Crypto & Digital Assets
A best-in-class US bitcoin miner that has rebuilt itself into a 0%-convert-funded, share-shrinking BTC-beta vehicle now sprinting late into the AI/HPC pivot — the equity is a leveraged call on bitcoin and on executing a datacenter business that today earns exactly $0.
Research
The verdict
A best-in-class US bitcoin miner that has rebuilt itself into a 0%-convert-funded, share-shrinking BTC-beta vehicle now sprinting late into the AI/HPC pivot — the equity is a leveraged call on bitcoin and on executing a datacenter business that today earns exactly $0.
CleanSpark is a US-domiciled (Nevada-incorporated, Henderson NV HQ) bitcoin miner that, as of FY2025, recast its self-description from "bitcoin miner" to "a data center developer, until recently focused exclusively on bitcoin mining". That sentence is the whole thesis in miniature — the company is mid-metamorphosis.
How it makes money (today): It owns/leases/operates a fleet of ASIC miners across Georgia, Tennessee, Mississippi and Wyoming, contributes 100% of its hashrate to a single mining pool (Foundry Digital), and earns daily bitcoin rewards under a Full-Pay-Per-Share ("FPPS") formula. Revenue is recognized as non-cash bitcoin consideration; the company then sells, holds, or pledges that bitcoin. 100% of revenue flows through Foundry — a single counterparty under a contract terminable by either party at any time.
Scale: 45.6 EH/s average operating hashrate at FY2025 year-end (peak 50 EH/s — "first US public company to hit 50 EH/s exclusively on American infrastructure" ), rising to 47.3 EH/s as of 2026-03-31. ~224,473 miners in service (of ~326,885 owned) at 16.2 J/TH average efficiency.
Contracted power: ~1,027 MW at FY2025 close, expanded to ~1,809 MW as of 2026-03-31 via Texas (Austin County 285 MW; Brazoria County ~300 MW, expandable to ~600 MW), Tennessee, and South Dakota acquisitions.
The pivot (tomorrow): AI & HPC hosting/leasing. As of 2026-03-31 this business has earned zero revenue — "we had earned no revenues from our AI and HPC services business". It is staffed (Jeffrey Thomas hired Oct 2025 as SVP AI Data Centers; Submer partnership for modular liquid cooling ) and funded, but pre-product.
Key payment-term flags: Revenue is single-pool concentrated (take-it-or-leave-it, no recurring contracts); the AI thesis depends on signing hyperscaler/colo leases that do not yet exist.
Upstream → CleanSpark → end-market, every named link:
Chokepoints: (1) Bitmain single-source ASIC dependency; (2) Foundry single-customer dependency; (3) bitcoin price as the only revenue driver. Three single points of failure stacked.
CleanSpark's moat is operational, not structural — and that is the central problem for a long-term holder.
Verdict on moat: thin and commoditized in mining; the only credible durable edge is the power/land portfolio, whose value is fully contingent on winning the AI-hosting transition that every peer is also chasing.
CleanSpark is effectively a single-segment business. Per the 10-Q, the CODM reviews the company as one operating segment; bitcoin mining revenue = 100% of total revenue in every period. segments.csv is empty because there is nothing to disaggregate yet.
By geography: all mining is at owned/leased US facilities (GA/TN/MS/WY) after the company exited all hosted-facility mining (Massena NY 50 MW hosting expired 2024-12-31; all hosted MW vacated by 2025-09-30).
Revenue trend (decelerating, then declining):
The future "segment" — AI/HPC — is $0 today. When (if) it generates revenue, this becomes the segment story; for now there is nothing to break out.
This is the most important lens for CleanSpark right now, because the latest print is brutal on GAAP and the reason is structural to the business model.
Headline:
What actually drove the loss — the ASC 350-60 mark: The net loss is dominated by a −$224.1M unrealized loss on the fair value of bitcoin plus a −$38.8M loss on bitcoin collateral, both because bitcoin fell from ~$87,500 (Dec 31) to ~$68,200 (Mar 31) during the quarter. Since Oct 2023, CleanSpark marks its entire bitcoin holding to market through the income statement every quarter (ASC 350-60). So the P&L is now a bitcoin-price derivative: BTC up = giant "earnings," BTC down = giant "loss." FY2025's $364.5M net income was, symmetrically, flattered by a +$425.6M fair-value gain.
