KEEL — Keel Infrastructure Corp.

Investment Read as of 2026-07-17

The Read

KEEL is a MIXED investment. While AI growth confirms massive underlying power demand, the recent signals highlight significant execution risks—specifically regulatory hurdles and local utility constraints—that KEEL must overcome to prove its pivot.

Bull case

• AI compute demand is forcing utilities to rethink generation plans across the US BofA says, validating the need for KEEL's large power pipeline.
• The industry is seeing clear revenue opportunities, as shown by Elisa identifying data center potential in Finland Elisa identifies....
• Major tech players are actively improving connectivity, with Microsoft adopting advanced beam technology to streamline AI infrastructure Microsoft taps 3M’s expanded beam tech....

Bear case / what breaks it

• The thesis remains unproven: KEEL has not yet signed a credible anchor tenant, which is the primary hurdle for converting its pipeline into revenue.
• Regulatory and environmental opposition pose clear risks; groups are petitioning to halt data center applications Coalition of environmental action groups..., and protests have already occurred at sites.
• Operational constraints are rising; planned data centers are being held back by issues like lack of available water Planned data center in Florida held back..., suggesting local utility capacity is a major bottleneck.

What the latest signal says

The signals confirm that while AI demand is driving infrastructure needs (rare earth, power), the immediate battleground has shifted from pure compute capacity to securing reliable water and navigating complex local regulations Planned data center in Florida held back.... KEEL must prove it can solve these physical, local constraints to validate its pipeline value.

Posterior history

Thesis detail

Core thesis

The earliest-stage, most speculative of the crypto-to-AI names. Keel (formerly

Bitfarms Canada) is repositioning as a North American digital + energy

infrastructure firm, with a ~2.2 GW power pipeline across Pennsylvania,

Washington, and Québec and ~$533M liquidity (cash + bitcoin). The bet: turn that

power pipeline into leased AI/HPC data-center capacity. The story is mostly

pipeline and balance sheet right now — it has not yet proven it can sign anchor

AI tenants, and Q1 showed real cash burn (revenue ~$37M, down year-on-year; net

loss ~$145M). Treat as a high-risk option on the pivot, not a proven operator.

Pillars (with priors)

1. The 2.2 GW power pipeline converts into leased AI/HPC capacity · P = 0.50

2. Signs at least one credible anchor AI/HPC lease to prove the pivot · P = 0.45

3. ~$533M liquidity funds the buildout without heavy dilution · P = 0.50

4. Geographic spread (PA/WA/QC) de-risks siting, power, and permitting · P = 0.50

Expected news (the prior)

• Anchor lease / customer announcements; interconnect & site milestones along the

2.2 GW pipeline; financing or dilution; quarterly cash burn vs. liquidity.

Residual = no anchor lease lands, cash burn forces a dilutive raise, or pipeline

sites stall in interconnection queues.

Thesis-breaking triggers (→ set P near 0)

• ☐ A full year passes with no credible anchor AI/HPC lease
• ☐ Cash burn forces a large, dilutive raise
• ☐ Key pipeline sites stall in interconnection/permitting
• ☐ Liquidity falls sharply (bitcoin drawdown + continued losses)

Leading vs lagging indicators

• Leading: lease signings, interconnect/site milestones, financing terms
• Lagging: leased revenue, cash burn rate, liquidity, share price

Key metrics

• Pipeline (GW) with firm interconnect vs. early-stage · anchor leases signed ·

liquidity ($ cash + BTC) · quarterly cash burn · dilution

Valuation anchor

Optionality on the power pipeline against the balance sheet — value the GW

pipeline at a per-MW data-center figure, discounted heavily for execution and the

fact that no anchor lease is yet signed. Most fragile of the three; highest beta.

Cross-arena sensors

B1 (AI compute demand), B2 (power/grid/interconnect), B5 (equities/crypto).

Posterior log

• 2026-07-17 · P 0.50→0.47 ↓ · MAINTENANCE · Stabilizing on borrowed strength — the bounce from 0.40 back to 0.50 is driven by a peer comp (CleanSpark's 20-year, $6.6B data-center deal)
• 2026-07-17 · P 0.45→0.50 ↑ · UP · Surprise: HIGH · https://www.datacenterdynamics.com/en/news/cleanspark-secures-tenant-with-20-year-lease-for-data-center-in-sandersville-georgia/
• 2026-07-14 · P 0.40→0.45 ↑ · UP · CleanSpark (a bitcoin miner) signed a 20-year, $6.6B-contracted-revenue data-center lease in Sandersville, GA — a direct comp proving a former miner can land a · https://www.datacenterdynamics.com/en/news/cleanspark-secures-tenant-with-20-year-lease-for-data-center-in-sandersville-georgia/
• 2026-06-26 · P 0.43→0.40 ↓ · MAINTENANCE · Weakening — two straight downgrades (0.50→0.47→0.43) on confirmed dilution, and the 2026-06-16 auditor change adds a governance flag on top
• 2026-06-19 · P 0.47→0.43 ↓ · MAINTENANCE · Weakening — the June filing cluster confirms the residual risk in the thesis (a dilutive raise) is happening: they sold unregistered equity
• 2026-06-12 · P 0.50→0.47 ↓ · MAINTENANCE · Weakening, slightly — the 2026-06-10 filing cluster confirms dilution ("sold unregistered equity") plus new debt, which is the residual risk
• 2026-05-28 · created · P 0.50 · — · AI-Energy-Thesis-Scaffold