66/100 risk score, and repeated execution misses despite the pivot toward AI infrastructure.
Valuation & Price Target
Engine price target $6.6 (-50.2% vs current), Low confidence.
- P/S Multiple: $4.64 (45% weight)
- Quality-adjusted: $11.02 (20% weight)
This view aligns with the quant baseline of SELL, though conviction is slightly higher than the engine's low-confidence call because the fundamental evidence is consistently negative. The engine target is $6.6, based on a sales-quality blend between $4.64 on P/S and $11.02 on quality adjustment, and the stock still trades at 6.27x sales despite -85.8% operating margin, -$3.69 FY2025 EPS, $-0.6B FCF, and $2.5B debt. The only credible reason to fight the quant call is the optionality in the power and AI pivot, but that upside is not yet matched by delivered earnings or balance-sheet strength. Sell-side consensus leans Buy, but with only 6 firms and a LOW signal, it does not outweigh MARA's own numbers.
Risk Assessment
Risk score 66/100 (High). The biggest risk is that leverage meets prolonged weak bitcoin economics, because MARA already has $2.5B debt, 54% of bitcoin collateral pledged, and a Q2 2026 net loss of $611M.
The Bottom Line
MARA rates a SELL. The company has real assets and real optionality, but the stock still prices too much future success into a business with negative free cash flow, worsening margins, and a weak forecasting record. The AI and power pivot may eventually work, yet right now it is an expensive promise funded with leverage, not a proven earnings stream. A more positive view would require evidence that Long Ridge closes on time, lease signings materialize, and contracted EBITDA begins to offset bitcoin volatility. Until then, downside toward the $6.6 base case remains the cleaner call.
Go deeper in the app
Live charts, options analytics, the full earnings-call breakdown, price alerts, and every other stock — updated in real time.
Get Tickr