Operating economics underneath the mark (the part that matters): Strip the FV noise and the operating business deteriorated badly:
Guidance: CleanSpark does not give formal revenue/EPS guidance (guidance.csv empty); it guides on hashrate/power capacity. Management framed the quarter around the balance sheet ("one of the strongest balance sheets in our sector") and the AI pivot, not the loss.
Balance-sheet flags: Cash up to $260.3M (from $43.0M) — funded by the $1.15B convert. Total debt now ~$1.79B long-term (up from $645M at Sept 30). Accumulated deficit blew out to −$913M (from −$126M). Total stockholders' equity halved to $986M (from $2.18B). The half-year burned through a third of the company's book equity, almost entirely via the bitcoin mark.
Market reaction: the stock has actually risen into mid-2026 (~$17, +8% on the print day per one report ) — the market is correctly looking through the GAAP mark to bitcoin and the AI optionality, not the reported loss. That tells you what CLSK trades on: BTC + AI narrative, not earnings.
No transcripts are on disk (transcripts/ empty), so this is web-grounded and necessarily lighter.
The dominant management narrative has shifted hard over the last ~4 quarters, from "pure-play US bitcoin miner / efficiency leader" toward "multi-gigawatt AI infrastructure platform". CEO Matt Schultz's recurring talking points are now: (1) "demand for scaled, AI-native compute continues to accelerate"; (2) transmission-level power access is "increasingly constrained" (i.e., our power portfolio is the prize); (3) balance-sheet strength. The phrase that has disappeared is "exclusively focused on bitcoin mining" — the 10-K literally rewrote the opening sentence to walk it back.
Tone read: promotional and pivot-forward, leaning on balance-sheet and power-portfolio talking points while deflecting from the deteriorating mining unit economics and the GAAP losses. The "improved profitability outlook" framing some outlets repeated is not supported by the actual Q2 numbers — it is bitcoin-recovery hope, not operating improvement.
Peer set = CleanSpark's own self-constructed peer group from the 10-K: MARA, RIOT, HIVE, BITF, WULF, CIFR, IREN, HUT — plus Bitdeer.
| Company | Ticker | USD mkt cap | Hashrate (EH/s) | EV/Sales | EV/EBITDA | P/E | Div yld | 5Y avg ROE |
|---|---|---|---|---|---|---|---|---|
| CleanSpark | CLSK | ~$4.83B | 47.3 | n/a | n/a (negative LTM EBITDA) | n/a — negative LTM EPS | 0% | n/a — net-loss years dominate |
| MARA Holdings | MARA | n/a | 61.7 | n/a | ~9x | n/a | 0% | n/a |
| Riot Platforms | RIOT | n/a | 36.4 | n/a | n/a | n/a | 0% | n/a |
| IREN (Iris Energy) | IREN | ~$18B | 43 | n/a | "cheaper group" | n/a | 0% | n/a |
| Bitdeer | BTDR | n/a | 69.5 | n/a | n/a | n/a | 0% | n/a |
| HIVE / BITF / WULF / CIFR / HUT | — | n/a | n/a | n/a | n/a | n/a | 0% | n/a |
Pattern (web-grounded, cross-checked to filings):
Read: the market reacts to bitcoin and the AI-pivot narrative, and largely ignores the GAAP loss line. CLSK is not traded on earnings; it is traded on BTC beta plus datacenter optionality.
A founder-led company that just changed its founder-CEO. This is the single most important governance fact in the file.
Archetype: founder-operators in capital-allocator mode — competent builders making increasingly financial bets, with a control structure that insulates them from common shareholders.
Acting as a forensic equity analyst.
