Runware Wants to Build AI Data Centers Like Cars — 30 MW a Day by 2028

Runware says it will produce 30 megawatts of data center capacity per day at its Romania factory by 2028 — nearly 1 gigawatt a month, or roughly the 10 gigawatts of data center power the entire U.S. was expected to add this year, compressed into a single company's annual output.

The AI infrastructure company told TechCrunch it plans to scale an assembly-line model for its Sonic Inference Pod, a modular data center built in a factory and shipped ready to install. That pace is ambitious. It's also the clearest signal yet of an industry-wide push to deploy AI compute faster and cheaper.

What does the factory-first model change?

Runware isn't first. Crusoe and Schneider Electric already sell prefabricated pods, and Nvidia announced its Vera Rubin AI factories and DSX POD earlier this year to speed up builds. The difference, according to Runware: it has actually built a factory, not just the reference designs others would use to build one.

CEO and co-founder Flaviu Radulescu said production ramps up gradually starting next year. Runware already runs one factory it has used for years to build pods, plus a second it started building earlier this year — that one finishes within months and comes online early next year.

Speed is the pitch. Pods should go from delivery to online within 48 hours, covering installation and setup — not build time. A pod once took three weeks to build; Runware says that drops to one day as production ramps. Customers can supply their own site or have Runware find one. They can rent compute, or bring their own GPUs and pay a monthly fee for pods and operations. Power and internet are billed separately.

Chip lead times actually work in Runware's favor, Radulescu argues. "Our pod delivery times are also shorter than current GPU lead times, giving us time to adapt the infrastructure to the hardware being ordered," he said. "For new GPU generations, our in-house hardware team can design, build, and test components itself, allowing us to adapt quickly."

Can it avoid Arrival's fate?

Execution is the open question, and there's a cautionary tale: EV startup Arrival, once valued at over $13 billion, failed to mass-produce vehicles cost-effectively. Runware has raised more than $60 million to date and plans a Series B, but is finishing and launching the factory on its $50 million Series A from last December.

"Our cost structure is very different," Radulescu said. "We do have headcount and materials cost. But our factory staff is cost-effective relative to the value of deploying compute for AI." He points to in-house system design, widely available affordable parts, and a hire meant to de-risk manufacturing: "We've brought in a highly experienced plant manager from the automotive industry to work alongside our chief of hardware and help us get that execution right."

Supply chains remain a weak spot. Pumps and fans come from Germany, and Runware still depends on Nvidia for networking.

Is the demand real?

Runware claims $1 billion in committed interest for Sonic pods since it began offering them to outside customers a month ago — "representing over 20 megawatts, which we're now working to deliver," Radulescu said.

On energy, the pods need only power and an internet connection. They use closed-loop liquid cooling with air-cooled radiators, so no water hook-up. "We're working with power providers on a site pipeline covering 397 megawatts, with potential to expand to 1.2 gigawatts," Radulescu said. The sites are far smaller than hyperscaler campuses, typically already powered, and the pods run "completely silent up to temperatures of 86 degrees Fahrenheit." Renewables are the target main energy source.

Next comes the factory launch early next year, then expansion to the U.S. and beyond. Radulescu's bet is that modular becomes "the standard for large AI deployment" — letting customers "plan a hyperscale data center around factory-built pods from the beginning and have it working in months instead of years."