Runware Reimagines AI Infrastructure with Assembly-Line Data Center Production
Runware is attempting to industrialize data center construction by manufacturing modular AI compute units on assembly lines, aiming to reduce deployment times to 48 hours.
The physical infrastructure required to support AI workloads is undergoing a shift as startups attempt to move data center construction from bespoke civil engineering projects to factory-based manufacturing. Runware, an AI infrastructure company, is scaling production of its Sonic Inference Pods—modular data centers designed to be built on an assembly line. According to TechCrunch, the company currently operates one factory and is finishing a second, which is expected to go online early next year. By shifting integration work into a controlled factory environment, Runware aims to deliver units that can be installed and brought online in 48 hours.
Runware’s manufacturing approach is designed to produce standardized, high-density compute modules that bypass the lengthy timelines of traditional hyperscale construction. CEO Flaviu Radulescu stated that the company plans to scale production at its Romania-based facility to 30 megawatts of data center capacity per day by 2028. While the company utilizes standard components for cooling and power systems, it maintains an in-house hardware team to design and test components, allowing the infrastructure to adapt to new GPU generations as they are released. Radulescu noted that the company’s pod delivery times are shorter than current GPU lead times.
The company faces the challenge of scaling complex manufacturing operations, a sector where other startups have historically struggled. To mitigate execution risks, Runware has hired a plant manager from the automotive industry to oversee its production lines. Radulescu emphasized that the company’s cost structure benefits from utilizing widely available parts and cost-effective factory labor. Runware has raised $50 million in Series A funding, which it is currently using to launch its second factory. The company also offers customers the option to rent compute capacity or supply their own GPUs, paying a monthly fee for the pods and operations.
Regarding traction, Runware claims to have received $1 billion in committed interest for its Sonic pods since offering them to outside customers a month ago, a figure representing over 20 megawatts of capacity. The company is currently working to fulfill these initial orders. The pods utilize a closed-loop liquid cooling system with air-cooled radiators, which the company says allows them to operate without a water connection. Runware is also coordinating with power providers on a site pipeline that covers 397 megawatts, with potential to expand to 1.2 gigawatts, according to Radulescu.
As the industry grapples with the speed of silicon iteration, Runware is positioning its modular systems as a solution for operators who need to add capacity without committing to multi-year construction cycles. The company intends to expand its factory footprint to the U.S. and other regions following the launch of its second facility. While competitors like Crusoe and Schneider Electric also offer prefabricated pods, Runware differentiates itself by emphasizing that it has established its own factory infrastructure rather than solely designing the systems for others to build. The company’s ability to maintain quality control during rapid scaling remains a key metric for industry observers.
Sources
- 01 Can AI data centers be mass-produced like cars? This startup aims to find out — TechCrunch — Startups