CleanSpark, Inc. (NASDAQ: CLSK) — The Best-Run Miner, Betting the Company on a Lease It Hasn’t Signed

Independent equity-research note. Report date: 2026-06-20. General information and analysis, not investment advice.

Price at analysis: $17.24 (2026-06-18) · Shares out: ~256.6M · Market cap: ~$4.42B · Enterprise value (less BTC + cash): ~$5.0B · FY end: Sept 30 · 52-week range (cycle): ~$6.72–$23.20 · Beta: ~2.95.

⚡ Claude’s Take

This block is the author’s own independent opinion and general information only — not investment advice. The analysis that follows takes no position and contains no price target; it discusses valuation only as embedded expectations and scenarios.

Call: HOLD / not-a-short / accumulate-on-weakness only into the low-$10s. Fair-value zone ~$11–16 on the mining business plus a probability-weighted HPC option; you only “back up the truck” on a BTC-led flush toward tangible-book (~$3–4) plus a still-unsigned lease, which rarely coincide. Conviction: medium-low.

CleanSpark is the best-operated pure-play Bitcoin miner in the public market — 50 EH/s at a top-quartile ~16 J/TH, ~100% owned-and-operated infrastructure, industry-leading uptime, and the most disciplined cost culture in the cohort. It is also, like every one of its peers, a moatless commodity price-taker in the bust phase of the post-halving capital cycle (hashprice at a five-year low, network difficulty just cut ~10%, ~20% of miners underwater), whose GAAP earnings are now a pure leveraged proxy for the Bitcoin price under fair-value accounting (a ~$378M Q2-FY26 net loss driven by a ~$224M non-cash BTC mark). The entire bull case rests on one pivot: converting its 1.8 GW of contracted power in Georgia, Texas and Wyoming into signed hyperscaler/AI leases. The problem is timing and proof — CleanSpark has zero signed HPC leases, only a “lead prospective tenant” at its 250 MW Sandersville, GA site, while IREN (a $9.7B Microsoft deal), RIOT (AMD), WULF, CIFR and Core Scientific have already banked the leases that re-rated their stocks. CleanSpark is the cheapest of the group on power it hasn’t yet monetized — which is the same statement as “the one the market is least willing to pay the HPC multiple for, because it hasn’t earned it.”

The framing is “a quality operator of a bad business, mid-leap to a better one, priced as a levered Bitcoin call with a free — but unsigned — AI option.” Two things keep me from a constructive call at $17. First, the balance sheet flipped: in ~14 months CleanSpark went from net-cash and all-equity-funded to ~$1.79B of 0%-coupon convertible notes (strikes $14.80 and $19.16, i.e., dilution deferred, not avoided), and BTC + cash ($1.18B) now sits below the debt — so the equity is not asset-backed (net liquid ≈ −$0.6B), unlike MARA. Second, governance is a yellow flag: an abrupt founder-CEO swap (Bradford out in days, ~$48.8M exit), ~$94M of combined CEO pay in a negative-operating-income year, comp keyed to EH/s/MW/TSR with no ROIC or per-share metric, a soft 76% say-on-pay, a 45-million-vote founder preferred block, and one open-market insider buy in five years. Flips bullish on a signed, investment-grade HPC lease at Sandersville (or the portfolio) at credible $/MW economics — that single event closes the valuation gap to the re-rated peers. Flips bearish if BTC breaks toward ~$45k (mining cash margins go negative and the converts’ 2028 put looms over a non-asset-backed equity) or the Sandersville tenant talks die with no replacement. Best operator on a sinking street — and they’ve mortgaged the house to buy lottery tickets on the new one.

📈 Stock Price Action — Five-Year Event Map

CleanSpark is a high-beta (~2.95) levered Bitcoin proxy that has round-tripped violently with each crypto cycle. From a sub-$2 Bitcoin-winter low (late 2022) it ran ~13x to a ~$23–24 cycle peak (early 2024 and again late 2025), and sits at $17.24 today — roughly −26% off the November-2025 cycle high (~$23.20) but +93% over the trailing twelve months and +110%+ off its January-2026 low (~$8.18). The price has no anchor other than the Bitcoin price, the network-difficulty ratchet, and — newly — the market’s appetite to pay for an unproven AI/HPC pivot. (Price moves are FACT; attributed drivers are INTERPRETATION.)

Cycle narrative. (1) CleanSpark bottomed near $2 in the 2022 winter, a leveraged victim of the BTC drawdown. (2) The 2023–24 spot-ETF approval and pre-halving run to a ~$73k BTC ATH drove an ~11x melt-up as CleanSpark aggressively energized hashrate. (3) The April-2024 halving halved block rewards into a rising-difficulty network, and the stock more than halved by autumn even as the all-stock GRIID deal closed. (4) A fresh BTC bull to ~$126k, the 50 EH/s milestone, two large convertible raises, and management’s AI/HPC framing drove a ~3.5x recovery to a ~$23 cycle peak in November 2025. (5) BTC’s winter-2026 crash toward ~$76k and the absence of any signed HPC lease knocked the stock back to ~$8. (6) A BTC stabilization and renewed bid for the AI option (helped by a Chardan price-target raise to $19 on Jun-8-2026) lifted it back to ~$17. Every leg is a Bitcoin-beta move first; the HPC pivot has so far added volatility and a narrative, not a durable re-rating.

1. Executive Summary

CleanSpark is the largest domestic (US) producer of Bitcoin hashrate and the #2 US-listed miner overall (~50 EH/s vs MARA’s ~72), distinguished within a structurally poor industry by genuine operational excellence: a ~16 J/TH fleet (top-quartile efficiency), ~100% owned-and-operated data centers (versus the hosted/JV models of peers), industry-leading uptime, and the most disciplined cost and capital-deployment culture in the cohort. In FY2025 (ended Sept-30-2025) revenue more than doubled to $766M (from $379M) and reported EBITDA reached ~$241M; the company mined Bitcoin and accumulated a treasury that peaked near 13,561 BTC (Q2-FY26). On the operating metrics that an analyst can verify, CleanSpark is the best-run name in Bitcoin mining.

That is necessary but not sufficient. Three findings dominate this analysis.

First, the industry is structurally bad and CleanSpark has no durable moat. Bitcoin mining is a near-perfect commodity business: the output (a Bitcoin) is fungible and globally priced, there are no customers to capture, no proprietary technology that persists (ASICs are bought off-the-shelf from Bitmain/MicroBT), and a negative network externality — every miner that adds hashrate raises difficulty and erodes every other miner’s economics. June 2026 is the bust phase of the post-April-2024-halving capital cycle in textbook Marathon form: network hashrate has rolled over from >1,000 to ~893–918 EH/s, a ~10% downward difficulty adjustment landed around June 13, 2026, hashprice sits at a five-year low (~$28–34/PH/day), BTC trades ~$64–70k versus a ~$126k 2025 high, and ~20% of legacy miners are mining at a loss. CleanSpark is a low-cost operator, not a low-cost franchise: its cheap power, land and interconnection are contractual and geographic advantages any competitor can replicate by signing the same PPAs and buying the same machines. The only quasi-durable asset is its incumbent interconnection-queue position and permitted megawatts in supply-constrained markets — a timing edge measured in quarters, not a structural barrier.

