MARA Holdings, Inc. (MARA)

Q1'26 revenue $174.6M, -18.4% YoY (lower BTC price/production); FY2025 revenue $907.1M, +38% YoY. FY2025 net loss $1.3B vs FY2024 net income $541.0M, the entire swing traces to a $1.5B negative change in digital-asset fair value (FY2025) vs a large gain in FY2024; GAAP 'earnings' here are a BTC mark-to-market artifact, not an operating signal (every $10K BTC move ≈ ~$350-450M swing per company disclosure).

Total debt fell from $3,599.9M to $2,416.2M (net of OID) in one quarter via BTC-funded buybacks, not earnings (10-Q Note 13). All five convertible tranches (conversion prices $18.89-$76.17) are out-of-the-money at $11.75, near-term conversion dilution is low, though a $150M line of credit and $48.1M of Dec-2026 notes mature soon. Cash $513.7M; BTC 35,303 (10-Q, 2026-03-31) / 36,303 (bitcointreasuries.net, 2026-06-16) worth ~$2.3B. Shares outstanding still +17% YoY (stockanalysis.com) reflecting the historical ATM-funded-BTC-purchase playbook, though management states the ATM has been unused since end-2025 in favor of BTC monetization.

Catalysts. Long Ridge close (H2 2026, +65% owned power capacity, $144M disclosed annualized adj. EBITDA); first signed AI/HPC tenant lease. Risks, hash price at a 5-year low (CoinShares Q1'26); BTC re-test of bear-market lows; resumed ATM dilution; guidance-credibility gap. Q9 segments: strongest = mining block rewards $156.25M (89.5% of Q1'26 revenue, still -20.4% YoY); weakest = Other/AI revenue $951K (-66% YoY), the AI pivot has zero income-statement traction yet.

Fear vs. Hope

No the automated screen verdict exists to anchor this against, so both sides: FEAR = another BTC bear leg combined with hash-price collapse (5-year low per CoinShares) renders mining thin-to-negative margin while $2.4B of debt remains, forcing a return to dilutive ATM issuance. HOPE = the AI/HPC pivot (Long Ridge, Starwood JV, Exaion) re-rates MARA toward Riot's richer ~17x NTM EV/Revenue multiple while BTC stabilizes or recovers.

Is It Priced Correctly?

Quantified via scenario decomposition: bear ~$5/share (-57% from $11.75) if BTC falls to $38,000 and the market cuts the ops-value credit to 0.6x; bull ~$23/share (+96%) if BTC reaches $110,000 and the market re-rates ops value to 1.6x (Riot-like). Base case (~$11.79) sits almost exactly at the current price, the stock is not obviously mispriced given today's BTC price and today's market credit for the AI pivot, but that credit itself could be wrong in either direction.

Inflection Signals

Hash rate +33% YoY to 72.2 EH/s (Q1'26) is a genuine positive inflection in mining capacity, but it comes after MARA missed its own mid-2025 guidance of 75 EH/s by year-end 2025 (actual: 53.2 EH/s), a credibility gap trailing data does not resolve. Separately, the ~30% one-quarter reduction in convertible debt (10-Q Note 13) is a real, verified balance-sheet inflection, achieved via opportunistic BTC-funded buybacks at a 9% discount, not via operating cash flow.

Reverse-DCF Reality Check

Not meaningful in standard DCF form (no the automated screen g_implied exists; MARA has no stable FCF to extrapolate). Reframed: the bull-case BTC price ($110,000) is not unprecedented. BTC traded above that level for much of 2025, reaching an ATH of $126,210.50 on 2025-10-06, so the bull case requires only a partial recovery, not a new record. No public miner, including MARA, has sustained hash-rate-share growth for 5 years without heavy share dilution, since global network difficulty typically grows faster than any single miner's absolute buildout pace.

