QumulusAI Secures $124 Million in GPU-as-a-Service Agreements

QumulusAI secured more than $124 million in customer subscriptions through three-year GPU-as-a-service agreements with Hyperbolic and another leading AI inference platform. The deals include nearly $21.9 million in combined upfront customer commitments.

To fulfill these agreements, the company is deploying 1,280 NVIDIA Blackwell GPUs across 160 Lenovo and Supermicro bare-metal servers. The cluster fabric is powered by Cisco Nexus One and utilizes B300 and B200 GPUs. These workload-optimized deployments are designed to reduce AI inference costs by approximately 20% compared to standard architectures by rightsizing memory, storage, and CPU cores.

The infrastructure targets high-throughput, low-latency applications, including automated coding systems and deep-research agents. Mike Maniscalco, CEO of QumulusAI, stated that AI infrastructure can no longer rely on one-size-fits-all designs. Jasper Zhang, CEO of Hyperbolic, noted that the partnership provides the performance and scalability necessary for customers building AI at scale.