Second, the equity is now a levered Bitcoin call wrapped around an unsigned AI option — and the balance sheet flipped. Under fair-value crypto accounting (ASU 2023-08, adopted FY2025), GAAP earnings track the Bitcoin price, not operations: FY2025 showed a +$364M GAAP “profit” that was overwhelmingly a non-cash mark-up on the coin stack, while Q2-FY26 (Mar-2026) showed a −$378M loss driven by a ~$224M non-cash mark-down. The honest read is the cash-flow statement, where operations consumed $461M of cash in FY2025 and free cash flow was −$1.02B. To fund the buildout, CleanSpark — historically debt-light and equity-funded — issued ~$1.79B of 0%-coupon convertible notes (Dec-2024 and Nov-2025) with conversion prices of $14.80 and $19.16, i.e., dilution deferred rather than avoided. Crucially, BTC treasury + cash (~$1.18B) now sits below the convertible debt, so the equity is not asset-backed (net liquid ≈ −$0.6B) — a materially weaker position than MARA’s roughly neutral net-liquid claim.

Third, the bull case is real but entirely prospective, and CleanSpark is a late mover. The correct strategic response to broken mining economics is to redeploy cheap power toward AI/HPC, and CleanSpark holds 1.8 GW of contracted capacity in Georgia, Texas and Wyoming with which to do it. But it has zero signed HPC leases — only a “lead prospective tenant” at its 250 MW Sandersville, GA site — while IREN, RIOT, WULF, CIFR and Core Scientific have already signed the hyperscaler/neocloud leases that re-rated their equities to $/MW data-center multiples. The market prices CleanSpark in the un-contracted-miner bucket (~$100M EV per EH/s; ~70th percentile of its own historical valuation), and that gap to the re-rated peers is the bull case — but it closes only on a signed lease. This memo takes no position and sets no price target outside Claude’s Take above. What follows is the evidence.

2. Business Overview

What CleanSpark does. CleanSpark self-mines Bitcoin: it owns and operates data centers full of application-specific integrated circuits (ASICs) that compete to validate Bitcoin blocks, earning the block subsidy (3.125 BTC per block after the April-2024 halving) plus transaction fees in proportion to its share of global network hashrate. Essentially all revenue is the fair value of Bitcoin mined. In FY2025 the company generated $766.3M of revenue (up 102% from $378.97M in FY2024 and $168.4M in FY2023), reflecting a fleet that scaled to a 50 EH/s peak hashrate (~4.3% of the global network) across ~33 sites and ~1,027 MW of operational capacity by fiscal year-end. (Source: FY2025 10-K, filed 2025-11-25; DEF 14A, 2026-01-22.)

A second, emerging identity — energy and data-center developer. Management now frames CleanSpark as evolving “into a digital infrastructure and data center development company,” leveraging its heritage as “energy natives.” The Q2-FY26 call (May-11-2026) repeatedly emphasized 1.8 GW of currently contracted power capacity (toward a ~2 GW portfolio) and a multi-gigawatt growth pipeline, positioning the company to convert select sites from Bitcoin mining to AI/HPC tenancy. This is the central strategic story and the source of the equity’s optionality.

Revenue composition and recurring vs. non-recurring. Today, ~100% of revenue is self-mined Bitcoin — non-recurring and non-contractual by nature (a price-taker’s claim on a fixed block reward). A small Digital Asset Management (DAM) desk monetizes the idle treasury via lending and derivatives overlays, generating ~$17.2M of cash fiscal-year-to-date through Q2-FY26 — immaterial against quarterly results. The prospective HPC business, if leases are signed, would introduce recurring, contracted revenue (long-duration leases with credit tenants) — the qualitative opposite of mining — but it is pre-revenue as of this writing.

Footprint. CleanSpark’s operating sites are concentrated in the US Southeast and Texas — Georgia (Sandersville, College Park/Norcross/Atlanta metro), Tennessee (Jackson and the former GRIID sites), Mississippi, plus development sites in Texas (Sealy, Brazoria) and Wyoming (Cheyenne). The portfolio is almost entirely owned and self-operated, a deliberate contrast with MARA’s historically hosted model and the structural enabler of any HPC conversion.

How it makes money, plainly. Revenue = (CleanSpark’s share of global hashrate) × (block subsidy + fees) × (BTC price), against electricity (~$0.052/kWh in Q2-FY26) and operating costs. Because the company historically held much of what it mined, reported revenue was partly non-cash, and cash electricity/opex routinely exceeded cash realized — the root of chronic negative operating cash flow. By mid-2026, with BTC down and the HPC build consuming capital, CleanSpark had effectively abandoned the pure-HODL posture, selling nearly all production (May-2026: 671 BTC produced, 654 sold at ~$79,934).

Verdict. A best-in-class commodity Bitcoin-mining operation in the early innings of a transformation into an owned-power, AI/HPC data-center developer. Revenue is ~entirely Bitcoin mining; the data-center business is contracted on the power side but pre-revenue on the lease side.

3. Industry Dynamics

The structure of Bitcoin mining. This is one of the purest commodity businesses in public markets, and a structurally poor one. The product is perfectly fungible and globally priced; a miner sells into a market with infinite, anonymous demand and earns exactly its proportional share of a fixed, periodically halving block reward. There is no contract, no customer, no pricing power, and no ability to differentiate output. Revenue is set by three exogenous variables — the BTC price, the global network hashrate (which determines your share), and the block subsidy — and one semi-controllable cost: electricity.

Where the cycle sits, June 2026 — textbook Marathon bust phase. The “Capital Returns” capital cycle is turning in real time:

• BTC price ~$64–70k, down ~45–49% from the ~$126k 2025 high. (Industry data, mid-June 2026.)
• Network hashrate ~893–918 EH/s and rolling over from a >1,000 EH/s peak in April/May 2026, triggering a ~10% downward difficulty adjustment around June 13, 2026 — one of the larger drops on record and a capitulation signal. (Hashrate Index / bt-miners, 2026-06.)
• Hashprice ~$28–34/PH/day — a five-year low, at or below cash breakeven for many operators; CoinShares estimates ~20% of legacy miners are mining at a loss. (CoinShares Q1-2026; Hashrate Index.)

The mechanism. The April-2024 halving cut the block subsidy to 3.125 BTC. High BTC prices through 2024–25 drew a flood of capital and machines into the network (hashrate roughly doubled toward ~1,000 EH/s), mechanically shrinking every miner’s slice of a fixed reward. With BTC off its highs, hashprice has compressed to breakeven, and for the first time in a while hashrate is contracting — the June-2026 difficulty drop is the visible signature of marginal miners curtailing or capitulating. High returns attracted capital → difficulty ratcheted up → margins compressed → marginal capacity is now exiting → difficulty falls. A brutal, recurring boom-bust.