Unpriced Optionality

The AI/HPC infrastructure pivot (Long Ridge $1.5B acquisition targeting 200-600MW, Starwood JV targeting >1GW, 64% Exaion stake) is real and anchored in signed agreements, but produces essentially zero income-statement revenue today (Other revenue $951K, -66% YoY) and no AI tenant leases are targeted before 2026-2028. A mechanical model built only from trailing mining financials would likely price this near zero; our research estimates it qualitatively via the ~$4.1B implied ops-value plug, but this is UNPRICED with any precision from filings alone.

Disconfirming Evidence

• MARA badly missed its own 2025 hash-rate guidance: it targeted 75 EH/s by year-end 2025 (announced mid-2025) but actually ended 2025 at 53.2 EH/s, a ~29% shortfall, an execution/credibility gap that undercuts confidence in forward guidance generally, including the AI/HPC pivot timeline.
• Sector economics are genuinely stressed: CoinShares' Q1 2026 report shows hash price collapsed to $28-30/PH/s/day, a five-year post-halving low, with the weighted-average public-miner cash cost to mine 1 BTC at ~$79,995 (Q4 2025) versus spot BTC of ~$62,000-63,600 (2026-07-07/08), implying the average public miner is mining near or below cash cost today, and ~15-20% of legacy machines globally are cash-flow negative.
• The historical dilution playbook is not fully retired: shares outstanding still grew ~17% YoY (stockanalysis.com) even as management says the ATM has been unused since end-2025; that policy is only ~2 quarters old and could reverse if BTC/hash prices stay depressed and the balance sheet needs refilling, since MARA has financed itself primarily via equity/converts for most of its history.

Falsification Condition

Bearish break: BTC trades sustained below ~$45,000 for a full quarter (pushes treasury value net of $1.9B net debt toward zero per the bear scenario) and/or hash price stays below $30/PH/s/day for two more quarters, forcing renewed ATM issuance. Bullish break (currently-undervalued signal): the Long Ridge acquisition closes on schedule (H2 2026) AND a signed AI/HPC tenant lease is disclosed before 2027, which would support re-rating the implied ops-value multiple toward Riot's ~17x NTM EV/Revenue from MARA's current ~9x.

Catalyst Horizon

Near-term (weeks): Q2 2026 earnings, next reporting date 2026-07-28, will show whether hash-rate growth and BTC-monetization pace continued. Medium-term (6-18 months): Long Ridge acquisition close (H2 2026) and any first disclosed AI/HPC tenant lease are the clearest re-rating catalysts. Unscheduled and dominant: the BTC price itself (macro liquidity, ETF flows, risk sentiment), no identifiable catalyst controls this variable, which drives most of the valuation range.

Segments

Strongest (by size): Mining operator - block rewards, $156.25M in Q1'26, 89.5% of total revenue, though -20.4% YoY on lower average realized BTC price and softer production (10-Q). Fastest-growing sub-segment: Mining participant revenue $15.32M, +42% YoY (pool/hash-rate-service fees), still only 8.8% of revenue. Weakest: Other revenue (includes Exaion/AI contribution) $951K, -66% YoY, and Hosting services $1.085M, roughly flat, the AI/HPC pivot has essentially zero income-statement traction as of Q1 2026.

KPI Pack

• realized_cost_vs_benchmark: Purchased energy cost per bitcoin $40,047 (Q1'26, company disclosure) vs spot BTC ~$62,000-68,000 during the quarter, energy-only cost still well below BTC price. Vs the industry-wide weighted-average public-miner cash cost of ~$79,995/BTC (Q4 2025, CoinShares), MARA's disclosed energy-cost metric looks more efficient than the sector average, though the two figures are not perfectly comparable (energy-only vs full cash cost).
• production_growth: BTC mined: 2,247 in Q1'26 vs 2,011 in Q4'25 (+11.7% QoQ, 10-Q/shareholder letter). Precise Q1'25 BTC-mined units UNAVAILABLE in sources reviewed; hash-rate growth (+33% YoY to 72.2 EH/s) is the disclosed proxy for output growth.
• hash_rate_vs_peers: MARA 72.2 EH/s (Q1'26, +33% YoY) vs Riot 42.5 EH/s (Mar 2026) vs CleanSpark 50.0 EH/s (Mar 2026). MARA remains the largest hash-rate US public miner, though it missed its own 75 EH/s year-end-2025 target (actual 53.2 EH/s).
• treasury_coverage_of_debt: 36,303 BTC (~$2.3B, bitcointreasuries.net 2026-06-16/2026-07-08) vs net debt ~$1.90B (10-Q, 2026-03-31), treasury alone covers ~121% of net debt at spot BTC, but a BTC decline to the bear-case $38,000 would push this cushion to roughly zero. No clean 'reserve life' equivalent exists for a non-extractive business; UNAVAILABLE as a multi-year metric.

