Moonshot Energy, QumulusAI, and Connected Nation Internet Exchange Points (IXP.us) announced a strategic commercial agreement to deploy a nationally distributed platform that colocates modular AI infrastructure directly at carrier-neutral Internet Exchange Points. The initiative targets 25 initial sites, scaling to 125 locations across U.S. university research campuses and municipalities, pairing IXP facilities with modular “AI Pods” optimized for low-latency inference workloads.
Moonshot Energy brings manufacturing-scale delivery of electrical systems, modular power, cooling, and AI-ready infrastructure to the partnership. Based in Texas, the company has focused on building standardized, repeatable infrastructure for neoclouds and hyperscalers, recently commissioning a 500,000-square-foot advanced manufacturing facility in Lewisville to support national deployments. By extending beyond manufacturing into GPU-as-a-Service operations through its joint venture with QumulusAI, Moonshot positions itself as both an infrastructure supplier and an operator of network-adjacent AI compute.
QumulusAI contributes the GPU orchestration layer, workload delivery platform, and commercial operating model designed for inference-heavy and latency-sensitive workloads. The company focuses on deploying GPU resources closer to where data is generated and consumed, addressing latency, transport cost, and data sovereignty challenges associated with centralized hyperscale data centers. Through QAI Moon, QumulusAI aligns its software and service stack with modular, power-constrained deployments that can scale incrementally based on local demand and available interconnection.
IXP.us provides the carrier-neutral interconnection foundation, with a national plan to develop Internet Exchange Points in 125 regional hub communities, primarily anchored at research universities. Structured as a joint venture between Connected Nation, Inc. and Newby Ventures, IXP.us emphasizes neutral physical infrastructure, dense network adjacency, and an operating model without monthly recurring cross-connect fees. The platform leverages existing relationships with DE-CIX for IX operations, TOWARDEX for fiber and conduit construction, and Connectbase for visibility and coordination of fiber, transport, and IP transit entering each site.
The first QAI Moon deployment is scheduled to begin by July 2026 at the IXP.us “alpha site” on the Wichita State University campus in Wichita, Kansas. This location serves as the archetype for future sites, integrating modular AI compute directly alongside dense, carrier-neutral interconnection to support real-time and inference-driven workloads in regional markets.
- National platform integrating carrier-neutral IXPs with modular, network-adjacent AI Pods
- Initial rollout at 25 sites, with a roadmap to 125 locations over five years
- Dual, geographically diverse 400G IP transit connections from four independent ISPs per site
- Redundant 400G IX ports on DE-CIXaaS switching at each IXP.us facility
- Direct adjacency to high-count dark fiber between interconnection and AI compute
- Approximately 2,000 kW initial module sizing per market, with flexible GPU configurations
- Focus on inference-driven, latency-sensitive workloads beyond centralized hyperscale data centers
“This partnership represents the physical convergence of power, compute, and interconnection at the exact point where AI demand is moving,” said Ethan Ellenberg, CEO of Moonshot. “By pairing Moonshot’s modular electrical and AI infrastructure with the IXP.us carrier-neutral interconnection model and QumulusAI’s GPU platforms, we are creating a repeatable national architecture that delivers ultra-low-latency AI without the constraints of hyperscale data centers.”
🌐 Analysis
The partnership reflects a broader industry trend toward distributing AI inference closer to networks and end users, as real-time applications place increasing pressure on latency and transport costs. By combining modular power and compute with carrier-neutral interconnection at research campuses and regional hubs, the model aligns with parallel efforts by telcos, IX operators, and regional data center developers seeking alternatives to hyperscale-centric architectures.
