CLSK · CleanSpark, Inc. · Stock research

and# CLSK — CleanSpark, Inc.

Current Thesis

The narrative leg here was never the mining P&L — it was that a signed hyperscaler lease would re-rate a bitcoin miner into an AI-infrastructure landlord. That leg was tested and failed on delivery. CleanSpark executed the 20-year triple-net Sandersville lease (announced 2026-07-14, detailed in the 2026-08-05 July operational update: 175 MW critical IT load, ~$6.6B contracted revenue, ~$11.6B with two five-year extensions, ~100% NOI contribution margin) — and the stock closed 2026-08-14 at $12.09, below the $16–17 shelf it traded on before the announcement, 47.9% under the $23.20 52-week high. The reason is dated and specific: first data-hall deliveries begin Q4 calendar 2027, and fiscal Q3 (2026-08-06) printed EPS $(0.89) vs $(0.32) consensus on revenue of $138.006M vs $148.941M, down 30.5% from $198.6M a year earlier. What an investor buys today is not a re-rate in progress; it is a $1.8–2.1B project-financing execution story with no lease revenue for roughly five quarters.

Bullish and bearish views on CleanSpark, Inc.

The model's bull view on CleanSpark, Inc. (CLSK), in brief: The contract exists and is investment-grade: 20-year triple-net lease, 175 MW critical IT load, ~$6.6B contracted revenue and ~100% NOI contribution margin, counterparty described as "a high investment-grade leading global technology company" (July operational update… The bear view: Revenue timing is the whole problem. Deliveries begin Q4 calendar 2027 (2026-08-05 update), so the $6.6B headline contributes nothing to FY2026 or most of FY2027 while the mining engine carries the P&L. The mining engine is shrinking: BTC produced 671 (May) → 614 (June) → 586… Both cases follow in full.

Bull Case

• The contract exists and is investment-grade: 20-year triple-net lease, 175 MW critical IT load, ~$6.6B contracted revenue and ~100% NOI contribution margin, counterparty described as "a high investment-grade leading global technology company" (July operational update, 2026-08-05).
• Execution has been de-risked at the front end: management said the anticipated equity portion of Sandersville is fully funded and all long-lead items are ordered and pre-paid, with project-based debt targeted for more than 90% of construction cost — i.e. no equity or equity-linked issuance planned for the project (Q3 call, 2026-08-06).
• Un-leased power remains the asset: 1.8 GW contracted against 808 MW utilized as of 2026-07-31 leaves roughly 1 GW of interconnected capacity across GA/MS/TN/WY that has not been sold to anyone yet.
• Mining still generates cash before marks: adjusted EBITDA was -$113M in fiscal Q3 but management stated it would have been +$20M excluding non-cash bitcoin mark-to-market (2026-08-06).
• Sell-side has not marked down: B. Riley maintained Buy and raised its target to $26 (2026-08-03); Cantor Fitzgerald reiterated Overweight, $26 (2026-08-07) — both after the fiscal Q3 date was known, the Cantor note the day after the print.
• Sector economics remain live: Bernstein counted more than $90B of AI deals across 3.7 GW of miner power capacity (2026-05-19) and reiterated the framework on Riot with an 80% upside case (2026-08-11).

Bear Case

• Revenue timing is the whole problem. Deliveries begin Q4 calendar 2027 (2026-08-05 update), so the $6.6B headline contributes nothing to FY2026 or most of FY2027 while the mining engine carries the P&L.
• The mining engine is shrinking: BTC produced 671 (May) → 614 (June) → 586 (July, reported 2026-08-05), three consecutive monthly declines, against 50 EH/s operational hashrate and 230,507 deployed miners.
• Financing is unclosed. Sandersville needs $1.8–2.1B at $10–12M per critical IT MW; more than 90% is intended to come from project debt that has not been announced (2026-08-06). Terms, cost and timing are the swing factor and none of it is public.
• Liquidity is bitcoin-denominated: $917M total liquidity at 2026-06-30, of which roughly $200M is cash and the balance ~14,000 BTC (13,931 held at 2026-07-31). A BTC drawdown shrinks both the collateral and the reported quarter.
• The counterparty is still unnamed as of the 2026-08-05 update — a year-plus of construction risk against a tenant no filing has identified.
• Relative preference inside the theme moved elsewhere: Bernstein's 2026-08-04 Texas data-center audit backed TeraWulf over CleanSpark and Core Scientific, and its 2026-08-11 call was on Riot.
• The strongest possible catalyst already fired. A signed $6.6B lease landed on 2026-07-14 and the three-month return through 2026-08-14 is -7.8% — the full arc from "in talks" (2026-06-08 reporting) through signature nets to a lower price.