Greenwald test. Mapping the industry to “Competition Demystified”: no barriers to entry (anyone with power, ASICs and capital can enter; ASICs are commodities), no demand captivity (Bitcoin is fungible — there are no customers to retain), and no proprietary technology that endures (efficiency gains are bought off the shelf and competed away). The only persistent variable is power cost, and cheap stranded/behind-the-meter generation is available to anyone willing to contract for it. This is precisely why essentially the entire cohort — IREN, RIOT, WULF, CIFR, Core Scientific, and now CleanSpark — is redeploying toward AI/HPC.

The AI/HPC end-market — a structurally better profit pool, but crowding fast. The adjacent opportunity (leasing power-ready data-center capacity to hyperscalers/neoclouds) is genuinely better: long-duration, contracted, credit-tenant cash flows at $/MW economics far above mining. But industry analysis of CIFR and WULF already flags that HPC-colocation returns are compressing as every miner, developer and hyperscaler races for the same megawatts; the window to capture premium $/MW economics is open now and narrowing.

Verdict: structurally bad core industry, with a better but rapidly crowding adjacent pool. Bitcoin mining has no barriers to entry, no pricing power, and classic capital-cycle commodity economics currently mean-reverting through breakeven. The HPC adjacency is the escape hatch the whole industry is running for — and CleanSpark is running for it later than the leaders.

4. Competitive Position

The blunt conclusion: CleanSpark is the best operator in mining, but has no durable franchise moat. Operational excellence in a moatless commodity business is a real advantage at the margin and a fragile one in aggregate — it lowers cost per coin but cannot prevent the industry-wide difficulty ratchet from competing those gains away.

Mapping to Greenwald’s taxonomy:

• Economies of scale + customer captivity? No captivity exists — Bitcoin is sold into a global fungible market with no customers to lock in. Scale buys marginal procurement leverage on ASICs and power, not defensible share. There is no minimum-efficient-scale that excludes entrants.
• Cost advantage? Real but replicable. CleanSpark runs a ~16.07 J/TH fleet (top-quartile; better than MARA’s ~18.6, competitive with IREN’s ~15) at ~$0.052/kWh power, ~100% owned and self-operated with industry-leading uptime. This delivers a genuine operating cost edge — but cheap power, land and interconnection are contractual/geographic, not proprietary, and competitors are signing the same kinds of PPAs.
• Supply / network / brand? None apply. There is no network effect and no brand premium for fungible Bitcoin.

The numbers confirm a cost edge without a moat. CleanSpark is widely regarded as among the lowest-cost, most-disciplined self-miners — the ~16 J/TH fleet and tight opex are real and verifiable. But the once-cited “~$34,000/BTC marginal cost” is stale (Q4-FY25); industry costs have inflated sharply with difficulty (RIOT’s all-in cost rose 54% YoY to ~$49,645/BTC in FY2025; IREN’s electricity-only cost ~$34,325). CleanSpark’s energy-only cost is likely in the high-$30k/BTC range with all-in (including depreciation and G&A) plausibly $50k+ — uncomfortably close to a ~$65k BTC price. A cost advantage that any competitor can replicate, and that the difficulty ratchet erodes each quarter, is not a moat.

The genuinely differentiated asset: ~100% owned infrastructure + a 1.8 GW contracted power bank. CleanSpark’s owned-and-operated model (vs MARA’s ~44%-hosted hashrate) is the single most defensible feature, because it gives full control of cost and uptime and the ability to repurpose sites for HPC. The 1.8 GW of contracted capacity in supply-constrained Georgia and ERCOT markets is a real, scarce asset. But scarcity of grid-connected power is an industry-wide condition, not a CleanSpark-specific moat; the quasi-durable edge is the incumbent interconnection-queue position and permitted megawatts — a head-start measured in quarters, not a structural barrier.

Scale vs peers. CleanSpark’s ~50 EH/s is #2 in the US public cohort behind MARA (~72), ahead of RIOT (~42.5) and IREN (~30+, targeting 50). But, as with MARA, scale has not produced a superior return on capital — it has produced a larger position in a business with no pricing power.

Verdict. No durable competitive advantage in mining — a best-in-class operator of a commoditized, moatless business. The only path to a defensible franchise is converting owned power into long-dated HPC leases; that would substitute contracted, captive-tenant economics for commodity exposure, but it is prospective.

5. Growth History and Forward Opportunities

Historical growth — real, mostly organic, but commoditized. Energized hashrate scaled rapidly to the 50 EH/s target by mid-2025, and revenue compounded from ~$39M (FY2021) to $131.5M (FY2022) to $168.4M (FY2023) to $379M (FY2024) to $766M (FY2025). Unlike MARA’s acquisitive sprawl, CleanSpark’s growth was predominantly organic — greenfield builds plus ASIC fleet expansion — with tuck-in site acquisitions (ATL Data Centers in 2020; the Sandersville, GA site from Mawson in Dec-2022; GRIID Infrastructure, an all-stock deal of ~$155M enterprise value closed Oct-30-2024 that brought three Tennessee sites and a >400 MW runway). This is cleaner growth than the peer set — but cleaner is not the same as value-creating.

Why the growth has been low-quality. Growing a moatless, price-taking commodity business is value-destructive at the margin: each incremental exahash the industry adds raises difficulty and compresses hashprice (now at five-year lows), and CleanSpark’s own +102% FY2025 revenue growth coincided with collapsing per-coin economics (cost to mine up ~50%+ industry-wide). The clearest evidence: Q2-FY26 revenue fell ~24.9% year-over-year despite a larger fleet, because the BTC price and difficulty — not CleanSpark’s execution — set revenue. Hashrate has now flatlined at ~50 EH/s (peak)/~46 EH/s (average) as capital and management attention shift to HPC. In a moatless business, scaling is not the path to value creation; exiting toward HPC is.

The forward opportunity — the entire thesis: HPC/AI conversion. CleanSpark’s 1.8 GW of contracted power is the raw material for a re-rating. The portfolio:

• Sandersville, GA — 250 MW, fully energized (the flagship monetization candidate; +122 acres acquired Jan-2026 for a greenfield data-center build; a “lead, high-credit-quality prospective tenant” is in active commercial talks — no signed lease).
• Sealy, TX — 285 MW approved, ~200 MW energizing 1H-2027.
• Brazoria, TX — 600 MW in two phases, ERCOT approval in hand for the first 300 MW.
• Cheyenne, WY — 110 MW, fence-line with an existing hyperscaler (potential interruptible-to-firm conversion).
• Jackson, TN — 60 MW; Metro Atlanta — +25 MW added recently.

Management emphasizes that prospective tenants increasingly engage on a portfolio basis (a single tenant wanting capacity across multiple geographies), that ~60–70% of a data-center build is factory-fabricated, and that it will not deploy major capital (“millions, not tens of millions”) until a lease is signed — a disciplined, capital-light posture.

Assessment — credible optionality, clear late-mover risk. The strategic logic is sound and the power bank is genuine. But CleanSpark is 6–18 months behind the peers that have already signed: IREN’s ~$9.7B Microsoft GPU-cloud deal plus a ~$2.1B Nvidia arrangement; RIOT’s AMD lease (expanded to ~50 MW); signed hyperscaler/neocloud leases at WULF, CIFR, Core Scientific and APLD. And the conversion is not trivial: mining load is interruptible/flexible, whereas hyperscaler AI training demands firm, redundant, low-latency power and fiber — requiring substantial incremental capex (substations, advanced cooling, redundancy) and the right interconnection firmness. The 1.8 GW is a necessary but not sufficient condition for a hyperscaler deal.

