on reindustrializtion

A few days ago, I found myself standing on the floor of a factory in Ferndale, Michigan, about fifteen minutes north of Detroit. The company is called Nox Metals. We invested at the pre-seed and seed stages.

Nox is trying to modernize a surprisingly antiquated part of American manufacturing: buying, cutting, and delivering metal. A customer can enter the dimensions of the aluminum they need, receive a quote almost instantly, and have the order routed directly into an (increasingly) automated production system. Nox describes what it’s building as America's first automated aluminum plate-cutting operation, serving customers across aerospace, defense, space, medical, and industrial manufacturing.

I knew all of this before I arrived, but seeing it for myself was different.

There is something about manufacturing that doesn’t translate well into a website or a deck. You read about throughput and software and lead times and gross margins, but none of that prepares you for the visual manifestation of the thing itself: enormous pieces of aluminum sitting on the floor, industrial-grade saws cutting through metal, forklifts moving heaps of material around the facility. Employees measuring, inspecting, moving, cutting, and shipping raw material that will eventually become pieces of airplanes, defense systems, medical equipment, and machines somewhere else.

I found the whole thing unexpectedly awe-inspiring. Some of that feeling had to do with Nox itself. But a lot of it had to with where this was all happening: Detroit.

People often forget that there was a period when Detroit was one of the most important cities in America. In 1920, it was the fourth-largest city in the country, behind only New York City, Chicago, and Philadelphia. By 1950, nearly 1.85 million people lived there. General Motors was the largest automobile company in the world, Ford had transformed manufacturing through mass production, and the city became synonymous with an industry that was reshaping American life itself: the auto industry.

Going through Detroit today, it is admittedly difficult to imagine just how economically consequential Detroit once was.

For a certain kind of American worker, the factory wasn’t just where you worked. Working in manufacturing created an entire social contract. You could build something tangible, earn a respectable living without an advanced degree, buy a home, raise a family, and believe with some degree of confidence that your children might live better than you did. Manufacturing jobs were, in many ways, a ticket to a middle-class life at a minimum.

But then, slowly, that world began coming apart.

The oil shocks of the 1970s altered what Americans wanted to drive. Japanese automakers arrived with smaller, more fuel-efficient cars, and often with manufacturing systems that were just… better. Foreign competition intensified through the 1970s and 1980s, while the Big Three American auto manufacturers struggled with cost structures, bureaucracy, quality problems, and a market that was changing fast. By the 1990s, employment at the Big Three had fallen by hundreds of thousands of workers from the 1980 levels.

What else happened? Globalization. For decades, American business increasingly organized itself around a very specific idea of efficiency. If something could be manufactured more cheaply somewhere else, move it there. If a supplier on another continent could shave several points off the cost of a component, use that supplier. The entire American business apparatus congregated around this concept. Consultants, executives, investors, economists, and business schools built sophisticated frameworks around optimizing global supply chains. McKinsey itself was publishing research in the early 2000s describing offshoring as a “win-win” and emphasizing the savings available to companies through lower-cost labor abroad.

In many cases, it worked extraordinarily well: consumers received cheaper goods, companies expanded margins, supply chains became global, and capital was allocated toward wherever production was most efficient.

Globalization, of course, had a dark side. From 2000 to 2010 alone, the United States lost roughly 5.8 million manufacturing jobs. McKinsey's own research later estimated that trade and offshoring directly accounted for roughly 1.3 million of those losses, while productivity growth was an even larger contributor. Those numbers were not evenly distributed across the country. Cities built around manufacturing absorbed more of the pain. Detroit absorbed a tremendous amount of it.

Then came 2008. Detroit had the misfortune of being exposed to several collapsing systems at once. The housing market was imploding. Credit disappeared. Consumers stopped buying cars, one of the easiest large purchases to postpone when you are terrified about your job and your house. GM and Chrysler were already carrying high structural costs and substantial debt, and both eventually required federal assistance as they approached collapse.

Between 2000 and 2009, metropolitan Detroit lost roughly 463,000 jobs, more than one-fifth (!) of the region's employment. Its economy contracted by nearly 18 percent during a decade in which the broader American economy actually grew. I find those numbers hard to comprehend.

Where did this all lead? The inevitable, of course. Detroit ultimately filed for bankruptcy in 2013, in what remains the largest municipal bankruptcy in American history.

But, it didn’t stop there. As if the region needed another reminder of how exposed it remained to physical industry, COVID arrived. In May 2020, employment in the Detroit-Warren-Dearborn metropolitan area was down nearly 20 percent from a year earlier, among the steepest declines of any large metropolitan area in the country. Manufacturing employment in the region again was hit particularly hard.

There is no single villain responsible for what happened to Detroit, even if our brains prefer stories simple enough to have one. It was foreign competition. Automation. Globalization. Management decisions. Public policy. Political dysfunction. Demographics. The financial crisis. A pandemic.

