QumulusAI (Nasdaq: QMLS), a neocloud infrastructure provider, said on Oct. 15 that the last NVIDIA B300 units under its USD 71.9 million three-year Blackwell agreement have been handed off and are now in production.

The company began billing on that agreement 10 weeks after signing. Deployment proceeded in two stages: B200 capacity went live on Sept. 8, and B300 capacity followed at the company's Kansas City site on Oct. 1.

How the Contract Book Grew Past USD 500 Million

The company disclosed that two agreements announced last month, worth a combined USD 240.9 million, pushed its total cumulative signed customer contract value to USD 523.4 million.

That figure represents a first-time crossing of the USD 500 million threshold and an 85% increase over the USD 282.5 million reported alongside QumulusAI's second-quarter 2026 results.

The two new agreements consist of a USD 211.9 million, three-year take-or-pay deal covering 3.75 MW at the metro Atlanta site announced in August and a separate USD 29.1 million agreement to be deployed at an additional site.

Together, the signed agreements equate to roughly USD 209 million in annual recurring revenue on initial terms, before any extension options.

Approximately USD 128 million of that total corresponds to the 8 MW expected to be active by month-end, about USD 71 million is tied to the 3.75 MW in Atlanta, and the remainder relates to the USD 29.1 million agreement.

Where the 11.75 MW Stand

QumulusAI also laid out activation timing for all 11.75 MW of its contracted capacity, which spans six colocation sites. Of that total, 7 MW is currently active and producing revenue.

Another 1 MW is already energized and slated for final handoff at the end of this month, which would lift active, revenue-generating capacity to 8 MW.

The remaining 3.75 MW at the metro Atlanta facility is targeted for activation in early 2027.

Repeat Customer Deepens Commitment

The customer that took delivery under the USD 71.9 million Blackwell agreement has since returned. The USD 211.9 million agreement for the Atlanta capacity is that customer's second contract with QumulusAI, and its contracted commitments to the company now total USD 283.8 million, excluding extension options.

Management Weighs In

"Crossing half a billion dollars in signed customer contracts is a milestone worth marking, and we got there by doing exactly what we said we would do: sign, deploy, and expand," said Michael Maniscalco, chief executive officer of QumulusAI.

"Our second-quarter revenue of USD 6.7 million reflected what we had deployed, not what we had sold. Revenue follows handoff, and the handoffs are happening now.

All of our capacity is under contract, and we already have a customer signed for the next site, so our focus is on securing the next block."

The Company Behind the Numbers

QumulusAI describes itself as a distributed AI cloud platform offering accelerated access to high-performance GPU compute.

Operating an inference-first, demand-led deployment model across multiple data center sites, the company positions compute closer to customer demand so AI teams and enterprises can scale production workloads.

It emphasizes rapid deployment and flexible private cloud infrastructure as an alternative to what it characterizes as the capacity constraints of traditional centralized and hyperscale cloud models.

Risk Disclosures Attached to the Outlook

The press release flagged several forward-looking statements, including the timing of the final handoff for the remaining 1 MW of energized capacity, activation of the remaining 3.75 MW in early 2027, deployment under the USD 29.1 million agreement, the annual recurring revenue attributed to signed agreements, and the company's ability to secure additional capacity.

Risks cited include reliance on a limited number of large customers; the availability and cost of power, network connectivity, and specialized hardware such as graphics processing units; substantial capital requirements and access to financing; competition and rapid technological change in the high-performance computing and AI markets; and the company's limited operating history and record of net losses.

Further risks are set out in the "Risk Factors" section of the company's amended Form S-1 registration statement (File No. 333-292514) and its Form 10-Q quarterly report.