The Re-Industrialization Economy: Why capital is coming back to the physical world

“We wanted flying cars, instead we got 140 characters.”

- Peter Thiel, co-founder, Founders Fund

For the better part of two decades, smart money largely treated the physical world as an inconvenience. Atoms were slow, capital-intensive, and low margin. Bits were fast, scalable, and, most importantly, someone else could pour the concrete. Many offshored the making of things, financialized the owning of things, and we told ourselves the exciting companies all lived in software. Thiel's line landed precisely because it stung: we'd pointed the most sophisticated capital allocation machine in human history at optimizing ad click-through rates and prompts.

This market dynamic is now changing. Capital is coming back to the physical economy - to factories, power plants, foundries, robots, and the unglamorous supply chains that make everything else possible. I call it the re-industrialization economy; but the label matters less than the shift. And after a career spent mostly on the bits side of the ledger, I've spent the last two years dragging my own attention back toward the physical, because I think this is the most important structural realignment of capital of the coming decade.

How We Got Here

This offshoring logic, specifically in the U.S., took thirty years to build and roughly thirty-six months to unravel. It didn't collapse because of one thing. It collapsed because of everything, all at once.

A pandemic that turned "just in time" into "just out of luck." A land war in Europe that reminded the West it had outsourced its arsenal. A near-total dependence on a single geopolitical rival for the semiconductors, rare earths, and pharmaceutical precursors that run modern life. And then a U.S. policy response of a scale we hadn't seen in generations — the CHIPS Act, the IRA, and a tariff regime that, whatever you think of it, priced in the political cost of building somewhere other than home.

The numbers followed the narrative, at least at first - and nowhere louder than in the U.S. Manufacturing construction spending more than doubled between 2020 and 2024, running north of $16 billion a month at the 2023 peak, per the Atlantic Council; semiconductor firms alone announced more than $645 billion across 140-plus projects since 2020; and in 2024, reshoring outpaced foreign direct investment by the widest margin on record - companies bringing their own operations home is a much more durable signal than foreigners chasing a subsidy. But the impulse was never uniquely American. Europe's Chips Act has since mobilized north of €80 billion against a €43 billion target, South Korea is building a single ~$470 billion semiconductor mega-cluster south of Seoul, and Japan and India are subsidizing fabs into existence from Kumamoto to Gujarat.

On paper, much of the industrialized world started building again at once.

So far, so bullish. But here’s where I put my other hat on.

The Part Nobody Puts on the Conference Slide

Using the U.S. as an example, if you only read the White House press releases, you'd think we were living through the largest reshoring wave in American history. If you read the construction data, you'd think something closer to the opposite.

Factory construction spending actually peaked in late 2023 and has been drifting down since - off more than 20% year-over-year through parts of 2024, dragged by a 44% slowdown in electronics and semiconductor fab spending from its mid-2024 high, per IoT Analytics. Strip out electronics and the "boom" since the tariffs is a whopping 5.6%. Business applications for new manufacturing facilities - the leading indicator for what actually gets built - fell nearly 40% year-over-year by mid-2025. Capgemini surveyed more than 1,200 executives and found that while three-quarters of organizations now have a re-industrialization strategy, actual planned investment outside the strategic core of semiconductors and defense has gone down, not up, as tariff and policy uncertainty made everyone more selective.

Announcements are cheap. Concrete is expensive. And the gap between the two is where a lot of well-intentioned capital is going to sit idle.

There's a deeper point buried in that data, though, and it's the one I want you to hold onto: the real physical build-out IS happening; it's just not happening where the nostalgic "bring back the factories" story says it should. In the U.S., the real engines of industrial growth right now are data centers, power infrastructure, and the defense industrial base - and the same concentration shows up wherever the build is genuinely underway, from Europe's energy and rearmament push to Asia's chip clusters. These are the handful of places where the demand, the capital, and national-security interests all point the same direction, and where investment has kept climbing even as broad-based reshoring stalled. That's not a footnote to the re-industrialization thesis. That IS the re-industrialization thesis, correctly understood.

What's Actually Being Rebuilt

When I map where the durable demand actually lives, it clusters into three reinforcing verticals – Physical World, Defense & Resiliency and Vertical AI. Notice how many of them are less about the finished product and more about the layer underneath it:

• Physical World

  • Semiconductors. The anchor tenant of the whole movement, and the one place the subsidies, the demand, and the national-security logic all point the same direction.

  • Robotics and automation. The answer to a labor problem that is very real — there are roughly 600,000 open manufacturing jobs and nearly 450,000 open construction jobs in the U.S. alone, and you cannot reshore what you cannot staff. Automation isn't optional here. It's the enabling condition.

  • Compute infrastructure. The largest cloud players have signaled something close to $700 billion of capex for 2026 alone. The industrialization of compute is a physical build — chips, campuses, cooling, and the power to run it all.

• Defense & Resiliency

  • The defense industrial base. The re-arming of the West, and the wider project of supply-chain resiliency that sits behind it. This was the subject of my last post; here it's the sharp end of that broader resiliency vertical, not a standalone story - which is the right way to see it.

  • Energy and power. This is the binding constraint on everything else. In 2026, access to power - not capital, not even labor - has become the universal governor on industrial growth; every available megawatt now commands a premium. That's why "nuclear everything" is suddenly in vogue, with fission startups pulling in over a billion dollars in recent rounds, and why grid, transmission, storage, and fast-deployable generation are where I'd argue the most under-appreciated opportunity sits.