Verdict: low-quality growth in the core franchise to date, with a credible but unproven and late forward option. Mining hashrate growth is commoditized; the HPC pivot could be transformational but is, as of this writing, entirely prospective — zero lease dollars — and behind the peers who have already banked them.

6. Financial Quality

GAAP earnings are noise. Under ASU 2023-08 (adopted FY2025), Bitcoin is carried at fair value with changes running through the income statement, so reported net income tracks the BTC price, not operating performance. The arc: net losses of −$21.8M (FY2021), −$57.3M (FY2022), −$138.1M (FY2023), −$145.8M (FY2024), then a +$364.5M GAAP “profit” in FY2025 — overwhelmingly a ~$1.19B non-cash mark-up on the coin stack as BTC rose into fiscal year-end, not operating earnings (operating income was −$107M). The whipsaw is already reversing: Q2-FY26 (Mar-2026) was a −$378.3M net loss on just $136.4M of revenue, driven by a ~$224M non-cash BTC mark-down (management referenced ~$263M of mark-to-market charges; retained earnings fell ~$408M in the quarter). Trailing twelve-month EPS is −$1.96. Any GAAP P/E is meaningless; the correct anchors are the cash-flow statement, EBITDA, and net asset value.

The cash-flow reality. Operating cash flow was −$17.2M (FY2023), −$233.7M (FY2024), and −$461.0M (FY2025); free cash flow was −$1.02B in FY2025 after $563M of capex (and −$1.04B in FY2024 after $806M capex). Like every miner, CleanSpark’s operations consume cash because it historically did not sell all it mined and because the fleet buildout is capital-intensive; the equity has been sustained by capital-markets issuance — first equity, latterly convertible debt. This is the central financial fact: on a cash basis the business has burned over $1B per year while building out, funded by the securities markets.

Margins and unit economics. Reported gross margin was 41.6% (FY2025) and EBITDA margin ~31.5% — but these are flattered by the BTC fair-value tailwind in the period and by depreciation accounting; the cash economics are far thinner. Energy cost per BTC is rising with difficulty (~$0.052/kWh at ~16 J/TH implies an energy-only cost in the high-$30k/BTC range; all-in plausibly $50k+ against a ~$65k BTC price). SBC was $45.3M in FY2025 (~6% of revenue) — meaningful but a minority of total dilution and modest relative to MARA’s ~$172M.

Balance sheet — large, but no longer asset-backed. At Mar-31-2026: cash ~$260M; a Bitcoin treasury of 13,561 BTC (~$925M at quarter-end, ~$0.9–1.1B at mid-2026 prices); net fixed assets ~$1.48B (the mining fleet and sites); goodwill $131.7M and intangibles ~$4.3M; total assets ~$2.91B. Against this sits ~$1.79B of 0%-coupon convertible debt plus a $71M tax payable. Total equity is ~$986M (BVPS ~$3.84; tangible BVPS ~$3.31, so P/TBV ~5.2x at $17.24). The critical observation: BTC + cash (~$1.18B) is now below the convertible debt (~$1.79B), so the net liquid claim is roughly −$0.6B — the equity is not backed by net liquid assets. Two BTC-collateralized credit lines (Coinbase $300M, Two Prime $100M) are available but undrawn, adding ~$400M of contingent liquidity that is itself BTC-price-sensitive.

ROIC/ROE. Not meaningful in the conventional sense — both are dominated by BTC fair-value swings (a financial-data aggregator shows return-on-invested-capital of ~−4% for FY2025 on an operating basis, versus a GAAP ROA of +14% flattered by the BTC mark). On the economically honest read — operating income over invested capital — CleanSpark does not earn its cost of capital in the current hashprice environment; the mining business, ex-BTC-appreciation, is roughly breakeven-to-loss-making on a fully-loaded basis.

Verdict: do economics improve with scale? No. Scale has produced more revenue and a larger coin stack, but not durable cash generation or returns above the cost of capital — because the output price is exogenous and the difficulty ratchet competes away efficiency gains. The economics are a leveraged function of the Bitcoin price, now amplified by ~$1.79B of debt.

7. Capital Allocation

The defining event: a balance-sheet transformation in ~14 months. CleanSpark spent most of its history as a net-cash, equity-funded miner (net cash of ~$55M and just $8.2M of long-term borrowings at Sept-2024). It then issued ~$1.79B of 0%-coupon convertible notes: a $650M 2030 tranche (Dec-2024; conversion price ~$14.80; capped calls with a $24.66 cap to blunt dilution; holder put June-2028) and a $1.15B 2032 tranche (Nov-2025; conversion price ~$19.16). At $17.24, the 2030 notes are in the money and dilutive on conversion; the capped call only protects the smaller tranche. The 0% coupons are genuinely cheap cash-flow, but these are deeply equity-linked options — dilution deferred, not avoided, layered onto a moatless business whose net liquid assets no longer cover the debt.

Equity issuance — a serial diluter, now moderating. Shares outstanding grew ~15x in five years: 17.4M (FY2020) → 37.4M → 55.7M → 160.2M → 270.9M → 284.3M (FY2025), funded by a large ATM program ($1.23B of stock issued in FY2024 alone). Management has since used note proceeds to repurchase stock ($145M in Dec-2024, ~$460M in Nov-2025), deliberately swapping future ATM dilution for convertible debt and shrinking the ex-treasury count to ~256M. This is a more shareholder-conscious posture than the prior ATM-at-any-price approach — but it has substituted leverage and deferred dilution for immediate dilution, not eliminated the cost of capital.

M&A — disciplined relative to peers. The GRIID acquisition (~$155M all-stock, Oct-2024) brought real Tennessee power and land at a modest size relative to CleanSpark’s then ~$3B market cap; earlier deals (ATL Data Centers, the Sandersville site) were capacity-sourcing rather than empire-building. Goodwill rose to $131.7M — modest. CleanSpark has not made the kind of large, value-destructive acquisition that has marred others; its capital sins are dilution and leverage, not overpriced M&A.

Incentive alignment — the clearest red flag. Compensation is keyed to scale and share price, not per-share value or returns on capital. The proxy frames performance as EH/s (50.0) and MW (1,027) milestones; incentive design uses operational targets plus market-performance stock units (MPSUs) tied to a $7.00 baseline price and relative TSR. There is no ROIC, no per-share, and no free-cash-flow metric anywhere in the plan — precisely the structure that green-lights dilutive, debt-funded scaling regardless of per-share economics. CEO compensation was ~$44.9M (Schultz, FY2025), and former CEO Bradford received $48.8M (including a $24.7M severance “All Other” component) — a combined ~$94M to two CEOs in a year of negative operating income. The FY2025 say-on-pay vote drew only ~76% support (soft, signaling institutional pushback).

