QumulusAI (NASDAQ: QMLS) has signed a three-year agreement valued at more than $71 million to supply NVIDIA Blackwell B300 and B200 GPU capacity to an unnamed AI inference platform provider. The deal, which includes renewal options, ranks among the company’s largest customer commitments announced to date and underscores growing demand for dedicated high-performance compute in production AI workloads. For investors, this agreement signals continued momentum in QumulusAI’s demand-led deployment model, adding significant contracted revenue visibility to its book of business.

The customer operates a platform that helps companies deploy large language models, vision models, speech models, and other AI applications with low latency and high reliability. Its clients specialize in LLMs, image generation, and video generation. The dedicated Blackwell capacity will provide the platform with the high-performance compute necessary to serve production inference workloads as demand from its customers scales. The agreement adds more than $71 million in contracted, multiyear commitments to QumulusAI’s total book of business.

Capacity under the agreement will be served from QumulusAI’s U.S. data center footprint and is expected to be ready for customer use in the third quarter of 2026. The company utilizes a demand-led deployment model that places capacity into available pockets of power across a distributed network of colocation and owned facilities. This approach enables QumulusAI to bring GPU capacity online in months rather than years, addressing the urgent infrastructure needs of AI enterprises.

"Inference is where AI meets the real world, and the platforms serving it can't afford to wait on capacity," said Michael Maniscalco, CEO of QumulusAI. "Our customer runs production workloads for companies that need the right combination of flexibility, access, cost, trust and speed in their infrastructure. A three-year commitment of this size reflects what our model is built to do: procure and deploy Blackwell capacity where demand already exists, and do it fast."

Recent Contract Momentum

This agreement extends a series of recent demand announcements made by QumulusAI since early June. The company has now announced more than $246 million in customer agreements during this period. Key recent deals include:

The rapid succession of these contracts highlights the scalability of QumulusAI’s distributed AI cloud platform. By combining rapid deployment with flexible private cloud infrastructure, the company aims to offer customers a faster, more adaptable path beyond the capacity constraints of traditional centralized and hyperscale cloud models.

What the Numbers Show

The concentration of large-value contracts over a short timeframe suggests strong market validation for QumulusAI’s niche focus on inference workloads. With more than $246 million in announced agreements since early June, the company is demonstrating its ability to secure multi-year commitments from diverse customer segments, including specialized inference platforms and global cloud marketplaces. This pipeline provides near-term revenue visibility while the company executes on its deployment timeline, with the latest $71 million deal coming online in Q3 2026.

QumulusAI, a neocloud infrastructure provider purpose-built for the AI computing era, has signed a two-year, take-or-pay agreement valued at more than $18 million to supply NVIDIA Blackwell B300 nodes to a GPU cloud marketplace used by AI teams across more than 100 regions worldwide. The contract makes QumulusAI a core Blackwell supplier to the customer, solidifying its position in the AI infrastructure market. Capacity under the agreement will be served from QumulusAI's active U.S. data center footprint, with initial deployments expected to come online this summer.

The company utilizes a demand-led deployment model to place capacity into available pockets of power across a distributed network of colocation sites. This approach enables QumulusAI to bring GPU capacity online in months rather than years. The new agreement adds a two-year, take-or-pay commitment to the company's book of business, complementing its existing contracted and near-term pipeline which includes more than $124 million in three-year inference agreements announced earlier this year.

"Our customers are building the AI economy, and they choose their suppliers carefully," said Mike Maniscalco, CEO of QumulusAI. "A two-year commitment for our Blackwell capacity says it plainly. The demand is real, and we are here to deliver." Maniscalco previously emphasized that meeting AI demand is a test of access, speed, and flexibility, noting the company's ability to source across multiple OEM partners such as Supermicro and Lenovo.

To support this and other agreements, QumulusAI has expanded its hardware inventory significantly. The company recently purchased 1,632 NVIDIA Blackwell B300 GPUs, delivered across 204 NVIDIA HGX B300 systems, marking one of its largest single capacity expansions. The purchase was funded primarily through financing arrangements with Technology Finance Corporation and USD.ai. Additionally, the order included 192 NVIDIA RTX PRO 6000 Blackwell GPUs to provide customers with workload-matched options spanning training, inference, and visualization.

This expansion reflects a period of rapid growth for QumulusAI. The company's deployed GPU fleet has expanded by more than 450% from June 2025 to June 2026. By converting signed, multiyear customer demand into deployed compute, QumulusAI aims to address the widening gap between AI infrastructure demand and supply.

Order Breakdown