Runware plans assembly-line factory to mass-produce modular AI data center pods

Runware wants to build AI data centers on an assembly line

AI infrastructure company Runware plans to scale a factory production line that assembles its modular data center pods much like cars, TechCrunch reports. The company, which unveiled its Sonic Inference Pod — a factory-built modular data center shipped ready for installation — a few months ago, says it will gradually raise output starting next year from its operations in Romania.

Runware is not the first to pursue prefabricated data centers. According to TechCrunch, Crusoe and Schneider Electric sell similar pods, and Nvidia earlier this year announced its Vera Rubin AI factories and the DSX POD to make AI data center construction faster and easier. Runware's distinguishing claim is that it has actually built a factory rather than only designing infrastructure for others to build.

An ambitious production target

Co-founder and CEO Flaviu Fadulescu told TechCrunch the company aims to reach 30 megawatts of data center capacity per day at its Romanian factory by 2028 — nearly a gigawatt per month. TechCrunch notes that at that pace Runware would produce roughly as much capacity in a year as the entire U.S. was expected to add in data center power this year, about 10 gigawatts.

Runware already has two factories: one it has used for several years to build pods, and a second it began constructing earlier this year that is expected to be finished within months and come online early next year.

Faster delivery and a flexible model

Once production ramps, Runware says a pod that used to take three weeks to build will take a single day. The company aims to bring delivered pods online within 48 hours, though that figure covers installation and setup, not construction. Customers can supply their own site or have Runware find one.

The commercial model is flexible: customers can rent compute from Runware, or bring their own GPUs and pay a monthly fee covering the pods and operations. Power and internet are billed separately, and failed components are replaced with spare parts.

Pods share a standard design but can be configured to customer requirements and the latest GPU chips. Fadulescu argued that long GPU lead times work in Runware's favor, since pod delivery is faster than chip delivery, leaving time to adapt the infrastructure to the hardware ordered. For new GPU generations, an in-house hardware team can design, build and test components itself.

Execution is the open question

As TechCrunch points out, these are plans until proven, and manufacturing at scale has humbled well-funded startups before. Electric vehicle maker Arrival, once valued at more than $13 billion, struggled to mass-produce vehicles cost-effectively compared with traditional plants. Runware has raised more than $60 million to date and plans a Series B, but for now it is finishing and launching the factory on its $50 million Series A raised last December.

Fadulescu said Runware's cost structure differs from Arrival's, with factory staff costs small relative to the value of deploying AI compute, and that designing its own systems with widely available, affordable parts keeps costs down. The company has also brought in an experienced plant manager from the automotive industry to work alongside its chief of hardware. Sourcing remains a risk: some parts, such as pumps and fans, come from Germany, and Runware still depends on Nvidia for networking.

Demand, at least on paper, appears strong. Runware claims $1 billion in committed interest for its Sonic pods since it began offering them to outside customers a month ago, which Fadulescu said represents over 20 megawatts the company is now working to deliver.

Power, siting and cooling

Fadulescu said the pods need only a power supply and an internet connection. They use a closed-loop liquid cooling system with air-cooled radiators to release heat, so no water hookup is required, and they run silent up to 86 degrees Fahrenheit. Runware is working with power providers on a site pipeline covering 397 megawatts, with potential to expand to 1.2 gigawatts, and says it targets sites much smaller than the campuses used by hyperscalers like Amazon, Microsoft and Google — sites that typically already have power available. Renewable energy is used when possible, and the company aims to make renewables its main energy source.

After bringing the new factory online early next year, Runware plans to expand to the U.S. and other regions. Fadulescu predicted modular systems will become the norm for large AI deployments, with customers able to plan hyperscale facilities around factory-built pods and have them working in months rather than years.

Why it matters

The AI buildout is constrained not only by chip supply but by how quickly data center capacity can be sited, powered and constructed. Runware's bet is that applying car-style mass production to data centers can compress that timeline from years to weeks, and its 30-megawatt-per-day target would, if reached, match an entire year of expected U.S. data center power additions on its own. The reported $1 billion in committed interest suggests real appetite for modular approaches among buyers. But the plan concentrates enormous execution risk — labor, inventory, supply chains and capital — in a domain where even richly funded manufacturing startups have failed, and the outcome will test whether AI infrastructure can really be built the way cars are.