AI Infrastructure
Runware plans to mass-produce AI data centers on factory assembly lines
The infrastructure startup is building a second factory in Romania to manufacture modular compute pods at automotive-style scale, targeting 1 gigawatt of capacity monthly by 2028.
TechCrunch — AI
Runware, an AI infrastructure builder, is pursuing an ambitious manufacturing vision: assembling data center pods on a factory production line the way automotive companies build cars. The company, which recently introduced the Sonic Inference Pod—a prefabricated modular data center—now wants to scale this approach into a full industrial operation.
Scale and Production Targets#
Co-founder and CEO Flaviu Fadulescu outlined plans to ramp production at Runware's Romania-based facility to 30 megawatts of capacity daily by 2028, translating to roughly 1 gigawatt per month. To contextualize this ambition: the United States was expected to add approximately 10 gigawatts of data center power in a single year. At full capacity, Runware would therefore generate in twelve months what the nation adds annually—a striking demonstration of how much the industry is pushing for faster deployment cycles.
While modular data center pods aren't new—competitors including Crusoe and Schneider Electric already offer similar products—Runware contends its manufacturing approach will enable faster delivery than rivals.
Manufacturing Strategy and Timeline#
The company is building a second factory facility that should open in early 2026, complementing an existing production site it has operated for several years. Production will begin ramping next year and follow a continuous assembly-line model. The company aims to commission pods within 48 hours of delivery to customer sites, though this timeline covers only installation and activation, not manufacturing itself. As production efficiencies improve, build time will shrink from the current three-week cycle to just one day.
Customers have several options: they can provide their own installation sites, arrange for Runware to identify one, rent compute power directly from the company, or supply their own GPUs and pay a monthly operational fee. Power and connectivity are billed separately.
Customization and Supply Chain Advantages#
Despite standardized designs, pods can be tailored to customer specifications and equipped with the latest GPU chips. Runware sees timing as a strategic edge—GPU lead times often exceed their pod delivery windows, allowing the company to match infrastructure with actual available hardware. The company's in-house hardware team can also design and build custom components for new GPU generations, enabling rapid adaptation.

