Runware aims to install modular AI data-centre pods in 48 hours and scale factory output by 2028. Both targets depend on execution and site readiness.
Runware says it wants to mass-produce modular AI data-centre pods at a Romanian factory and bring installed units online within 48 hours. The AI infrastructure startup outlined its plans to TechCrunch on October 30. Its timeline refers to installation and setup at a prepared site, not manufacturing the pod itself. The company’s much larger production targets remain future goals rather than demonstrated capacity.
Runware’s modular AI data-centre pods
Runware’s Sonic Pods are container-sized systems designed for AI inference, the computing that runs trained AI models. The company says each pod packages about one megawatt of computing capacity into a 20-foot container. It says the design includes roughly 1,200 GPUs and closed-loop liquid cooling.
Runware introduced the pods earlier this year as a way to deploy compute without building a conventional data-centre campus. Its 
official product page says a unit needs ground, power and network access, and can be deployed in days. The page also describes an average build time of three weeks, a distinction from the 48-hour onsite setup target CEO Flaviu Fadulescu described to TechCrunch.
Forbes previously reported on the pod design, including its container form and one-megawatt capacity. The systems arrive assembled and tested, according to Runware’s descriptions cited in that coverage. That reporting supports the existence and design of the product, but not the future factory’s proposed production rate.
Runware’s 48-hour installation target
Fadulescu told TechCrunch the company aims to bring pods online within 48 hours after installation begins. The target assumes a customer site is ready and has the required power and network connection. It does not mean a fully functioning data centre can be manufactured from scratch in two days.
Runware says building a pod currently takes around three weeks, and it expects the process to fall to one day as production expands. Those figures are company estimates, not independently verified manufacturing results at the planned scale. The startup says customers can provide sites themselves or ask Runware to find locations.
The company offers more than one way to use the infrastructure. Customers may rent computing capacity from Runware, or provide their own GPUs and pay for pod operations. Power and internet costs are separate, according to TechCrunch’s account of the company’s model.
Production plans and funding
Runware told TechCrunch it aims to raise factory production to 30 megawatts of data-centre capacity per day by 2028. The company said it has two factories: an existing facility and a second site under construction. It expects the newer factory to finish within months and begin operating early next year.
Those targets require a sharp scale-up in manufacturing, supply chains, staffing and site access. Runware has said it will raise a Series B, but TechCrunch reported the company is completing its factory work with a $50 million Series A announced in December 2025. The company’s funding announcement says Dawn Capital led that round, with several other investors participating.
Runware also told TechCrunch it had received $1 billion in “committed interest” for its pods, representing more than 20 megawatts it is working to deliver. That description should not be treated as completed sales or installed capacity. The company did not disclose the customers or contract terms in the report.
Power, cooling and execution challenges
Runware says the pods use a sealed cooling loop and need no water connection. Its design uses radiators to release heat, while sites still require substantial electrical power and reliable internet. The company told TechCrunch it is working with power providers on a site pipeline, but those sites are not all necessarily live or secured.
The CEO said the factory model could make AI infrastructure quicker to deploy than traditional construction. “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.” — Flaviu Fadulescu, Runware co-founder and CEO. That is the company’s vision, not a verified result of the planned production system.
Ultimately, Runware is pursuing factory-built pods to speed AI compute deployment and scale manufacturing. The product exists, but the 48-hour setup target and 2028 output goal have not been proven at scale. Factory completion, power availability and customer deliveries will be key milestones. Runware’s next test is turning its ambitious projections into operating capacity