1. Fair-value-of-bitcoin earnings (the biggest "quality of earnings" issue — but it's GAAP-mandated, not a manipulation). Under ASC 350-60, the entire bitcoin holding (11,920 BTC, fair value $813.2M at 2026-03-31 ) is marked to market through net income every quarter. This makes "net income" almost meaningless as an operating signal — FY2025's +$425.6M FV gain and Q2 FY2026's −$224.1M FV loss are price moves, not performance. Watch: management's Adjusted EBITDA excludes the FV mark — but the non-GAAP reconciliation explicitly notes it does not strip the FV loss/collateral loss, so even Adj EBITDA was −$241.2M. Cash flow vs. earnings: revenue is non-cash (paid in BTC), so reported revenue ≠ cash; the company must sell BTC or draw credit to fund operations.
2. Bitcoin-collateral derecognition + rehypothecation. Bitcoin pledged to Coinbase/Two Prime is derecognized and replaced by a "Receivable from bitcoin collateral" that the lenders can sell, pledge, and rehypothecate. At 2026-03-31, $111.9M sat in this receivable. This is counterparty/credit risk to crypto lenders dressed as a receivable — and a −$142.5M six-month "loss on bitcoin collateral" ran through other income as BTC fell.
3. Prior-year ICFR MATERIAL WEAKNESSES (now remediated). As of 2025-09-30 management concluded ICFR was effective, but explicitly disclosed it had remediated two material weaknesses that existed at FY2024: (#1) IT general controls — change management, logical access, segregation of duties over the general ledger, affecting essentially every material line item; (#2) controls over PP&E and deposits on miners. A company that had pervasive ICFR weaknesses through FY2024 warrants ongoing scrutiny that remediation holds, especially given the balance-sheet complexity (converts, capped calls, BTC derivatives, collateral).
4. Tariff contingency — up to $130M, unreserved. Since May 2025, US CBP has asserted Chinese-origin import tariffs on miners imported in 2024; potential liability up to ~$130M plus statutory interest, with no provision recorded (management deems loss "not probable"). Given Bitmain is 72% of miner supply, this is a live, material, off-balance-sheet exposure.
5. SBC ramp + valuation allowance. Stock-based comp jumped to $22.2M for H1 FY2026 (from $6.1M) under the new 2025 LTIP — non-cash, but real dilution pressure partly offsetting the buyback. The company carries a full valuation allowance on US deferred tax assets (increased $127.8M in H1 on the BTC mark) — i.e., the accounting confirms it does not expect to use those losses against future taxable income soon.
6. Goodwill $131.7M from the GRIID acquisition (all-stock, $128.2M, closed Oct 2024 for TN power) — not yet impaired, but a watch item if the AI/TN buildout disappoints.
Regulatory findings (required sub-section):
Why this is unusually hard for CLSK: GAAP EPS is dominated by the un-forecastable bitcoin FV mark. Consensus EPS figures are therefore close to noise — analysts model ~$0.73 FY2026 and ~$0.32 FY2027, but the FY2027 range is −$2.71 to +$2.65, which simply restates "we are guessing the bitcoin price." (The aggregated "$277B revenue" consensus figure is a clear data-vendor error — flagged, not used.) I therefore model mining gross profit / cash economics, not GAAP EPS, and treat any FV gain/loss as a separate BTC-price overlay.
Anchor actuals: FY2025 revenue $766.3M; Q2 FY2026 revenue $136.4M (≈$546M annualized run-rate at the Q2 BTC price); cash cost to mine ~$45.4K/BTC; ~20-21 BTC/day; 47.3 EH/s rising; bitcoin ~$66.4K at the time of writing.
Base case (BTC ~$70K avg, hashrate share holds ~4.5%, AI revenue still ~$0 in FY2026, first small AI contribution FY2027–28):
Bull (BTC $100K+, AI leases signed at multiple sites): revenue re-accelerates and, more importantly, the multiple re-rates toward IREN's AI-inclusive valuation — the equity upside is the re-rate, not the EPS.
Bear (BTC $50–60K sustained, AI pivot slips): mining runs at/under full-cost breakeven, the $1.8B converts loom (first 2030-note put June 2028), equity keeps absorbing FV losses, and the stock de-rates to a distressed-miner multiple.
