Runware targets mass production to accelerate AI data center deployment

The company, which has raised over $60 million to date, operates a manufacturing facility in Romania and is preparing to bring a second site online early next year. CEO Flaviu Radulescu contends that by controlling the hardware design and utilizing standard components, Runware can adapt its infrastructure to the latest GPU releases faster than traditional construction firms. This approach addresses the persistent bottleneck of GPU lead times, allowing the firm to integrate components just as they become available.

While the model promises efficiency, the startup faces significant logistical hurdles. Industry analysts point to the cautionary tale of the electric vehicle manufacturer Arrival, which collapsed after failing to achieve cost-effective mass production. To mitigate these risks, Radulescu has recruited a plant manager from the automotive sector to oversee the transition to high-volume assembly. The company claims to have already secured $1 billion in committed interest for its Sonic pods, representing an initial delivery pipeline of 20 megawatts.

Runware’s pods utilize closed-loop liquid cooling, a design choice intended to minimize water usage and reduce the environmental footprint compared to hyperscale facilities. By deploying these modular units on sites that already possess grid access, the company hopes to avoid the community resistance often associated with large-scale data center projects. Following the launch of its second factory, Runware plans to expand its manufacturing footprint into the United States and other international markets.