SKYWAY SUMMITInvestments · CoverageKEEL is the former Bitfarms, a Bitcoin miner that rebranded April 1, 2026, redomiciled US, and is converting its energized power sites into AI/HPC data-center colocation. The cash is there ($533M liquidity), the sites are zoned, the permits are tracking for mid-to-late summer, and lease negotiations are in progress with multiple counterparties. The peer comp is striking: KEEL trades at ~$3M per 2027 MW versus peer transactions at $10.7M per MW, a roughly 3.5× gap if the pivot is real. That gap is the trade, and the catalyst is binary: lease signs (re-rate) or lease slips (the multiple compresses on continued cash burn). This is not a buy-and-forget holding; it is a position with a specific date-driven thesis.
From Bitcoin mining to AI power-landlord, on the same land
The simplest way to understand KEEL is the pivot itself: the company spent five years building the one thing AI suddenly needs more than anything, energized power capacity in the right places, while doing something else with it (mining Bitcoin). The infrastructure was a means to an end. Now the end has changed.
What makes the pivot credible is what was already in place. The five campuses, Panther Creek (PA, 350 MW), Sharon (PA, 110 MW), Moses Lake (WA, 18 MW), Sherbrooke (QC, 96 MW), and Scrubgrass (PA, 1.3 GW pipeline), span 2.2 GW of total multi-year pipeline capacity, with 341 MW already energized and 430 MW secured. The Pennsylvania footprint matters specifically: it sits near hyperscale clusters with one of the most constrained power markets in North America, where new interconnection queues run multi-year. Keel did not have to build that queue position; it inherited it.
The demand pull is unambiguous. The Blackwell (B200/GB200) generation is sold out through mid-2026 and hyperscaler-scale deployments have moved from 20-50 MW pilots to 200-600 MW commitments. Cipher Mining (a close peer) signed a 15-year, 300 MW deal with Amazon worth $5.5B in total contract value, or roughly $367M of annual recurring revenue, in late 2025. Keel's flagship Panther Creek site is 350 MW. A comparable contract would meaningfully exceed Cipher's deal value. Whether Keel can sign such a deal is the entire question.
Power is the constraint; Keel owns the constrained thing
In the AI build-out, GPUs are the visible bottleneck, but power is the deeper one. NVIDIA can ship the chips; if there is nowhere to plug them in, the chips do not generate revenue. New utility-scale power interconnections in the United States take four to seven years; permitted, energized, site-controlled megawatts in good locations are inventory that cannot be manufactured on AI timescales. That is the structural moat KEEL is selling.
Total energized: 341 MW · total secured: 430 MW · total pipeline: 2.2 GW. The three near-term sites under permit review (Panther Creek, Sharon, Moses Lake) sum to 478 MW, the inventory directly addressable for 2026-2027 lease execution.
Two pieces of leverage matter. Operating leverage: a colocation lease is a long-duration, take-or-pay contract; once Panther Creek signs, the cash flows are essentially annuities for 15 years and the unit economics swing from heavy operating losses today to material recurring EBITDA. Strategic leverage: KEEL's customers bring their own GPUs (unlike IREN, which buys and resells GPU compute), so KEEL is not exposed to NVIDIA generational risk or chip-cycle margin pressure. KEEL is selling power, real estate, and cooling, the parts of the stack that compound rather than depreciate.
Specifically: the power-landlord tier, not the compute tier
Sitting in Tier 2 is the deliberate strategic choice. Tier 1 (chip design) is a duopoly; Tier 3 (GPU cloud) is fiercely competitive with thin margins and high reinvestment needs; Tier 4 (model labs) burns cash on training. Tier 2 owns the genuinely scarce resource and rents it on long contracts to creditworthy counterparties. The hyperscalers want the floor space, the cooling, and the power; they want to bring their own silicon. That is precisely what Keel sells.