But standing inside Nox, I kept thinking about how many times this particular part of the country had been told, in one form or another, that the future was happening somewhere else. And then I was standing in a factory full of young, hungry people trying to build it there where it felt like it truly belonged.

Factories have traditionally been organized around fairly rigid boundaries. Engineers design things -> software teams build systems -> operators use those systems -> management collects information from the floor -> sends it upward -> meetings happen -> priorities get established, and perhaps eventually some change makes its way back down.

It was striking. I remember watching floor workers walk over to engineers and talk about the software they were using. A person actually cutting and moving metal could encounter something inefficient, confusing, or simply annoying, explain it to someone sitting a few feet away, and potentially have the system changed. Right then and there. The feedback loop was absurdly short.

I’ve spent much of my career around software businesses, where we talk constantly about being close to the customer: talk to users, watch how they behave, collect feedback, ship quickly, and iterate. The reality, however, is that inside most software companies, the people writing the code are often very separated from the people using it. At Nox, sometimes the user is standing twenty feet away.

I started to realize that there was something more interesting happening there than a manufacturing company using software. The manufacturing operation itself was informing the software on how to become better.

The floor workers understand things the engineers cannot fully understand because they spend every day physically interacting with the system. They know which sequence of actions feels unnatural; they know which piece of information arrives too late, and which screen requires three clicks when it should require one.

But at Nox, the reverse is also true. The engineers sit close to the operation, so they cannot remain abstracted from the physical consequences of what they build. They see the saws and machinery. They see the inventory sprawled across the factory floor. They see what happens when scheduling logic is wrong, for example.

For decades, we as an industry treated software and manufacturing almost as opposite ends of the economy.Software was clean, scalable, capital-light, fast-moving, and located in places like Silicon Valley. Manufacturing was physical, capital-intensive, messy, slow-moving, and increasingly something America was supposedly better off letting someone else do.

That distinction is beginning to break down. The interesting manufacturing companies being built now may not simply be factories with software layered on top. The software may become inseparable from the factory itself. Each and every physical action creates information. Each worker becomes a real-time source of product feedback. The physical operation and the digital system begin teaching each other.

Nox calls its internal system Factos. It uses software and machine learning to make critical decisions around cutting, scheduling, inventory, and how material should be used. The company's stated goal is not simply to automate individual tasks, but to coordinate the factory more intelligently as a system.

Much of the intelligence, however, doesn’t come from an algorithm. It comes from the person standing beside the saw, or a forklift driver.

We spent decades thinking about automation primarily as a way to remove people from manufacturing. Perhaps some of the most interesting factories of the future will work because software allows the knowledge of the people inside them to travel much faster.

A floor worker notices something -> An engineer understands it -> The software changes -> The operation improves -> the next worker encounters a slightly better system.

Eventually, it’s not hard to imagine thousands of tiny observations accumulating inside the company. That feels like a very different model of manufacturing to me.

It also made the culture at Nox easier to understand. People seemed genuinely happy to be there. I realize how subjective that observation is, and perhaps I happened to visit on a particularly good day, but the energy inside the building was hard to miss. There was none of the strange social distance I expected between the people building software and the people operating machines. Everyone seemed to understand that they were working on the same problem.

In my Uber to the airport, I left thinking about Detroit again.

There is a tendency when we talk about bringing manufacturing back to America, almost to make the conversation nostalgic. We imagine recreating some earlier industrial age, reopening the factories that closed, and somehow returning to some version of the country that existed before globalization hollowed parts of it out.

I’ll be candid: I do not think that is what is going to happen. The old factory is probably not coming back. Something better is coming. A factory where software engineers understand the shop floor and machine operators understand the software. Where automation doesn’t simply replace human judgment, but captures it and iterates.

There was something poignant to me about seeing that being built outside Detroit. A century ago, this region helped invent modern industrial production. Then it spent decades experiencing many of the consequences of that system moving elsewhere. Now, inside an old industrial building in Ferndale, a group of people is trying to figure out what the next version might look like.

I don't know how large Nox ultimately becomes. One of the strange privileges of early-stage investing is that you encounter companies before the answer exists.

What I do know is that I walked into the factory thinking mostly like an investor. I was looking at equipment, people, throughput, software, inventory, and trying to understand how all the pieces fit together. I left thinking about something much larger. For years, Americans have talked about the decline of manufacturing as though it were a force of nature, something that happened to us because history inevitably moves in one direction. Maybe history is not quite that tidy.

Industries leave, sometimes cities decline, and inevitably, sometimes knowledge disappears. And sometimes, years later, a twenty-something founder rents a factory fifteen minutes outside Detroit, fills it with metal and software and people who actually seem excited to come to work, and decides to start building again.

Long America.