Keel Infrastructure (KEEL) reported a third-quarter loss of $0.10 per share, missing the consensus estimate of a $0.07 loss by 42.9%. Revenue of $27.6 million beat the $21.44 million consensus by 28.7%. Neither headline number is the real story. The most consequential development was management abandoning its target of signing three data-center leases in 2026 and declining to guide on lease timing at all. For a company whose investment case rests on converting powered land into contracted HPC revenue, that withdrawal matters more than the size of the EPS miss.
The miss itself is largely an accounting artifact. The loss from continuing operations widened to $62 million from $11 million a year earlier, driven primarily by a $55 million non-cash unrealized loss on the capped calls tied to the 2025 convertible notes. Underneath that item, the core business is contracting by design. Revenue fell from $68 million a year ago as Keel stopped U.S. Bitcoin mining in the second quarter. Sequentially, revenue slid from $37 million to $30 million to $28 million over the past three quarters. Gross margin was negative 53%. G&A doubled to $34 million on senior hiring, professional fees, and higher stock-based compensation. Adjusted EBITDA moved to negative $27 million from positive $19 million a year earlier and has worsened each quarter this year. The reported operating loss narrowed to $8 million from $18 million, but the quality of that improvement deserves scrutiny. It was helped by a $73.6 million fair-value gain on digital assets, only partly offset by a $32.2 million realized loss on sales. Keel sold 2,520 Bitcoin, its entire treasury plus the quarter's production, for $192 million. That leaves holdings nominal and largely ends treasury price exposure.
The call showed physical progress coexisting with commercial delay, and that is the central tension. On the positive side:
- Moses Lake moved from permitting to full permits and active construction, funded from the balance sheet. It is expected to be the first site generating HPC revenue in 2027, with an optional path to Vera Rubin architecture.
- Sharon's final environmental permit is now in draft.
- Contracted power at Baie-Comeau rose to 30 MW from 22 MW at no cost.
- A potential acquisition in Allentown, Pennsylvania, a fully constructed powered shell with 75 MW of 2027 power, is in diligence.
Management also said hyperscaler demand exceeds supply and that no prospective tenant has walked away. Even so, the tone shifted noticeably from last quarter's confidence to a more defensive framing. The call placed heavier emphasis on rising financing costs, yield on cost, stage-gated spending, and openness to powered-shell and colocation structures. Openness to alternative structures could widen the pool of counterparties, but it may also signal some pressure on lease economics.
The Pennsylvania picture contains a conflict investors should note. The release says revised Sharon and Panther Creek plans comply with the state's new executive order with no meaningful changes to project costs or timelines for tenants. On the call, however, management acknowledged that the order removed data centers from fast-track review and enlarged the permitting backlog. Permits once expected by mid-to-late summer now have no firm timeline. The two statements can be reconciled if delivery dates for tenants hold while permit milestones slip, but that leaves less room for further delay. Similarly, the release describes the Panther Creek and Scrubgrass expansion load studies as on track for the first quarter of 2027. The call analysis shows that the update itself slipped from a December/January expectation. Add higher interest rates, tighter debt markets, and a shift in political control of the Pennsylvania Senate, and bears have legitimate ammunition. Liquidity is still substantial at about $780 million, and management says no debt or equity financing is needed before signing a lease. But that figure is down from $819 million in August, with no HPC revenue yet contracted.
The market had already priced in much of this disappointment before the report. As of November 4, shares stood at $2.10, down 45.5% from the $3.85 open after the August report and 45.5% below the 200-day moving average. The stock fell roughly 50% from its September 22 inter-earnings high of $4.23 to a 52-week low of $1.86 on November 2, and entered the report only about 13% above that low. AVWAP readings remain negative. Investor sentiment improved modestly, from slightly negative to slightly positive, but remains close to neutral. That suggests cautious interest rather than conviction.
The bottom line is that Keel's quarter was less about the EPS miss or the revenue beat than about credibility on timing. The sites are advancing, the balance sheet is sound, and Moses Lake gives the company a tangible 2027 revenue path. But dropping the three-lease target and withdrawing timing guidance shifts the burden of proof back onto management. Until a creditworthy lease is signed, the shrinking legacy business and steady cash use will keep the stock tied to execution risk rather than the size of the opportunity.