Forget robots on assembly lines. Foundational Industries wants AI to run the entire factory

Jonathan Winer has a theory about why the U.S. keeps losing the manufacturing argument, and it's not the one you hear in Washington. His startup, Foundational Industries, just closed a $25 million seed round led by BoxGroup and Zigg Ventures, with Abstract Ventures, Adverb Ventures, Buckley Ventures, and Offline Ventures also writing checks. The pitch isn't another robotic arm bolted onto an existing assembly line. It's a factory designed from the ground up to be operated by software, starting with something unglamorous but urgent: custom rack enclosures for data centers.

Winer spent six years at Sidewalk Infrastructure Partners, Alphabet's infrastructure arm, deploying more than $1 billion. He's using that seed money not to build a giant plant but a minimum viable one — the whole thing has already been simulated in software emulators, he says, before a single piece of metal gets touched. The first customers are data-center developers, neoclouds, and chipmakers who need enclosures built fast because new AI silicon runs hotter and at different voltages than anything that came before. Names weren't disclosed, but the timing tracks with an industry scrambling to keep up with chip generations that change faster than factories can retool.

The real argument, though, is about China. Winer thinks the standard line — that China wins on cheap labor and knockoff quality — is years out of date. He describes Chinese factories running some of the most advanced automation on the planet, increasingly built on homegrown tools rather than imported ones, backed by more than $1 trillion in state investment over the past decade. Copying that playbook, he argues, is a losing move for the U.S.: not enough skilled labor, and even with it, no way to match Chinese economics on their terms.

So instead of competing on China's turf, Foundational is trying to change the game entirely. Winer's bet rests on two things America still has plenty of: strong AI models and, for now, more compute. Feed a customer's product intent into the system and it can spit out a bill of materials and manufacturing process almost instantly — a task that used to eat months of engineering time. He also thinks China's advantage has a hidden cost. Those state-subsidized automated factories need constant utilization to justify their existence, he says, which locks them into a kind of forced busyness. An AI-native factory, cheaper and faster to build with each new one, doesn't carry that baggage.