Startup Runware Wants to Mass-Produce AI Data Centers Like Cars: The 48-Hour Goal and the Risks

What happened

According to TechCrunch, AI infrastructure company Runware told the outlet it wants to build a factory that assembles modular data center pods on a production line, much as cars are made. A modular data center is one built in a factory and installed once shipped, rather than constructed entirely from scratch on site. The Sonic Inference Pod, which Runware unveiled a few months ago, is this kind of product.

TechCrunch notes that Crusoe and Schneider Electric also sell prefabricated pods, but Runware says it can ship faster. Nvidia also announced Vera Rubin AI factories and DSX POD earlier this year, likewise aimed at helping companies build AI data centers faster. TechCrunch says the difference is that Runware says it has actually built a factory, rather than just designing the infrastructure that would let others build one.

The company's goals and business model

Runware co-founder and CEO Flaviu Radulescu told TechCrunch that the company plans to scale its factory in Romania to 30 megawatts of data center capacity per day by 2028, close to 1 gigawatt per month. Megawatts and gigawatts are units of power, with 1 gigawatt equal to 1,000 megawatts, used here to express data center scale. TechCrunch notes that at this pace, Runware's annual output would roughly equal the data center power the US was expected to add this year, about 10 gigawatts.

Radulescu said the company will ramp up gradually starting next year. It already has two factories: one it has used for a few years to build the pods, and a second it started building earlier this year, which is expected to be finished within months and go online early next year.

Why it matters and what questions remain

If this factory mass-production model works, AI computing power could be set up faster. Radulescu said customers should be able to plan hyperscale data centers (the giant facilities run by the largest tech companies) around factory-built pods from the start and have them running in months rather than years. But that is his claim, and whether the plan comes true depends on execution.

TechCrunch also notes that data centers in the US and elsewhere have drawn criticism for their energy use and their impact on power grids and other resources. Radulescu said the pods need only power and an internet connection and run completely silently at temperatures up to 86 degrees Fahrenheit. He said the company is working with power providers, and its planned site list covers 397 megawatts, potentially expanding to 1.2 gigawatts. He added that the sites typically already have power and are much smaller than those of hyperscalers such as Amazon, Microsoft and Google, so they should not further strain the grid or disrupt communities. He also said the company uses renewable energy whenever possible and aims to make it the main source.

Funding and supply chain challenges

Runware has raised more than $60 million so far and plans a Series B round; for now it is using the $50 million Series A raised last December to finish and launch the factory. Series A and B are different stages of a startup's fundraising. Radulescu said designing its own systems and buying widely available, affordable parts helps keep costs low, and that the company has brought in a highly experienced plant manager from the car industry to work alongside its chief of hardware.

On the supply chain, Runware says it designs and builds its cooling and other systems itself from standard parts, but some components such as pumps and fans come from Germany, and it still depends on Nvidia for networking. The company says that once the new factory is online, it will expand to the US and other parts of the world.

Browse the latest news in this topic

• AI news