Governance and control. A Series A Preferred block of 1,000,000 shares held by the founder group carries 45,000,000 votes — entrenching control well above economic ownership — and is entitled to a dividend of 2% of EBITA ($10.7M paid in FY2025). This preferred dividend is the source of the ~$0.04/share figure in data feeds; CleanSpark has never paid, and does not plan to pay, a common dividend. The preferred payout is a related-party cash drain to insiders, not a shareholder return. The August-2025 founder-CEO swap (Bradford out abruptly; Schultz, CEO 2014–2019, reinstated) was a founder/board reassertion of control with no transition runway.

Insider behavior. Across a five-year, 231-Form-4 record, there was exactly one open-market purchase (code P): S. Matthew Schultz on 2021-07-07 — five years ago. Every recent Form 4 is routine equity-comp churn (option/RSU exercise, tax withholding, grants). Insiders are net sellers with zero open-market conviction at current prices — a notable contrast with the bullish insider tapes seen at some other the author-covered names.

Verdict: mixed-to-weak. Genuinely disciplined on M&A and operations, and recently more thoughtful on dilution (converts + buybacks). But the capital story now rests on ~$1.79B of equity-linked debt against a non-asset-backed equity, a comp plan that ignores ROIC and per-share value, founder super-voting entrenchment, a related-party preferred dividend, weak say-on-pay, and no insider buying. Management has allocated capital competently within a bad business; it has not demonstrated that scaling that business creates per-share value.

8. Changes and Headwinds — Last Two Years

Strategic pivot to AI/HPC (2025–2026). The defining change: CleanSpark reframed itself from pure-play miner to “digital infrastructure and data center development company,” contracting 1.8 GW of power and marketing flagship sites (Sandersville, GA) for HPC tenancy. As of the May-2026 call, this remains pre-lease — a “lead prospective tenant” in talks, no signed deal. (Strengthens the thesis if executed; today it is optionality, not earnings.)

Balance-sheet leverage (Dec-2024 → Nov-2025). Two convertible raises totaling ~$1.79B transformed a net-cash miner into a levered one, funding buybacks and the HPC buildout. (Weakens the thesis: leverage on a BTC-price-sensitive, non-asset-backed equity.)

Leadership change (Aug–Sep 2025). Co-founder/CEO Zachary Bradford resigned abruptly (Aug-10-2025); co-founder Matthew Schultz (Chairman, prior CEO 2014–2019) returned as CEO; CFO Gary Vecchiarelli was named President. (Yellow flag: an abrupt founder swap with a large exit package at the moment of the leverage pivot.)

Accounting change (FY2025). Adoption of ASU 2023-08 fair-value crypto accounting turned GAAP earnings into a Bitcoin-price derivative — a +$364M FY2025 “profit” followed by a −$378M Q2-FY26 loss, both dominated by non-cash marks. (Neutral mechanically, but it makes headline EPS analytically useless and amplifies reported volatility.)

Post-halving margin compression and the 2026 difficulty cut. The April-2024 halving plus a flood of network hashrate compressed hashprice to five-year lows; the ~10% downward difficulty adjustment around June-13-2026 confirms marginal-miner capitulation. (Headwind to the core business; partial relief for survivors via lower difficulty.)

HODL posture abandoned. As BTC fell and the HPC build consumed cash, CleanSpark shifted from accumulating to selling nearly all production (May-2026). (Rational liquidity management, but it removes the “MSTR-for-miners” treasury-accumulation narrative.)

Verdict. The two-year change set is net thesis-defining rather than thesis-strengthening: the HPC pivot creates the upside option, but the leverage, the founder-CEO churn, the BTC-driven earnings whipsaw, and the deteriorating mining economics all raise the risk profile. The stock’s fate now hinges on a single binary — a signed HPC lease — overlaid on the Bitcoin price.

9. Risk Analysis (Risk Matrix)

Catastrophic-loss read. A total loss is unlikely — CleanSpark owns real, saleable assets (an efficient ASIC fleet, owned-and-permitted megawatts in scarce markets, a ~$0.9–1.1B Bitcoin treasury) and generates revenue. But because the equity is not net-liquid-asset-backed (debt > BTC + cash), a severe BTC drawdown toward ~$45k would push mining cash margins negative and erode the treasury below the converts, making the equity a deeply out-of-the-money option on the HPC pivot. The realistic bad case is a large, leverage-amplified drawdown — not impairment to zero.

10. Valuation Discussion (Embedded Expectations)

The right lenses. GAAP P/E is meaningless (BTC-driven). The useful frames are: (a) net asset value / sum-of-the-parts; (b) EV per EH/s (mining); © EV per MW (the HPC option); (d) own-history valuation percentiles; (e) EV/EBITDA with the caveat that EBITDA is BTC-sensitive.

Sum-of-the-parts / net asset value. At $17.24 and ~256.6M shares, the market cap is ~$4.42B. Subtracting the Bitcoin treasury (~$0.9–1.1B) and cash (~$0.26B) and adding back the ~$1.79B of converts yields an operating/strategic enterprise value of ~$5.0B. Against ~50 EH/s of mining, that is ~$100M of EV per EH/s — mid-pack and roughly in line with MARA (~$103–111M), i.e., the market is not paying CleanSpark a premium mining multiple, nor crediting it the $/MW HPC multiples (~$22.5M+ per MW, ~15x an implied ~$1.5M/MW NOI) that re-rated peers with signed leases. Critically, net liquid value is ≈ −$0.6B (converts exceed BTC + cash), so essentially the entire $4.42B equity value is the going-concern mining business plus the unsigned HPC option — a more demanding setup than MARA, whose net-liquid claim is roughly neutral.

Own-history percentiles (a market-data provider). CleanSpark trades at the ~70th percentile of its own historical valuation (composite): P/B 4.68 (~82nd percentile — elevated) and P/S 6.78 (~58th percentile — middling); P/E is null (negative TTM EPS). So the stock is not at a cyclical-trough valuation versus its own range — it is moderately rich, consistent with a market already paying for some HPC optionality.

EV/EBITDA. ~16–20x TTM EBITDA (ROIC: EV/TTM-EBITDA ~16x on the FY2025 basis, ~20x on the Mar-2026 TTM) — but EBITDA is a BTC-price artifact and will compress further if hashprice stays at five-year lows. Not a reliable anchor.

Embedded expectations — what the price implies. At ~$100M EV/EH/s with no premium for signed HPC, the market is underwriting: (1) the mining business survives at roughly current scale through the hashprice trough; (2) the Bitcoin price holds in a range that keeps the treasury and cash margins viable; and (3) a non-trivial probability that at least one HPC lease gets signed, which the ~70th-percentile own-history multiple partially prices. What the price is not yet paying for is a consummated HPC re-rating — that is the asymmetric upside if Sandersville (or a portfolio deal) signs, and the downside if it does not and BTC weakens.

Scenario analysis (illustrative; not a target).