A balance-sheet story, in transition
The Q1 2026 income statement is, by design, terrible. Revenue of $37M is down 23% year-over-year (Bitcoin mining is being wound down), the operating loss is $98M, the net loss from continuing operations is $128M, and adjusted EBITDA is negative $17M. The company has accelerated depreciation as it transitions and is taking the losses now, ahead of the pivot's revenue ramp. The right way to read this quarter is as the trough, not the run-rate.
| $M | Q1'25 | Q1'26 ● | FY26E | FY27E (lease case) |
|---|---|---|---|---|
| Revenue | 48 | 37 | ~150-200 | ~400-600 |
| Adj. EBITDA | ~+5 | -17 | ~-50 to -20 | ~+100 to +200 |
| Liquidity | ~$0.3B | $533M | $300-400M | $200M + lease cashflow |
| Energized MW (HPC) | 0 | 0 | ~50-100 | ~478 |
Q1'26 figures from the 11 May 2026 release. Forward columns illustrative build pending, engine will compute. FY27E lease case assumes Panther Creek (350 MW) signs in 2H 2026 with 2027 commencement, plus partial revenue from Sharon and Moses Lake.
Method: a per-MW comparable. KEEL trades at ~$3M per 2027 MW on its 478 MW near-term capacity; peer transactions (Cipher-AWS, BTIG's reference on CORZ) sit at $7-10.7M per MW. Pre-lease, the discount is justified by execution risk; post-lease, it should compress. Pre-lease fair value of ~$4.50-5.50 (at $4-5M/MW, halfway to peers) brackets the current price and matches the $5.40 consensus. Post-lease fair value of ~$8-10 (closer to peer multiples on 478 MW) matches Alliance Global's $8 target. The asymmetry is the trade: limited downside at current levels if the lease slips, ~60-90% upside if it signs. Cantor Fitzgerald is the bearish outlier at $3, reflecting the real risk that the pivot fails to monetize the secured power.
The valuation gap is the entire bull case
The gap is real and it is the cleanest single argument for owning KEEL: the market is pricing the megawatts at roughly one-third of where comparable assets transact. The bear interpretation is that KEEL has not yet proven its megawatts are equivalent to peers' megawatts (no signed leases, no demonstrated AI customer base, history of Bitcoin operations). The bull interpretation is that the gap is execution-discount, not asset-discount, and it closes the moment a lease lands.
Sell-side cautiously bullish; one named hedge fund is loading up
Two reads to combine here. First, the analyst tape: most of the desks have raised targets after Q1, with the bullish spread anchored by Alliance Global at $8 (citing the 478 MW in active permitting) and the bearish anchor at Cantor's $3 (citing the unproven pivot and ongoing burn). Average sell-side target ~$5.40 sits at the current price, meaning the upside in consensus is the post-lease re-rate, not a near-term price move.
Second, and arguably more informative: Situational Awareness LP, Leopold Aschenbrenner's fund (the same fund we noted in the TE brief), is publicly long KEEL as part of a broader basket across the Bitcoin-to-AI infrastructure pivot (also holding Applied Digital, Bitdeer, CleanSpark, Core Scientific, HIVE, IREN, Riot). This is a fund whose entire thesis is the AI build-out, and it has chosen KEEL specifically over Cipher (exited) and Hut 8 (exited), trimming Core Scientific. The signal is selective, not basket-buying. Combined with the analyst-driven re-rating, the bullish positioning is real, named, and current.
Three things, in order of probability. (1) Lease slippage, the live risk. Permits land mid-to-late summer; the bull case requires a colocation signing by Q3-Q4 2026. If those negotiations drag into 2027, the cash burn extends, dilution becomes more likely, and the multiple compresses on lost time. The Cantor downgrade to $3 prices this risk explicitly. (2) Dilution, the structural risk. Pre-revenue infrastructure-builds typically issue equity, and even with $533M of liquidity, a 350 MW Panther Creek build requires capital. The market reaction to issuance can be ugly on a stock with beta 3.9. (3) The pivot itself, the tail risk. KEEL's own 10-Q discloses that the transition could fail outright: supplier risk, project delays, customer concentration once leases sign, residual Bitcoin exposure on the unencumbered BTC ($197M of the liquidity). The path is sound; the execution risk is genuine.
Prototype Skyway Coverage brief · facts current to 27 May 2026 (Q1'26 results, post-Q1 analyst actions, peer comp valuation references, named institutional positioning all sourced). Forward estimates and fair-value derivation illustrative pending the quant engine. Not investment advice. The brain advises, the trader decides.
Company & opportunity first; the Skyway edge framework is context, not the verdict. Prototype. Illustrative figures not yet engine-computed.