The Manufacturing Transition:
A Generational Bet

The United States is making a generational bet on reindustrialization.

We cannot afford to lose. Over the next decade, our economic and national security depend upon our capacity to produce critical goods, from drones to medical devices. Our industrial base has been hollowed out, and the battlefield of the future may depend on production we no longer have. The pandemic showed we couldn't make what we needed; now a looming conflict with China reminds us how dependent we've become.

The U.S. is making two simultaneous bets: one on reindustrialization, one on AI and compute. On the surface, they seem separate. They aren't. AI is the mechanism that enables reindustrialization. The bet will be decided by whether we can leverage our advantage in AI models to surpass our rivals’ industrial capacity before it's too late.

America makes bets in a messy and decentralized way. In manufacturing, that has left us weakened and exposed. China, by contrast, has been ruthlessly focused for decades. It’s poured more than $1 trillion into advanced manufacturing, deploys over 300,000 industrial robots a year — more than the rest of the world combined — and now makes roughly a third of everything on Earth, with that share heading towards 45% by 2030. Its supplier clusters are densely networked, and its state capital functions as a strategic national venture fund.

The biggest cliché in U.S. venture capital right now is that, as "software margins asymptote to zero," "physical AI" will be the only real moat left. The cliché is half right. Reindustrialization will be won by whoever turns AI compute into physical manufacturing capacity.

The same messy, decentralized instinct that left us vulnerable in manufacturing built our AI advantage. A mobilization is already underway in compute, and it holds the key to our reindustrialization. The estimated $7 trillion pouring into AI data centers is the largest national industrial project since the Interstate Highway System, predicated on the idea that, at sufficient scale, compute becomes the chokepoint through which the global economy runs. By some measures, it's paying off: today, the U.S. holds 75% of the world's AI compute capacity, proof that America can still mobilize at industrial scale.

Walk the streets of El Segundo, and you can see the next industrial base taking shape. Dozens of companies are building “in atoms, not bits”: robotics, drones, satellites, defense systems, energy systems, factories, and the physical infrastructure of the AI age.

This is how America leapfrogs.

Not by recreating the physical world China already dominates. Not by trying to win the last war on labor cost, scale, and legacy supply chains. But by reinventing the means of production itself, with AI at the center.

The pieces are now in place: Robot Foundation Models that can generalize physical tasks, inference at the robotic edge that can bring intelligence onto the factory floor, and supply chain ecosystems that can be coordinated through software.

But the window is narrow. We are in an arms race toward autonomous production. The advantage will accrue to the companies and countries that first convert AI from a digital capability into physical output at industrial scale. Whoever builds generalizable autonomous factories first will not just make products cheaper. They will define the next geography of manufacturing, the next structure of supply chains, and the next basis of industrial power.

Production follows constraints.

For the last half century, the binding constraint was labor. Manufacturing moved to wherever labor was cheapest, most abundant, and most scalable. When labor is no longer the constraint, production no longer needs to chase low-cost labor markets around the world. It can return to the places with the deepest demand, the most available capital, the most secure inputs, and the strongest strategic need to build.

We are at an inflection point.

The autonomous factory is not just a cheaper factory. It is a different kind of industrial system. Every unit produced generates data. Every process run improves the model. Every factory deployed expands the operating playbook. Every customer served deepens the ecosystem. The first autonomous factory makes the second easier. The second makes the third faster. The network compounds.

Labor arbitrage was the organizing principle of the last era of manufacturing. Autonomy will be the organizing principle of the next one.

This is the manufacturing transition; it is the defining arms race of the next decade.

Foundational Industries

The manufacturing transition will not be won by retrofitting the old industrial base.

It will not come from adding a robot arm to a legacy line, a vision model to an inspection station, a humanoid to a warehouse aisle, or an AI scheduler on top of yesterday’s factory. Those tools will improve productivity. They will not create autonomous production.

Autonomous manufacturing requires a different starting point.

The factory has to be designed for AI from the ground up. Every process, every work cell, every material movement, every inspection step, every scheduling decision, every quality loop has to be built around a common intelligence layer. The goal is not a factory assisted by AI. The goal is an agentic factory orchestrated by AI.

That is Foundational’s mandate.

We are building a network of AI-native factories to reshore critical supply chains. Starting from a blank page, we are designing the production system for the next industrial era: design ingestion, process planning, robotic execution, inspection, scheduling, quality, and supply chain operations, all connected by software and improved by data. In this model, AI is not a tool inside the factory. AI is the architecture of the factory.

Winning does not mean matching China factory for factory or worker for worker. That race was decided a long time ago.

The opportunity is to build on a different curve: factories that become cheaper, faster, and more capable with each one deployed. Not because we add more labor, but because the system learns. The model improves. The playbook gets sharper. The next factory benefits from everything the last one discovered.

That is where America can compound its advantage: compute, data, software, capital, and iteration. And it creates a new industrial workforce in the process: robotics technicians, AI system operators, manufacturing engineers, and people who know how to run factories that improve over time.

It also gives us something we no longer have today: control over supply chains. The next pandemic, blockade, or war should not leave us waiting on someone else’s factory to decide whether America can build.

Foundational was assembled for this moment.

Our team brings together the two capabilities this transition requires: the ability to build intelligent manufacturing systems and the ability to run manufacturing at global scale. Few organizations possess both. Foundational was purpose-built by combining leaders from the frontier of physical AI with leaders who spent decades operating one of the world's largest global manufacturing networks. We believe autonomous factories will be built at the intersection of those two disciplines, and that's exactly how we assembled our team.

I have spent 25 years building technology-enabled infrastructure, deploying more than $3 billion of capital, and creating first-of-its-kind physical AI companies at Alphabet’s Sidewalk Infrastructure Partners. The path to Foundational began with a simple realization: even when America has the capital, the demand, and the strategic need, too often we no longer have the production capacity. In one of our digital infrastructure projects, we needed components that no domestic supplier could build. The problem was not financing. The problem was the factory.

So we decided to build it.

We are pleased to announce a $25M seed investment led by BoxGroup and Zigg Ventures, with Abstract Ventures, Adverb Ventures, Buckley Ventures, and Offline Ventures. They share our conviction that the next industrial base will be built by manufacturers that own the factory and the operating system and are accountable for delivering qualified production at scale.

We are launching our first facility at the end of this year. It will be the first node in a network of factories designed from the ground up to optimize AI. Our initial focus is manufacturing capacity for AI data-center equipment, where new silicon is driving new demand for power, cooling, interconnect, and physical infrastructure.

The United States built the foundational models. Now we have to build the physical world those models enable.

The next industrial base will be autonomous.

And it has to be ours.

Jonathan Winer

Founder & CEO, Foundational