• Bear (~30%): BTC drifts toward ~$45–55k; hashprice stays at trough; no HPC lease signs through 2026; the converts’ 2028 put looms over a non-asset-backed equity. Mining cash margins go negative; the stock de-rates toward tangible book and net-liquid stress — a ~$7–11 range (a large, leverage-amplified drawdown, not impairment).
• Base (~45%): BTC ranges ~$60–85k; CleanSpark holds ~50 EH/s at top-quartile cost; one mid-sized HPC lease (or continued credible Sandersville progress) keeps the option alive but unconsummated; multiple holds ~70th-percentile own-history. The stock sits roughly in its current ~$13–19 band.
• Bull (~25%): CleanSpark signs an investment-grade HPC lease at Sandersville (and/or a portfolio deal) at credible $/MW economics, and BTC holds firm. The market begins paying $/MW data-center multiples on the converting megawatts, closing the gap to re-rated peers — a ~$24–35+ re-rating (back above the cycle high), with the magnitude set by the MW leased and the tenant’s credit.

The payoff is bimodal and governed by two variables: the Bitcoin price (continuous) and the HPC lease (binary). No price target; no recommendation.

11. Variant Perception

Consensus view. Sell-side and crypto-equity investors broadly regard CleanSpark as the best-run pure-play miner — lowest-cost, most operationally disciplined, ~100% owned — and increasingly as a cheap, un-contracted call option on the AI/HPC pivot given its 1.8 GW power bank (e.g., Chardan reiterated Buy and raised its target to $19 on Jun-8-2026). The bullish consensus is “great operator + scarce power + AI optionality, at a discount to re-rated peers.”

Strongest bull case. CleanSpark owns and controls 1.8 GW of contracted power in the exact supply-constrained markets (Georgia, ERCOT Texas) where hyperscalers are desperate for capacity; it runs the most efficient fleet at the lowest cost; and it is one signed lease away from the $/MW re-rating that already happened to IREN, RIOT, WULF and CIFR. If Sandersville signs an investment-grade tenant, the market re-prices the converting megawatts at data-center multiples (~$22.5M+/MW) versus the ~$100M/EH/s mining multiple it pays today, and the stock re-rates sharply. Meanwhile the survivor benefits from the difficulty cut as marginal miners capitulate.

Strongest bear case. CleanSpark is a moatless commodity miner at the trough of the capital cycle, now levered with ~$1.79B of converts against an equity that is not asset-backed (BTC + cash < debt). GAAP earnings are a BTC-price derivative; operating cash flow is deeply negative; the comp plan rewards scale, not per-share value; insiders don’t buy; and the entire upside rests on an HPC lease the company has been “in talks” on for quarters without signing — while later-moving than peers who already closed deals. If the lease doesn’t come and BTC weakens, the stock is a levered short-the-cycle instrument with deferred-dilution overhang and a 2028 convert put.

The 3–5 assumptions that matter most:

1. Does Sandersville (or a portfolio deal) sign a credible HPC lease, and when? The single binary that closes or sustains the valuation gap. Falsifies the bull case if talks die with no replacement; falsifies the bear case the day an investment-grade lease is signed.
2. Where does the Bitcoin price go? Continuous driver of revenue, treasury value, cash margins, and the net-liquid backstop. A break toward ~$45k falsifies the constructive case; a sustained hold above ~$80k supports it.
3. Does the network hashprice mean-revert up, or stay at five-year lows? Determines whether the core mining business is cash-generative or a cash sink while waiting for the pivot.
4. Can owned, interruptible mining megawatts be converted to firm HPC load at acceptable capex and interconnection firmness? The technical gating factor on the pivot’s economics.
5. Will management create or destroy per-share value with the leverage? Comp ignores ROIC/per-share; dilution history is adverse; this is the governance swing factor.

Factor-positioning read (factor & price-action analysis). CleanSpark screens as a high-beta (~2.95), high-idiosyncratic-vol, small-cap “Fintech”-industry levered Bitcoin proxy — its dominant factor loadings are Market (~1.85–2.13 across models) and SmallSize (~1.63), with the bulk of variance idiosyncratic to the BTC/crypto-equity complex (factor-similar peers are MARA/RIOT leveraged ETFs, Bitfarms, Bit Digital, HIVE — a pure miner cluster). The tape is momentum-positive, not a falling knife: rs_12m +93.7%, rs_6m +45%, rs_YTD +70%, currently ~−26% off the November-2025 cycle high and ~+110% off the January-2026 low. The risk-adjusted track record is too regime-dependent to annualize meaningfully (the factor model returned no stable leaderboard). The read for Variant Perception: consensus is leaning long the AI-option narrative into a BTC-stabilization bounce — a crowded, high-beta momentum posture that would unwind violently on either a BTC break or a failed Sandersville process, and would extend sharply on a signed lease. This is positioning evidence, not a price call.

12. Fact vs. Interpretation

13. Open Questions

1. When, and on what economics, does Sandersville sign? What tenant credit quality, lease tenor, $/MW, and capex split — and is it a single site or the portfolio deal management hints at?
2. What is CleanSpark’s true all-in cost to mine in 2026? The “$34k” figure is stale; an updated fully-loaded number (energy + depreciation + G&A) would clarify cash breakeven against BTC.
3. How firm is the 1.8 GW? How much is firm vs interruptible load, and what is the interconnection-queue/energization timeline by site (Sealy 1H-2027; Brazoria phase 2)?
4. What is the conversion capex per MW to take an interruptible mining site to hyperscaler-grade firm, redundant, liquid-cooled capacity — and who funds it (tenant build-to-suit vs CleanSpark balance sheet)?
5. How does management handle the 2028 convert put if BTC and the stock are weak and the equity remains non-asset-backed?
6. Will the comp plan ever incorporate ROIC or per-share value — or does scale-keyed comp continue to incentivize dilutive growth?
7. What is the post-Bradford strategic cadence under Schultz — does the abrupt swap signal a change in pivot urgency or risk appetite?

14. What Must Be True

For the bull case (constructive at/above current levels):

1. CleanSpark signs at least one investment-grade HPC/AI lease (Sandersville and/or portfolio) at credible $/MW economics within the next several quarters. Falsification: the Sandersville process ends with no signed tenant and no comparable replacement by year-end 2026.
2. The Bitcoin price holds broadly above ~$60–65k, keeping the treasury and mining cash margins viable while the pivot matures. Falsification: a sustained BTC break below ~$50k.
3. The converting megawatts re-rate at data-center $/MW multiples, closing the gap to peers, without a capital raise that heavily dilutes the option. Falsification: a large equity issuance or convert conversion that swamps the per-share benefit of the lease.

For the bear case (avoid / structurally impaired):

1. No HPC lease materializes and CleanSpark remains a pure commodity miner at the trough. Falsification: a signed investment-grade lease.
2. Hashprice stays at five-year lows (or BTC falls), keeping mining cash-flow-negative and stressing the non-asset-backed equity against the 2028 put. Falsification: a durable hashprice recovery or a BTC move back above ~$90k.
3. Governance/capital allocation continues to favor scale over per-share value — further dilution, no ROIC in comp, continued insider non-participation. Falsification: a comp redesign incorporating per-share/ROIC metrics and/or meaningful insider open-market buying.

The bull and bear converge on a single pivot: does the 1.8 GW of contracted power convert into signed, credit-tenant HPC leases before the Bitcoin/hashprice cycle and the convert maturities force the issue? Everything else is second-order.

