Jonathan Winer believes the United States is waging the wrong manufacturing war. His startup, Foundational Industries, has secured a $25 million seed round—reported exclusively by Fortune—to develop factories where artificial intelligence manages the entire operation rather than simply layering automation onto traditional assembly lines.

The funding round was led by BoxGroup and Zigg Ventures, with participation from Abstract Ventures, Adverb Ventures, Buckley Ventures, and Offline Ventures. The round reflects growing investor appetite for startups targeting the intersection of AI and physical production, an area that has attracted attention as Washington ramps up industrial policy commitments—including the CHIPS and Science Act's $52 billion semiconductor manufacturing incentives and broader efforts to onshore critical supply chains.

Rather than retrofitting existing facilities with individual robot arms and vision sensors, Foundational Industries manufactures physical products—starting with data-center hardware—in factories designed from the ground up to be operated by software. Winer, who spent six years at Alphabet's Sidewalk Infrastructure Partners deploying more than $1 billion in capital, said the seed funding will not yet finance a full-scale factory.

"What this is allowing us to do is to sort of build a minimum viable product," Winer said. "We've actually built the entire factory in software already using software emulators."

Foundational Industries' initial customers include data-center developers, neoclouds, and chipmakers that require custom rack enclosures as new AI silicon operates at different voltages and cooling requirements. The demand for such customized hardware has intensified alongside the global AI infrastructure buildout, where hyperscale operators and emerging cloud providers are racing to expand capacity and adopt next-generation accelerators from companies like Nvidia, AMD, and a growing field of custom silicon designers. The company declined to disclose customer names.

Winer's broader argument centers on China. The prevailing view in Washington attributes China's manufacturing dominance to cheap labor and lower-quality imitation. Winer considers that assessment outdated.

"Many of their factories are some of the most advanced automated factories in the world, and they are increasingly using not just international industrial automation, but also homegrown and home produced solutions," he said.

China has invested more than $1 trillion in advanced manufacturing over the past decade, supported by state subsidies and what Winer describes as "a really dense industrial ecosystem" that enables rapid product design and launch.

That density is precisely why Winer believes directly replicating China's approach will not succeed. "We just don't have the people or the skill sets to do it," he said. "And even if we did, it's probably not economically competitive to China."

Instead, Winer's strategy is to bypass that competition and construct a fundamentally different type of factory. His thesis relies on two American advantages: sophisticated AI models and researchers—a talent gap he acknowledges is shrinking—and significantly greater AI compute capacity. Combined, these factors allow Foundational Industries' system to take a customer's "product intent" and almost instantly generate a bill of materials and manufacturing process, a step that traditionally required months of manual design work.

Winer is also betting against what he sees as a structural vulnerability in China's model. He argues that Chinese factories, however advanced, remain built around older-style automation that requires constant utilization to justify state subsidies.

"They kind of need to feed the beast now," he said.

His wager is that AI-native factories—becoming cheaper and faster with each iteration—can provide the United States with a much-needed competitive edge in manufacturing.

This story was originally featured in Fortune.com.