LE PERSPECTIVE / THE STORY

Runware wants to mass-produce AI data centers on assembly lines, like cars. Its factory-built pods are designed to arrive ready for installation instead of requiring years of custom construction.

Each Sonic Pod fits about 1,200 GPUs and one megawatt of IT capacity into a 20-foot container. Runware aims to bring a pod online within 48 hours of delivery.

The Bigger Bet Is Manufacturing
Making a working pod is just the start. The difficult part is repeatable production at scale, with dependable components and power-ready sites. Runware wants AI capacity to be a product companies can order, not a facility they must build from scratch.

Co-founders Flaviu Rădulescu, the CEO, and Ioana Hreninciuc assembled their first liquid-cooled GPU servers on Hreninciuc’s kitchen table in 2023. Now they’re aiming for factory-scale production.

Runware is targeting 30 megawatts of daily output in Romania by 2028, with a second factory expected online in early 2027. These are plans, not current production.

The 48-hour clock covers installation and setup, not manufacturing. Runware also aims to cut assembly time from about three weeks to one day.

The biggest test is scaling supply chains, power access and manufacturing without losing speed. If it works, AI compute capacity may become something companies order, rather than build from scratch.

Get stories like this in your inbox every week. Subscribe to The Week in Perspective.