15. Source Appendix

See the separate Source Appendix (CLSK_source_appendix.md) for the full, dated citation list. Primary sources include: CleanSpark FY2025 Form 10-K (filed 2025-11-25); Form 10-Q for the quarter ended Mar-31-2026 (filed 2026-05-11); DEF 14A proxy (2026-01-22); Form 8-Ks (GRIID close 2024-10-31; CEO transition 2025-08-11; convertible-note issuances Dec-2024 and Nov-2025); the Form 4 insider corpus (CIK 0000827876); the Q2-FY2026 earnings-call transcript (2026-05-11, via a financial-data aggregator); CleanSpark monthly operational updates (May-2026); financial-data aggregators (fundamentals and valuation percentiles), a quantitative factor model, financial news feeds; and industry data on Bitcoin network hashrate, difficulty and hashprice (Hashrate Index, CoinShares, bt-miners), accessed June 2026. Peer cross-reads drawn from prior published analysis of peers MARA, RIOT, IREN, WULF and Core Scientific.

APPENDIX A — Standard Diligence Questionnaire

CleanSpark, Inc. (NASDAQ: CLSK) — supplemental diligence. Report date: 2026-06-20. Price $17.24.

Supplemental to the memo; grounded in the underlying analysis. Labels: F = Fact, I = Interpretation, A = Assumption.

General

What thoughtful questions have other investors asked about this company? (I) Three dominate: (1) When does the AI/HPC pivot produce a signed lease, and on what $/MW economics? — the binary that re-rates or de-rates the stock. (2) Is CleanSpark’s operational edge (lowest cost, ~16 J/TH, ~100% owned) a moat or just good execution in a moatless business? (3) How safe is the balance sheet now that ~$1.79B of converts sit above a BTC + cash claim of only ~$1.18B? A fourth, growing, question: was layering on convertible leverage at a non-asset-backed equity the right call versus staying debt-light?

Cyclicality & Earnings Nature

Are earnings at a cyclical high or low? (F/I) GAAP earnings are meaningless (BTC-price-driven under fair-value accounting): +$364M FY2025, −$378M Q2-FY26. On operating economics, the business is near a cyclical low — hashprice at five-year lows, BTC off its high, Q2-FY26 revenue −24.9% YoY despite a larger fleet.

Driven by the external environment or internal actions? (I) Overwhelmingly external — the Bitcoin price and network difficulty set revenue. Internal actions (fleet efficiency, uptime, power contracts) move cost per coin at the margin but cannot offset the exogenous price/difficulty drivers.

How stable are revenues? (F) Highly unstable — a leveraged function of the BTC price and the difficulty ratchet; no contracts, no recurring revenue. The HPC pivot, if signed, would introduce contracted recurring revenue for the first time.

Outlook for products/services? (I) Mining: structurally challenged (post-halving, hashprice trough). HPC/AI: large and growing TAM, but CleanSpark is pre-revenue and a late mover.

How big will this market be? (F/I) Bitcoin mining is a fixed-reward, halving market (shrinking subsidy). The AI/HPC data-center market is large and supply-constrained on power — the reason the whole cohort is pivoting. CleanSpark’s addressable slice depends entirely on converting its 1.8 GW.

Business Quality & Competitive Moat

Is the industry getting more or less competitive? (I) Mining: more competitive and capital-cycle-compressed (difficulty ratchet). HPC: intensely and increasingly competitive as miners, developers and hyperscalers chase the same megawatts.

How profitable is the business (ROIC, ROE)? (F/I) Not meaningfully profitable on an economic basis — operating income −$107M (FY2025); operating ROIC ~−4%; GAAP ROA/ROE flattered by BTC marks. The business does not earn its cost of capital at trough hashprice.

How profitable is the industry; barriers to entry? (I) Poor, and no barriers — fungible output, off-the-shelf ASICs, no customer captivity. A textbook commodity price-taker.

Can the business be easily understood? (F) The mining mechanics are simple; the GAAP statements are confusing (BTC fair-value noise) and require cash-flow/NAV analysis.

Undermined by foreign low-cost labor? (I) Not labor — but by anyone with cheaper power anywhere. Power cost, not labor, is the competitive axis.

Do brands matter? (F) No. Bitcoin is fungible; there is no brand premium.

Nature of competition / switching costs? (F) Zero switching costs (no customers in mining). In HPC, the competition is for tenants and megawatts; tenant relationships and interconnection position are the differentiators.

Financial Condition & Balance Sheet

Assets not fully recognized on the balance sheet? (I) The interconnection-queue position and permitted-but-unenergized megawatts have option value not reflected at book; conversely, the ASIC fleet depreciates fast and obsolesces.

Off-balance-sheet liabilities? (F) None material flagged; ~$400M of undrawn BTC-collateralized credit lines (Coinbase $300M, Two Prime $100M) are contingent, BTC-price-sensitive liquidity.

How conservative is the accounting? (I) Fair-value crypto accounting is mandated (ASU 2023-08) and makes earnings volatile but is not aggressive. The Series A Preferred (2%-of-EBITA dividend, 45M super-votes) is a related-party structure to watch.

How CapEx-hungry is the business? (F) Very — capex $806M (FY2024), $563M (FY2025); the HPC build would add substantial firm-power/cooling capex per MW. A chronic cash consumer.

Capital Allocation & Management

How much FCF; how is it used; philosophy? (F) FCF deeply negative (−$1.02B FY2025). “Use of cash” has been buildout funded by securities issuance; recently note proceeds funded buybacks (~$605M Dec-2024 + Nov-2025) to swap deferred-dilution for converts. Philosophy: scale first.

Significant acquisitions recently? (F) GRIID Infrastructure (~$155M all-stock, Oct-2024, Tennessee power/land); earlier ATL Data Centers and the Sandersville site. Disciplined relative to peers.

Buying back shares? (F) Yes — ~$605M repurchased with note proceeds — but net share count still grew ~15x over five years; buybacks offset, not reverse, dilution.

Issuing large amounts of stock to insiders? (F) Large RSU/option grants plus monthly BTC payments under employment agreements; CEO pay ~$44.9M (Schultz) and ~$48.8M (Bradford, incl. severance) in FY2025.

Compensation policy / incentives? (F, red flag) Keyed to EH/s, MW and TSR/MPSU ($7 baseline) — no ROIC, no per-share, no FCF metric. Say-on-pay ~76%. Series A Preferred dividend = related-party drain.

Motivations of management? (I) Founder-controlled (Schultz reinstated as CEO Aug-2025; 45M-vote preferred). Incentives reward scale and share price, not per-share value — a structural concern in a moatless, dilution-prone business.

Valuation & Market Data

ADR, MLP, or K-1 issuer? (F) No — a Nevada C-corp common stock on NASDAQ; standard 1099 treatment. No K-1.

Dividend policy? (F) No common dividend, ever, and none planned. The ~$0.04 figure in data feeds is the Series A Preferred (2%-of-EBITA) dividend to founders.

How profitable is the business? (F/I) Economically near breakeven-to-loss-making at trough hashprice; GAAP profit is a BTC artifact.

Is net income diverging from cash from operations? (F) Massively — FY2025 GAAP NI +$364M vs operating cash flow −$461M (a ~$825M divergence), driven by non-cash BTC marks. Cash flow is the honest read.

