A Romania-based artificial intelligence infrastructure company is pushing the boundaries of how data centers are built and deployed. Runware, which drew attention earlier this year after introducing its Sonic Inference Pod - a modular data center assembled in a factory and delivered ready for installation - is now setting its sights on something considerably more ambitious: a full-scale manufacturing facility that produces these pods along an assembly line, much the way automobiles roll off a production line.

The concept of prefabricated, modular data centers is not without precedent. Companies such as Crusoe and Schneider Electric already offer pre-built pod solutions. What Runware argues sets it apart is the speed at which it can get those units to customers and the scale at which it intends to produce them.

Gigawatt-Scale Ambitions

Flaviu Radulescu, co-founder and chief executive of Runware, has outlined a production target that, if achieved, would place the company in rare company. The plan calls for scaling output at its Romanian factory to 30 megawatts of data center capacity per day by 2028 - a rate that would translate to nearly one gigawatt per month. To put that figure in context, the entire United States was projected to add roughly 10 gigawatts of data center power capacity over the course of this year. Runware's stated goal would, in theory, match that figure within a single year of full-scale operation.

That level of ambition reflects a broader trend across the technology industry. Nvidia earlier this year announced its Vera Rubin AI factories and the DSX POD initiative, both designed to accelerate and simplify the process of building AI-ready data centers. The distinction Runware draws is that it has already built a functioning factory rather than simply designing the blueprint for others to follow.

Radulescu confirmed that the company currently operates two facilities. The first has been in use for several years and has been producing pods throughout that time. Construction on a second, larger facility began earlier this year and is expected to be completed within months, with operations anticipated to begin in early 2026. Production will scale up incrementally starting next year.

From Weeks to Hours

The assembly-line model Runware is pursuing would dramatically compress the timelines typically associated with data center deployment. Under its proposed operational framework, customers can either provide their own site or have Runware identify a suitable location. Once a site is prepared, Runware aims to bring the pods online within 48 hours - a window that covers installation and commissioning, not the manufacturing process itself.

Building a single pod once required approximately three weeks. As the company ramps up its production capabilities, Radulescu says that build time will fall to a single day. The company will also maintain a supply of spare parts to address component failures. On the commercial side, customers have two options: they can supply their own GPUs and pay a recurring monthly fee for the pods and operational support, or they can rent compute capacity directly from Runware. Power and internet connectivity are billed separately.

Customization is also part of the offering. While the pods follow a standard design, they can be configured to meet specific customer requirements and adapted to accommodate the latest GPU generations. This flexibility becomes particularly relevant given the well-documented delays in GPU supply chains. As Radulescu noted, the gap between ordering chips and receiving them actually works in the company's favor. "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."

Manufacturing at Scale: Lessons from the EV Industry

Ambitious manufacturing plans, however, come with well-documented risks. The history of technology hardware startups attempting to reinvent production at scale includes cautionary examples. Arrival, the electric vehicle startup that once carried a valuation exceeding $13 billion, became a prominent cautionary tale after struggling to produce vehicles cost-effectively using its microfactory model when compared to established automotive plants.

Runware has raised more than $60 million to date and is currently operating on the strength of a $50 million Series A round closed last December. A Series B is in the planning stages, but for now the company is using existing capital to complete and launch the new factory. Radulescu acknowledged the comparison to Arrival but argued that Runware's cost structure is fundamentally different. "We do have headcount and materials cost," he said, "but our factory staff is cost-effective relative to the value of deploying compute for AI." He added that designing systems in-house and sourcing widely available, affordable components helps keep expenses manageable.

To shore up the manufacturing execution side of the business, Runware has brought in a plant manager with deep experience in the automotive sector to work alongside its chief of hardware. The intent is to apply proven production discipline to what remains a relatively novel approach to data center manufacturing.

Supply Chain and Component Dependencies

Even with a vertically integrated approach to designing its cooling and pod systems, Runware still relies on external suppliers for certain components. Pumps and fans, for instance, are sourced from Germany, and the company continues to depend on Nvidia for networking infrastructure. These dependencies represent potential pressure points in any supply chain disruption scenario.

Despite those constraints, Runware reports strong early commercial interest. Since opening its Sonic pods to outside customers approximately one month ago, the company says it has received $1 billion in committed interest, representing more than 20 megawatts of capacity that it is now working to fulfill.

Energy Use and Environmental Considerations

Data centers, both in the United States and internationally, have faced increasing scrutiny over their energy consumption and the strain they place on power grids and water supplies. Runware's design choices appear aimed at addressing at least some of those concerns. The pods require only a power connection and internet access. Cooling is handled through a closed-loop liquid system with air-cooled radiators that dissipate heat without requiring a water supply - a meaningful distinction in regions where water access is constrained.

On the power side, Radulescu said the company is working with energy providers on a site pipeline covering 397 megawatts, with potential to expand to 1.2 gigawatts. The company is deliberately targeting smaller deployment sites - ones that would not typically be considered by hyperscale operators such as Amazon Web Services, Microsoft Azure, or Google Cloud. Many of these sites already have power infrastructure in place, which Runware argues reduces the likelihood of disrupting local grids or communities.

From an acoustic standpoint, the pods are described as completely silent in ambient temperatures up to 86 degrees Fahrenheit. The company also states that it uses renewable energy sources wherever they are available and is actively working to make renewables its primary energy input across its network of sites.

A Global Rollout on the Horizon

With the second factory expected to come online in early 2026, Runware's next steps involve expanding its footprint beyond Romania into the United States and other regions. Radulescu expressed confidence that modular, factory-built systems represent not just a viable alternative to traditional data center construction, but the direction in which the entire industry is heading.

"Customers should be able to plan a hyperscale data center around factory-built pods from the beginning and have it working in months instead of years," Radulescu said.

Whether Runware can translate its production targets and commercial momentum into a sustainable, large-scale manufacturing operation remains to be seen. The company is entering territory that has proven difficult for even well-funded startups, and the gap between stated ambition and operational reality in hardware manufacturing can be significant. But with a concrete factory already in use, a second under construction, and early customer commitments in place, Runware is at least further along than many of its predecessors when they made similar claims.