QumulusAI (QMLS) delivered a quarter that was strong on its own terms but short of what Wall Street expected. The neocloud provider reported a loss of $0.29 per share on revenue of $14.76 million for the September quarter. Consensus called for a loss of $0.27 per share and revenue of $18.0 million, so the company missed by 7.4% on earnings and 18.0% on the top line.

The miss came even though revenue more than doubled sequentially for a second straight quarter. It rose from $3.4 million in Q1 to $6.7 million in Q2 and $14.8 million in Q3, and the company put year-over-year growth at 376%. Remaining performance obligations nearly tripled to $515.0 million from $173.1 million. That is the central tension: demand is clearly not the problem, but the Street had modeled a steeper ramp than the company's energized capacity could bill.

The quality of the quarter deserves scrutiny in both directions. GAAP results were noisy. The company reported a GAAP net loss of $14.4 million, or $0.43 per share, which differs from the $0.29 normalized figure. That loss reflects two large offsetting items:

• a $26.4 million non-cash charge for shares issued for professional services tied to the direct listing;
• $32.3 million of fair-value gains on the convertible note, the related option and the warrant liability.

The operating loss widened to $41.6 million, driven by the listing charge and a $13.4 million jump in depreciation and amortization. Adjusted EBITDA turned positive for the first time at $3.7 million, a 25% margin, versus roughly -$0.8 million in Q2. That milestone matters, but investors should note what the measure excludes: amortization and interest on finance leases. The company expects to pay about $13.0 million on those leases in the remainder of 2026 and $53.3 million in 2027. Similarly, nine-month operating cash flow of $54.3 million was flattered by a $79.0 million build in deferred revenue, which reflects customer prepayments rather than recurring cash generation.

The conference call reinforced the demand story. Several commercial metrics improved:

• 2026 signings reached $529 million, up from $282.5 million at Q2.
• Two repeat customers committed $241 million to three-year take-or-pay Blackwell B300 contracts.
• Pricing on new contracts improved to $19-21 million per MW from $18-20 million.
• Customer deposits on Q3 signings totaled $71 million.
• The 8 MW of capacity promised last quarter is fully active and billing.
• Of the backlog, $29.1 million is scheduled for Q4, which implies roughly another sequential doubling.

Management's tone was noticeably more confident, captured in the line "we're just getting started."

Bears have legitimate ammunition on execution and the balance sheet:

• Capacity: contracted but not-yet-billing capacity is only 5.45 MW. The 3.75 MW Metro Atlanta site slipped to Q2 2027, Columbia, South Carolina adds 0.5 MW, and the new 1.2 MW in Denver is still unsold. Together with the 8 MW active, that leaves the July target of 18 MW by year-end out of reach, and management offered no update.
• Guidance: no formal outlook was provided, and the earlier $300 million ARR goal was not reaffirmed in Q2.
• Leverage: debt and finance leases climbed to $248 million. Net interest expense rose to $5.4 million from $2.0 million, and both interest and depreciation step up again in Q4.
• Financing: funding for one September B300 deal is still in process, with only a 20% prepayment received.
• Supply: colocation operators are reportedly refusing non-investment-grade tenants.
• Concentration and regulation: the customer base remains concentrated in AI-native inference platforms, and regulatory risk hangs over larger sites planned for 2027-28.

On the positive side, the company expanded financing without issuing new equity. It added a $51.5 million loan, made another USD.AI draw and enlarged its TFC lease facility to $123.6 million. Shareholders' equity also recovered to $25.3 million from a slight deficit at Q2.

The market had already turned decisively skeptical before the report. Shares closed at $3.13 on the eve of earnings, down 51.6% from the $6.47 open after the August report and 48.1% below the 200-day moving average of $6.03. The stock peaked at $8.40 on September 23, then slid to a 52-week low of $2.58 on November 5 before bouncing about 21%. Investor sentiment sits at -0.30, moderately negative, and the AVWAP trend remains negative. In other words, much of the disappointment over capacity timing and leverage appears to have been priced in during the slide, though a revenue miss of this size gives skeptics little reason to step back.

The bottom line is that QumulusAI's problem is converting demand into energized capacity fast enough, not finding demand. A $515 million backlog, improving pricing and a first positive adjusted EBITDA quarter support a credible growth story. The 18% revenue miss, the Atlanta delay, the unaddressed 18 MW target and rising interest and lease obligations show how much depends on execution and financing. Until capacity additions catch up with signings, the stock is likely to trade on deployment milestones rather than backlog headlines.