Risks & Downside

What factors would cause the stock to decline? (I) A Bitcoin price decline; a failed/absent HPC lease; hashprice staying at trough; a dilutive capital raise or convert conversion; the 2028 put under a weak stock; governance disappointment.

Risk of a catastrophic loss? (I) Elevated drawdown risk but low total-loss risk. Real assets (efficient fleet, owned scarce-market megawatts, ~$0.9–1.1B BTC) and revenue exist; but the equity is not net-liquid-asset-backed (debt > BTC + cash), so a deep BTC drawdown is severely leverage-amplified.

Chance of a total loss? (I) Low. The realistic bad case is a large, leverage-amplified drawdown to tangible book / net-liquid stress, not a zero.

Recent News & Events

Has the business environment changed recently? (F) Yes — June-2026 saw a ~10% network-difficulty cut (marginal-miner capitulation), BTC volatility (a tumble then partial recovery), and a Chardan price-target raise to $19 (Jun-8). The HPC pivot remains pre-lease.

Significant acquisitions / accounting changes / management changes? (F) ASU 2023-08 fair-value accounting (FY2025); abrupt founder-CEO swap (Bradford → Schultz, Aug-2025); two large convertible raises (Dec-2024, Nov-2025); the strategic re-branding toward “digital infrastructure and data center development.”

Recent changes — new markets, facilities, management? (F) New Texas (Sealy, Brazoria) and Wyoming (Cheyenne) development sites; +122 acres at Sandersville for a greenfield data center; Metro Atlanta +25 MW; new executive team under Schultz/Vecchiarelli.

APPENDIX B — Source Appendix

CleanSpark, Inc. (NASDAQ: CLSK) — sources. Report date: 2026-06-20. Price at analysis: $17.24 (2026-06-18).

Primary sources first. Each material claim in the memo traces to one of these. Third-party aggregated data (financial-data aggregators and a factor model) is reconciled to filings where possible; filings are authoritative.

Primary — SEC filings (CIK 0000827876)

1. Form 10-K, FY2025 (fiscal year ended 2025-09-30), filed 2025-11-25 — revenue $766.3M; operating income −$107M; GAAP NI +$364M; fair-value crypto accounting (ASU 2023-08); 50 EH/s; ~1,027 MW / ~33 sites; no common dividend policy; risk factors. https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=0000827876&type=10-K
2. Form 10-Q, Q2 FY2026 (quarter ended 2026-03-31), filed 2026-05-11 — net loss −$378.3M; revenue $136.4M; ~$224M non-cash BTC mark-down; balance sheet (cash ~$260M; BTC 13,561; converts ~$1.79B; equity ~$986M; ~256.6M shares); convertible-note terms; BTC-backed credit lines. https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=0000827876&type=10-Q
3. DEF 14A proxy, filed 2026-01-22 — executive compensation (Schultz ~$44.9M; Bradford ~$48.8M incl. severance); incentive metrics (EH/s, MW, MPSU $7 baseline, TSR; no ROIC/per-share); ~76% say-on-pay; Series A Preferred (1.0M shares / 45.0M votes / 2%-of-EBITA dividend); ~255.75M shares at record date.
4. Form 8-K, 2024-10-31 — completion of GRIID Infrastructure acquisition (~$155M all-stock; three Tennessee sites).
5. Form 8-K, 2025-08-11 — CEO transition: Zachary Bradford resignation; S. Matthew Schultz appointed CEO; Vecchiarelli named President (effective 2025-09-04).
6. Form 8-Ks / indentures, Dec-2024 and Nov-2025 — $650M 2030 0% convertible notes (conversion ~$14.80; capped calls, cap $24.66; holder put Jun-2028) and $1.15B 2032 0% convertible notes (conversion ~$19.16); related share buybacks.
7. Form 4 insider corpus (CIK 0000827876) — 231 filings reviewed; one open-market purchase (code P, S. Matthew Schultz, 2021-07-07); otherwise routine M/F/A/S; insiders net sellers.

Primary — Company disclosures

8. CleanSpark Q2 FY2026 earnings-call transcript, 2026-05-11 (via a financial-data aggregator) — “digital infrastructure and data center development company” framing; 1.8 GW contracted; Sandersville “lead prospective tenant”; Sealy/Brazoria/Cheyenne/Jackson detail; HODL 13,561 BTC ~$925M; DAM $17.2M YTD; power $0.052/kWh; 1,799 BTC mined.
9. CleanSpark monthly operational updates (May-2026) — 50.0 EH/s peak / 46.2 EH/s average; ~16.07 J/TH; 13,470 BTC; 671 produced / 654 sold at ~$79,934.

Secondary — Quantitative data services

10. financial-data aggregator — income statement, balance sheet, cash flow, enterprise value, profitability/valuation multiples (FY2020–Q2-FY2026), accessed 2026-06-20.
11. Own-history valuation percentiles — own-history percentiles: P/B 4.68 (81.8th), P/S 6.78 (58.5th), composite ~70th; P/E null. Accessed 2026-06-18.
12. Financial news aggregator — recent items incl. Chardan PT raise to $19 (2026-06-08), BTC-tumble/crypto-equity moves (2026-06-05/06-15). Accessed 2026-06-20.
13. Adjusted price history (split/dividend-adjusted OHLCV) — five-year price arc; 52-week range; EMAs/beta. Accessed 2026-06-20.
14. quantitative factor model — stock-info (beta ~2.95, alpha +0.06, rs_12m +93.7%, rs_6m +45%, rs_peak −76% vs 2018 ATH); stock-loadings (Market ~1.85–2.13, SmallSize ~1.63, Industry: Fintech); related-stocks (MARA/RIOT 2x ETFs, Bitfarms, Bit Digital, HIVE). Accessed 2026-06-18.

Secondary — Industry and peer data

15. Bitcoin network data — hashrate ~893–918 EH/s; ~10% difficulty cut ~2026-06-13; hashprice ~$28–34/PH/day (five-year low); ~20% of miners at a loss. Sources: Hashrate Index, CoinShares Q1-2026 mining report, bt-miners; accessed June 2026.
16. Peer disclosures / coverage — IREN ~$9.7B Microsoft + ~$2.1B Nvidia deals; RIOT AMD lease (~50 MW); cost-to-mine comparisons (RIOT ~$49,645/BTC FY2025; IREN ~$34,325 electricity-only). Sources: company releases, DataCenterDynamics, CoinShares; accessed June 2026.
17. Prior the author coverage (internal cross-read) — MARA (2026-06-13), RIOT (2026-06-14), IREN (2026-06-10), WULF (2026-06-12), CORZ (2026-06-20) full reports, used for industry framing and peer comps.

Analytical frameworks

18. Competition Demystified (Greenwald & Kahn) — barriers-to-entry / moat-type taxonomy; applied to the no-moat verdict.
19. Capital Returns (Marathon Asset Management / Chancellor) — supply-side capital-cycle analysis; applied to the post-halving bust-phase verdict.

Note: prices, BTC levels, network statistics and valuation multiples are as of the access dates above and move continuously; figures cited reflect the most recent available data at report date.