Setup & Price Structure

Reference close 2026-08-14: $12.09. That is 47.9% below the $23.20 52-week high, with RSI(14) at 41.4 and a three-month return of -7.8%. The 2026-07-14 announcement gap is fully retraced and then some: the 2026-06-18 reaction high near $17.24 and the ~$16.50 early-July shelf now sit overhead as supply, and price is under the pre-signing range rather than consolidating above it. RSI in the low 40s is mid-range — no washed-out reading to work from, and no base has formed since the 2026-08-06 print.

The narrative is dead for this name's re-rate leg. The dating is unambiguous: contract signed 2026-07-14, first read on it 2026-08-06 (miss on both lines, delivery pushed out to Q4 CY2027), price below the pre-catalyst range by 2026-08-14. The sector theme is not dead — Bernstein was still publishing upside cases on peers on 2026-08-04 and 2026-08-11 — which makes the failure name-specific rather than thematic, and that distinction matters for anyone reading CLSK weakness as a sector signal.

Crowding and positioning observables (stated, not judged): targets cluster at $26 (B. Riley 2026-08-03, Cantor 2026-08-07) against a $12.09 close, a gap of more than 2x with no post-print reductions observed; the stock trades below, not above, its longer moving averages after retracing the July gap; CLSK appeared on most-shorted "squeeze watch" screens on 2026-07-13 and was not among the names headlined in the 2026-08-13 most-shorted roundup (Ondas, Avis Budget, RH); 13F disclosures from 2026-05-20/22 showed Leopold Aschenbrenner's Situational Awareness increasing a CLSK stake alongside an $8.5B short against Nvidia/Oracle/Broadcom (2026-05-18) — a paired power-versus-chips construction, and quarterly filings are stale by the time they publish. No Form 4 activity surfaced in the reviewed filing feed through 2026-08-14.

Catalyst Calendar (next 30 days)

• ~2026-09-03 (est.) — August 2026 operational update. Monthly cadence confirmed by the 2026-07-07 (June) and 2026-08-05 (July) releases. Reads: whether BTC production breaks the 671→614→586 sequence, treasury count against 13,931, and any Sandersville construction milestone.
• Outside the window: fiscal Q4/FY2026 results, ~2026-12-01 (est.), fiscal year ends 2026-09-30. First data hall ready for service, Q4 calendar 2027 per the 2026-08-05 update.

Elapsed catalysts

• No fixed date — Sandersville project-debt close. Management targeted >90% debt funding of a $1.8–2.1B build on the 2026-08-06 call; there is no announced timeline, so this can land inside or well outside the window. (passed 20d ago)
• No fixed date — an 8-K naming the lease counterparty. Nothing had been filed identifying the tenant as of 2026-08-05. (passed 21d ago)

What Would Change Our Mind

The repair case does not run through the lease — that is signed and priced at zero right now. It runs through financing and a base. Announced project debt on the $1.8–2.1B Sandersville build at a disclosed rate, or an 8-K naming an investment-grade tenant, would put a dated number against the piece of the story the market currently discounts entirely; either would justify re-opening the re-rate frame. So would an August operational update (~2026-09-03) that ends the three-month production slide while the stock holds above the August range.

On the downside, a weekly close below $11.00 carries price through the post-print August range with no prior reaction shelf until the zone that preceded the May lease-talk rally, and would confirm the structure is broken rather than repairing. A second, independent break: the ~2026-09-03 update passing with a fourth consecutive monthly BTC decline and still no project-debt announcement, which would leave the whole equity a levered bitcoin proxy waiting on a Q4-2027 delivery date. Sell-side targets holding at $26 while price sits near $12 is a gap that resolves by one side moving — a wave of PT cuts would be the marker that the street has capitulated to the delivery timeline.

Correlation Notes

• Bitcoin spot is the dominant daily driver until lease revenue prints: it sets mining margin, the mark on 13,931 BTC, and the value of $917M of stated liquidity (2026-06-30) that is mostly BTC.
• Miner peers IREN, RIOT, WULF, CORZ trade the same AI-lease template; intra-group rotation is visible in Bernstein's 2026-08-04 preference for TeraWulf and its 2026-08-11 Riot call, so CLSK-specific weakness has been coexisting with peer upgrades.
• Credit conditions and data-center financing markets matter more than usual here: with >90% of a $1.8–2.1B build earmarked for project debt, spread widening in data-center ABS/private credit transmits directly into the equity story.
• Hyperscaler capex headlines move the group's multiple, but CLSK's contracted tenant is unnamed and its first delivery is Q4 2027, so capex revisions hit sentiment long before they could hit this contract.
• Prior AI-power beta was partly expressed through the Situational Awareness paired construction (long CLSK, short the chip complex, per 2026-05-18/20 disclosures); if that pairing unwinds, the flow is not correlated to CleanSpark fundamentals at all.