QumulusAI, Inc. Common Stock (QMLS) Earnings
QMLS has beaten EPS estimates in 0 of its last 1 reported quarters (average surprise -105.7% over the last four).
| Report date | EPS est | EPS actual | Surprise | Revenue | Rev. surprise |
|---|---|---|---|---|---|
| Aug 25, 2026 | $-0.35 | $-0.72 | -105.7% | $7M | +107.5% |
Source: company filings + earnings calendar. For informational purposes only — not investment advice.
Earnings call summary
Q2 FY2026 · August 25, 2026
AI summary of management’s prepared remarks and analyst Q&A. For informational purposes only — not investment advice.
Management highlights
- Strategic Positioning: Cumulus AI operates as a GPU cluster operator rather than a data center developer, focusing on 'hyperspeed' deployment into existing power shells (2-50 MW) rather than building gigawatt-scale campuses. - Capacity Growth: The company more than tripled its GPU fleet in Q2. It announced a co-location agreement in Metro Atlanta for an initial 3.75 MW with rights for up to 7 MW additional, contributing to a combined footprint of ~57 MW going into 2027. - Commercial Momentum: Signed 21 new direct customer contracts totaling $169.7 million in Q2, bringing total signed contract value to $282.5 million across 40 contracts. Direct customers now account for >96% of recurring revenue. - Unit Economics: Recent Blackwell-based contracts generate $18-20 million annualized revenue per megawatt, with a blended rate of ~$16 million/MW across the installed base. - Customer Base: Transitioned away from marketplace dependence to direct multi-year agreements. Primary near-term growth drivers are 'AI natives' (VC-backed inference/model-as-a-service companies) and select enterprise clients.
Guidance
- Capacity Guidance: Management did not explicitly reaffirm the July 14, 2026 guidance of reaching 18 MW of HPC capacity by year-end, but stated they are still pursuing this target through smaller-scale available capacity. - Revenue/ARR Guidance: Did not reaffirm the previously stated $300 million ARR target. - Forward Outlook: Management indicated demand is not constrained; the primary bottleneck is securing land and power shells. They expressed confidence in the land/power team's ability to secure capacity for 2027 and 2028.
Segment performance
Revenue was $6.7 million, up 118% year-over-year. Compute power revenue grew to $5.6 million (84% of total revenue), while legacy mining/hosting/crypto mining contributed $1.1 million (16%). Gross margin expanded to 66% from 55.1%. Operating loss widened to $7.7 million due to $6.9 million in depreciation/amortization from new HPC assets coming online.
Risks & headwinds
- Supply Chain Constraints: The business is heavily dependent on securing GPUs and power infrastructure; supply constraints remain the primary operational bottleneck. - Execution Risk: The strategy relies on rapid deployment into existing power shells, which may face logistical or regulatory hurdles compared to greenfield builds. - Financial Volatility: Significant non-cash items, including a $19.2 million loss from convertible note issuance, impact reported net income, though cash flow remains supported by financing activities. - Contract Renewal Risk: A portion of the contract book comes up for renewal annually, exposing the company to potential repricing risks if GPU market dynamics shift.
Analyst Q&A
Q: Michael Donovan asked about the economics and timeline of the AI XP initiative, specifically CapEx contributions and site activation. /
A: Mike Maniscalco explained that partners provide land, power, and shell, while Cumulus rolls in racks and GPUs once ready. The first site in Wichita is currently energizing, with additional sites further out in the pipeline.
Q: Brett Knobloch queried whether the 18 MW year-end capacity target remains valid and when the $16M/MW revenue metric would fully reflect on the active base. /
A: Maniscalco confirmed they are still targeting the 18 MW goal by finding smaller pockets of available power. CFO Scott Krosnowski added that all CapEx for the initial 8 MW is financed and expected to be revenue-producing by year-end.
Q: Bill Papagnastasio sought details on customer profiles for the new direct contracts and updates on the GPU financing environment. /
A: Maniscalco identified 'AI natives' (VC-backed inference companies) as key near-term drivers, alongside growing enterprise interest. He noted improved financing options via USDI, Permian Labs, and traditional leases, emphasizing alignment of capital to assets.
Q: Papagnastasio also asked if the $300M ARR guidance held and how the demand funnel compares to previous expectations. /
A: Management declined to reaffirm the $300M ARR figure but stated demand exceeds current supply capabilities. The main constraint is securing land and power, not customer interest, with focus shifting to 2027+ capacity acquisition.
Q: Alan Klee asked about competitive differentiation and GPU financing amortization terms. /
A: Maniscalco highlighted speed and access to NVIDIA/OEM partnerships as key differentiators against neoclouds, enabling faster deployment of sub-50 MW clusters. Krosnowski noted most GPU financing is structured over three years to align with compute contracts.