QumulusAI (NASDAQ:QMLS) reported second-quarter revenue growth and an expansion in its deployed GPU fleet as the recently public AI infrastructure provider focused on bringing customer-contracted compute capacity online.
Revenue for the second quarter was $6.7 million, up 118% from $3.1 million a year earlier. Compute revenue rose 328% to $5.6 million and represented 84% of total revenue, compared with 43% in the prior-year period, according to Chief Financial Officer Scott Krosnowski.
Contracts and Capacity Buildout
QumulusAI signed 21 new direct customer contracts totaling $169.7 million during the quarter. Total signed contract value reached $282.5 million across 40 contracts, with a weighted average term of 2.2 years. The company reported remaining performance obligations of $173.1 million as of June 30, a narrower accounting measure covering direct customer compute contracts signed as of quarter-end.
Maniscalco said direct customer relationships accounted for more than 96% of recurring revenue, compared with less than 10% a year earlier, as the company shifted away from dependence on a single marketplace and toward direct multiyear agreements.
The company said it has 8 megawatts of compute capacity that is fully sold, with final GPU deployments underway. It also has long-term leases on sites in Oklahoma and Texas totaling roughly 39 megawatts of available power, in addition to its existing high-performance computing data center lease footprint.
Last week, QumulusAI announced a Metro Atlanta colocation agreement for an initial 3.75 megawatts. The agreement includes a right of first offer for up to 7 additional megawatts at the same site, potentially bringing the Atlanta location to 10.75 megawatts.
Management’s July guidance contemplated 18 megawatts of HPC capacity by year-end, including 8 megawatts already in place and 10 megawatts to be developed. Maniscalco said the Atlanta agreement represents the first 3.75 megawatts of that planned additional capacity and that the company continues to pursue smaller sites that can be ready for service this year.
“We’re not currently demand constrained,” Maniscalco said in response to an analyst question. He said the primary near-term constraint on growth is securing land, power and shell capacity to support additional deployments. The company did not reaffirm its previously issued $300 million annual recurring revenue target.
Margins, Losses and Cash Flow
Gross profit rose to $4.5 million from $1.7 million a year earlier, while gross margin expanded to 66.6% from 55.1%. Krosnowski said margin also improved from 37.5% in the first quarter as GPU activations increased faster than associated HPC colocation costs. The quarter also included approximately $0.4 million in curtailment credits at the company’s Oklahoma site for returning power to the grid during peak-demand periods.
Operating loss widened to $7.7 million from $2.2 million in the year-ago quarter. Depreciation and amortization totaled $6.9 million, versus $1.1 million a year earlier, reflecting HPC assets entering service. Adjusted EBITDA was a loss of $0.8 million, improving sequentially from a $2.8 million loss in the first quarter.
Net loss was $22.8 million, compared with net income of $12.1 million a year earlier. The company said both periods were affected by non-cash accounting items. The prior-year result included a $14.5 million non-cash gain tied to the remeasurement of its investment in The Cloud Minders, while the current quarter included a $19.2 million non-cash loss associated with convertible notes, partly offset by $6.2 million in non-cash fair-value gains on the notes and related option.
For the first six months, operating cash flow was positive $22.3 million, compared with $0.8 million used in the prior-year period. Krosnowski attributed the improvement to customer prepayments, with deferred revenue increasing by $30.5 million. The company said major multiyear contracts typically include prepayments ranging from 10% to 35%.
Cash and restricted cash totaled $39.9 million at quarter-end, up from $11.7 million at year-end. The total included $19.9 million that had been restricted pending the company’s direct listing and became available after QumulusAI began trading on Nasdaq on July 16.
Deployment Strategy and Financing
Maniscalco said QumulusAI is pursuing a “hyper-speed” deployment model centered on existing or near-ready pockets of power and colocation capacity, generally ranging from 2 megawatts to 50 megawatts, rather than multiyear, gigawatt-scale data center construction projects.
The strategy is designed to serve customers seeking near-term compute availability, including AI-native companies experiencing rapid demand growth as well as enterprise customers. Maniscalco said enterprise customers are part of the company’s pipeline but generally make decisions more slowly than AI-native customers.
QumulusAI was approved as an NVIDIA Cloud Partner on July 17, one day after its Nasdaq trading debut. Management said the partner status supports access to the supply chain, while public-market access supports funding for deployments.
Krosnowski said QumulusAI’s most recent Blackwell contracts generate annualized revenue of between $18 million and $20 million per megawatt, compared with roughly $16 million per megawatt across its installed base. He attributed the difference to pricing power, a focus on Blackwell chips and customer willingness to pay for capacity that can be activated quickly.
All capital expenditures scheduled for the company’s 8 megawatts of sold capacity have been ordered and financed, Krosnowski said. Most GPU financing arrangements run for three years, though some are longer. The company expects its cost of capital to decline as it expands deployments and seeks lower-cost financing options.
About QumulusAI (NASDAQ:QMLS)
QumulusAI is a cloud infrastructure company specializing in rapid deployment of graphics processing unit (“GPU”)-powered solutions for artificial intelligence (“AI”) applications, serving a critical market that is often overlooked by large-scale cloud providers (“hyperscalers”), which operate massive, standardized computing infrastructures primarily serving the largest enterprises. Our platform delivers flexible, competitively priced, and customizable solutions for underserved small and mid-market customers—including machine learning teams, AI infrastructure startups, and research institutions—while also supporting the scale and complexity requirements of large enterprises, such as long-term deployments or supplemental on-demand compute capacity.
-pt-telekomunikasi-indonesia-tbk-logo-1200x675.png){kind=logo}