The honest one-liner: FY2026-28 "EPS" is unknowable because it's a bitcoin-price forecast in a trench coat. The forecastable variables are hashrate (rising), cost/BTC (~$45K cash), and whether AI revenue ever shows up (zero so far).
Per --watchlist rules, no Brier forecast was logged (forecast.ts create skipped). A reasonable scoreable line for a future /thesis pass: "CLSK FY2027 (ending 2026-09-30 → use FY2028 ending 2027-09-30) generates >$0 of AI/HPC hosting revenue, p≈0.55."
Bull case. CleanSpark is a top-3 US bitcoin miner (47.3 EH/s, best-in-class 16.2 J/TH efficiency) that has quietly assembled the scarce asset of the AI era — ~1.8 GW of contracted, grid-connected, mostly-US power, including fresh ERCOT transmission-level capacity — and funded it with $1.8B of 0%-coupon converts while shrinking its share count. If even a fraction of that power portfolio converts to AI/HPC leases, the equity re-rates from a ~$4.8B "miner" multiple toward IREN's ~$18B AI-inclusive multiple on less hashrate. Meanwhile the 11,920-BTC treasury is a free call on bitcoin, and a BTC recovery to six figures swings GAAP back to large reported "profits." Strong-Buy consensus, median PT $19 (+11% from ~$17), range to $26.
Bear case (2–3 things that permanently impair). (1) The AI pivot is late and unproven — IREN, Hut 8, Cipher, RIOT, MARA are all chasing the identical "power → AI" re-rate; CleanSpark has $0 AI revenue, no signed hyperscaler lease, and a brand as a miner. If hyperscalers favor incumbents with track records, CleanSpark's power becomes merch-mining capacity, not AI gold. (2) Mining economics are structurally deteriorating — full-cost-to-mine hit 136% of revenue in Q2; share of global hashrate is falling despite capex; post-halving block rewards plus rising network difficulty squeeze the core business every quarter BTC doesn't rally. (3) $1.8B of converts + $130M tariff exposure + a certified securities class action — the leverage and contingencies that look benign at BTC $70K become acute at BTC $50K, with the first 2030-note put in June 2028.
Pre-mortem (it's Dec 2027 and the thesis broke): Bitcoin chopped around $55–65K, so the BTC-beta trade never paid; the AI buildout slipped (permitting, GPU access, no anchor tenant) so the re-rate never came; full-cost mining losses + SBC + the converts steadily eroded equity; and the Smith/Hasthantra litigation produced a settlement and a governance overhang. The stock is a sub-$10 distressed miner and the 0% converts that looked clever now cap the upside (capped calls) without protecting the downside.
Are multiples too high? On mining fundamentals, yes — CLSK is in the "expensive" EV/EBITDA cluster for a business mining at full-cost-breakeven. The valuation only makes sense as an option on the AI pivot + BTC upside, not as a cash-flow multiple.
Contrarian view (what the market refuses to see): The bullish framing treats the $1.8B of 0% converts as "cheap capital / smart balance sheet." The contrarian read is that CleanSpark has quietly become a leveraged bitcoin-and-AI bet: it borrowed $1.8B to buy back stock and hold BTC, halving its book equity in six months via the mark. That's MicroStrategy-style financialization grafted onto a commodity miner — brilliant in a bull market, a doom-loop (forced BTC sales, convert refinancing into a closed window) in a bear one. The market is pricing the optionality and ignoring the leverage.
Dismantling the bull case.
A levered, sub-NAV Ethereum proxy run by the best operators in the trade — own the discount-closing buyback, not the "treasury premium" that already died; the bet is ETH plus a 0.83x→1.0x mNAV mean-reversion, fighting a 95% drawdown and a structurally bigger rival.
A Bitcoin miner that won the AI-infrastructure lottery — NVIDIA both rents IREN's GPUs and bought its equity at $70, validating the liquid-cooling pivot Culper shorted; the bet is now execution and dilution, not survival.
A 25 EH/s green Bitcoin miner that liquidated its entire treasury to fund a credible-but-unfunded pivot to AI compute — the stock now trades on a CAD $3.5B Gigafactory dream that $23M of cash cannot pay for.