• BTC holdings 35,303 (fair value $2,408.9M) at 2026-03-31 (10-Q Note 5), down from 53,822 BTC ($4,711.7M) at 2025-12-31. MARA sold ~15,133 BTC (~$1.1B) between 2026-03-04 and 2026-03-25 to fund a $1.0B convertible-note buyback at a 9% discount, cutting convertible principal ~30% (8-K, 2026-03-26). Most recent tracked holdings: 36,303 BTC, ~$2.3B (bitcointreasuries.net, count as of 2026-06-16, value as of 2026-07-08).
• Energized hash rate 72.2 EH/s in Q1 2026, +33% YoY (10-Q/shareholder letter), but MARA missed its own mid-2025 guidance of 75 EH/s by year-end 2025 (actual: 53.2 EH/s), a ~29% shortfall.
• Q1 2026 revenue $174.6M, -18.4% YoY; FY2025 revenue $907.1M, +38% YoY vs FY2024 $656.4M (10-K/10-Q). TTM revenue computed at $867.8M, matching stockanalysis.com's $867.82M.
• FY2025 net loss $1.3B vs FY2024 net income $541.0M; Q1 2026 net loss $1,259.6M attributable to common stockholders ($3.31/share) on a $714.7M non-cash change-in-fair-value-of-digital-assets loss. GAAP earnings are dominated by BTC mark-to-market, not operations (10-Q).
• Total debt fell from $3,599.9M (net of OID, 2025-12-31) to $2,416.2M (2026-03-31) via BTC-funded buybacks. All five convertible tranches (conversion prices $18.89-$76.17) are out-of-the-money at the $11.75 stock price (10-Q Note 13).
• Purchased energy cost per bitcoin $40,047 in Q1 2026; cost per kWh $0.04 at owned sites (both FY2025 and FY2024); energy cost was 38.5% of owned-mining revenue in FY2025 vs 40.8% in FY2024 (10-K).
• Industry-wide: CoinShares' Q1 2026 report shows hash price collapsed to $28-30/PH/s/day, a five-year post-halving low, with weighted-average public-miner cash cost to mine 1 BTC at ~$79,995 (Q4 2025), roughly 25-30% above spot BTC (~$62,000-63,600, 2026-07-07/08); ~15-20% of legacy mining machines globally are cash-flow negative.
• $1.5B agreement to acquire Long Ridge Energy & Power (505MW gas plant, 1,600 acres, Ohio), expected to close H2 2026, adding ~65% owned power capacity and a disclosed $144M in annualized adjusted EBITDA, targeting 200-600MW of AI/HPC buildout by 2028 (8-K, 2026-04-30).
• Shares outstanding 380,873,087 at 2026-03-31 (10-Q), +17.02% YoY (stockanalysis.com); management states the ATM equity program has been unused since end of 2025, funding operations via BTC monetization instead.
• MARA (72.2 EH/s, ~9x NTM EV/Revenue) vs Riot Platforms (42.5 EH/s, ~17x NTM EV/Revenue, market cap $6.29B) vs CleanSpark (50.0 EH/s, ~9x, EV ~$4.6B). MARA is the largest hash-rate miner but trades at CleanSpark's multiple despite a larger AI-infrastructure pipeline than either peer's.