  • Critical minerals and batteries. The stuff we spent three decades letting someone else refine, now treated as strategic infrastructure.

• Vertical AI

  • Industrial AI. Digital twins, AI-native factory design, AI-driven materials and process science - the software that makes a modern plant outrun a legacy one.

  • Supply-chain and logistics intelligence. AI that turns a brittle, blind supply chain into a self-orchestrating one - demand sensing, multi-tier visibility, dynamic sourcing and routing. Rebuilding supply chains is half physical and half informational, and this is the informational half.

  • Embodied AI and autonomy. The brains that turn a robot, drone, or autonomous platform from expensive hardware into an actual worker - perception, planning, and control. The hardware in the Physical World bucket is only as valuable as the intelligence riding on top of it.

The Flywheel Everyone Missed

Here's the piece that took me a while to see clearly, and it reframes the whole thing.

The lazy version of the story pits atoms against bits - the physical economy making a comeback at the expense of software and AI. That's exactly wrong. The truth is that AI has become the single largest driver of the physical rebuild, and simultaneously its most important enabler.

It’s the driver, because the entire AI buildout is a ferociously physical undertaking - the data centers, the chips, the transformers, the gigawatts. It is also the enabler, because the robotics, automation, and industrial intelligence that AI makes possible are the only plausible answer to the labor and productivity constraints that killed the last several attempts at re-industrialization. Atoms and bits aren't competing for capital. They've fused into a single trade. The physical economy may turn out to be the largest AI opportunity of them all.

“The allocators who win will be the ones who can see the whole board rather than just their square of it and invest across the capital stack.”

What It Means for Capital Allocators

So, what do you actually do with this over the next few years? I obviously have a few thoughts here.

• This is an atoms game, and atoms have different physics. Capital intensity is higher, duration is longer, and the J-curve is deeper than anything most software investors are conditioned to tolerate. The flip side is real assets, defensible moats, and pricing power that a SaaS company can only dream about. If you're underwriting this on a software clock, you will either bail at the bottom of the J or never enter at all.

• It spans the entire capital stack. This is the part I find most under-discussed and most fascinating. Re-industrialization is not a venture trade or a private-equity trade or an infrastructure trade - it's all of them at once. The same thesis expresses itself through early-stage robotics, growth-stage energy, private credit against equipment and offtake, build-to-suit industrial real estate, and public-market picks-and-shovels. Whatever chair you sit in, there's a version of this that fits your mandate. The allocators who win will be the ones who can see the whole board rather than just their square of it and invest across the capital stack. It’s the all-weather / absolute return investment thesis writ large.

• Own the picks and shovels. In a capital-intensive gold rush, the most reliable returns are in the enabling layer (the power, the compute, the automation, the components) not the flashiest end-product. Power, above all. Whoever solves megawatts wins, because everyone downstream is gated by them.

• Underwrite for policy that survives a policy cycle. A meaningful share of this rests on industrial policy - subsidies, tariffs, tax credits - that can shift with an election. I try to separate the demand that exists because of a subsidy from the demand that would exist anyway and merely gets accelerated by one. The former is a trade. The latter is a business. Only one of them survives the next administration.

• Respect the constraints, because they're the whole ballgame. Strip away the vertical-specific detail and every sector above answers to the same three questions — the ones that separate the projects that get built from the ones that stay press releases. First, where does the power come from? Compute, semiconductors, and energy are gated by megawatts before anything else; in 2026 the megawatt is the scarcest input in the economy. Second, is there demand at real scale, or just a pilot? A design win from the Pentagon, a proof-of-concept with a manufacturer, a letter of intent for offtake - those are the easy part. Converting them into contracts with true scale, and in defense real in-theater applicability, is the hard part. And third, who staffs it? You can't reshore a fab, run a plant, or field a fleet of robots into an industrial base you don’t have the people to run - a gap that bites across the advanced economies and bites hardest in the aging ones. Power, demand, and talent - those are the first questions I ask now, not the last.

The Rebuild Is Real. The Builders Are Rarer.

Let me land where I started. Re-industrialization is real, it's structural, and it's generational - the biggest reallocation of capital toward the physical world in most of our lifetimes. And it is emphatically not an American story. The same forces are redrawing Europe's push for strategic autonomy, Japan's and Korea's grip on chips and batteries, India's "China-plus-one" manufacturing surge, and the Gulf's sovereign-funded diversification away from oil - each with its own subsidies, its own bottlenecks, and its own version of the announcement-versus-build gap. Capital doesn't respect borders here, and neither should the thesis. I'm bullish on the arc, wherever it's being drawn.

But that arc will reward the ones who can actually pour the concrete, secure the power, staff the floor, and survive a policy cycle, not the ones who can merely announce, regardless of which flag they fly. The announcement wave and the build wave are two different things in every geography, and over the next few years the distance between them is where fortunes will be made and lost; from Arizona to Dresden to Gujarat.

We spent twenty years pointing our best capital at 140 characters. The chance to point it at flying cars - at the chips, grids, robots, and industrial backbone of the physical economy itself - doesn't come around often, and this time the opportunity set spans the globe.

Just don't confuse the ribbon-cutting with the factory.

#ONWARD