What is CleanSpark, Inc. (CLSK) worth?

CleanSpark is a US bitcoin miner (~45-50 EH/s) plus a ~14,000-BTC treasury that has executed its first institutional AI/HPC lease - a 20-year NNN lease at Sandersville, GA (175 MW IT, USD6.6B contracted, ~USD330M average / ~USD251M year-1 NOI, investment-grade tenant) - and holds an LOI + exclusivity over an 885 MW Texas platform with the same tenant. Despite the feed's 'Capital Markets' label, this is an energy-and-compute infrastructure story.

Why we cut the report. Our prior Buy/USD20 correctly identified that the pivot had de-risked from 'will they sign anything' to 'how much converts,' but it mis-sized the answer. Two corrections matter: (1) Sandersville should be capitalised on year-1 NOI at a 6% cap (~USD5.7/sh), not average NOI at ~14x (~USD7.2/sh); and (2) Texas cannot be a single signing-probability scalar. Funding is the dominant variable, and our prior report never priced it. At 55-70% project-debt LTV (the 2026 benchmark, not 80%), the Texas equity slice is USD2.2-3.3B - more than CleanSpark's entire liquidity including selling every bitcoin it owns.

Corrected math. Contracted floor (Texas = 0): ~USD5.35/sh. Full-notional ceiling (Texas 100%, no dilution): ~USD22.65. Our decomposed SOTP carries Texas at a ~42% credit (market-implied ~44% ~= 90% sign x 49% realisation) and lands at ~USD12.6; the six-node scenario tree EV is ~USD13.4. Fair value USD13.00.

Risk/reward. The distribution is wide (USD5.36 - USD22.65) but centred right at the price. The upside legs all require Texas to convert AND Brazoria's ~285 MW to be real AND the build to be funded without heavy dilution - a product of three uncertain conditions, not one. We move to Hold and would re-engage on a funding announcement that is not dilutive at a low price, or on evidence that Brazoria's capacity is secured.

Probability-weighted SOTP / NAV with decomposed Texas conversion credit (P(sign) x MW realisation x funding x timing)

Every figure is computed from SEC filings by the analyst’s calculator tools and fact-checked by a second, independent model.