Runware, an AI infrastructure startup, is pioneering a new approach to data center production by aiming to build modular AI data centers on an assembly line, similar to automobile manufacturing. After unveiling its Sonic Inference Pod earlier this year—a prefabricated, ready-to-install data center—Runware now plans to scale up production dramatically from its Romania-based facility. The company intends to reach a manufacturing capacity of 30 megawatts per day, translating to almost 1 gigawatt monthly by 2028, which roughly matches the entire U.S. data center power capacity that’s expected to be added this year. This ambitious scaling aligns with a global industry push toward faster, more efficient data center deployment to meet AI’s growing computational demands.

Runware distinguishes itself by having already built a factory for pod production, unlike competitors like Crusoe and Schneider Electric that offer designs or limited prefabricated solutions. The new factory is expected to come online early next year after the completion of a second facility that began construction earlier this year. Using an assembly-line process, pods can be manufactured within a day—a significant improvement over the previous three-week build time. The modular pods are customizable to client specifications and updated to accommodate the latest GPU hardware, offering flexibility that matches the rapid evolution of AI technology. Customers can either provide their own deployment sites or use locations sourced by Runware, with pods designed to go online within 48 hours of installation.

Financially, Runware has secured $50 million in Series A funding as of December 2025 and plans to raise a Series B to support its factory scale-up. CEO Flaviu Radulescu explains that the company’s cost structure benefits from efficient factory operations, experienced staff sourced from the automotive industry, and the use of standard, widely available components rather than proprietary parts. Despite these advantages, challenges remain, such as supply chain dependencies and securing components including cooling systems from Germany and networking equipment from Nvidia. Nonetheless, Runware reports having over $1 billion in committed interest for its pods from customers within just a month of offering them commercially, reflecting strong demand.

Environmental considerations are part of Runware’s design philosophy. The pods utilize closed-loop liquid cooling and air-cooled radiators, needing only power and internet connectivity, without requiring water connections. The company emphasizes energy efficiency by using renewable power sources wherever possible and deploying data centers on smaller existing sites to minimize grid strain and community disruption. As the new factory gears up, Runware envisions modular AI data centers becoming the industry norm, allowing hyperscale deployments to be planned and operational within months rather than years. This innovation could redefine how AI infrastructure is built and deployed globally.