Modular infrastructure company Moonshot Energy, GPU-as-a-service provider QumulusAI, and Connect Nation Internet Exchange Points (IXP.us) have unveiled a joint initiative to create QAI Moon Pods at 25 sites across the country, with plans to scale up to 125 cities. This collaboration aims to establish a national infrastructure for inference and AI workloads, leveraging carrier-neutral interconnection, modular AI infrastructure, and GPU-as-a-service to extend AI compute capabilities beyond traditional data center boundaries, reducing latency effectively.
The initial deployment is scheduled to commence in July 2026 at Wichita State University in Kansas, with expansion to 25 additional cities in the pipeline. By identifying potential sites at US university research campuses and municipalities, the companies are striving to establish a widespread network of AI pods to meet the growing demand for AI compute at the edge.
According to IXP.us co-CEO Hunter Newby, the focus is on building internet exchange points and AI models rather than conventional data centers, with a swift deployment timeline of several months for the AI pod buildouts. This approach emphasizes the importance of agility and responsiveness in meeting evolving AI infrastructure needs.
Modular Approach
The collaborative effort involves the design and construction of 2,000 KW modular units by Moonshot, complemented by QumulusAI’s GPU infrastructure expertise. By offering low-latency AI compute capabilities at the edge without the limitations associated with large hyperscale data centers, the AI pods are poised to revolutionize the accessibility and efficiency of AI processing.
Ethan Ellenberg, CEO of Moonshot, emphasized the significance of this partnership in integrating power, compute, and interconnection at the intersection of rising AI demand. The strategic alignment of resources and expertise is geared towards meeting the evolving needs of various industries requiring efficient AI processing at a local level.
Steven Dickens, CEO and analyst at HyperFrame Research, underscored the importance of bringing inference capabilities closer to the edge to address underserved areas. With a projected surge in GPU deployment for inference applications in diverse sectors such as computer vision, smart retail, manufacturing, and healthcare, there is a growing need for accessible AI infrastructure with minimal latency.
Inference Market Play
The collaboration between Moonshot Energy, QumulusAI, and Connect Nation Internet Exchange Points aims to bridge the gap in inferencing demand by offering GPU compute capabilities at the network edge. Mike Maniscalo, CEO of QumulusAI, highlighted the shift towards inference-driven, latency-sensitive workloads that require scalable and practical AI compute solutions beyond traditional data centers.
This strategic partnership is focused on making high-performance AI compute more accessible and economically viable outside of hyperscale data centers, catering to the increasing demand for distributed and latency-sensitive workloads. By placing GPU compute resources directly at the network edge, the companies aim to enhance the efficiency and scalability of AI processing for a wide range of industries.
