Let It Crash: How to Steer What Comes After

I’m a venture capitalist, and I’m telling you to root for the crash that torches my own asset class. I mean it. The valuations are silly, the data-center spending is feverish, and half the people I talk to are quietly bracing for the fall. But here’s the argument almost no one will say out loud: the coming crash would be the best thing that happens to this technology. And it’s a necessary part of the Renaissance cycle I wrote about last time: the pattern by which a disruptive new technology forces a society to rebuild its picture of what a human being is for.

Economist Carlota Perez has found a common pattern: every great technological surge of the last 250 years took the same course: a revolution, a financial bubble, a collapse, and only then a golden age. Canals, railways, steel, cars, and computers all ran it. AI isn’t bound to repeat that course, though. What makes the pattern more than coincidence is its mechanism: the crash after the bubble.

The bubble and the crash that ends it are not a detour around the golden age. They are the road to it. But if done wrong, the same turning point that built the American suburbs built the gulag. The only variable is how we respond.

Today, that variable is a choice almost no one is naming: when the rebuild comes, do we tax the rich or spread the means by putting the machine in many hands? This essay is the case for the second.

Why we want a crash

The products of a technology bubble are more durable than the bubble itself. Technological fervor drags an absurd amount of capital into the new thing and pours it into infrastructure that no rational, sober investor would have built that fast. Britain got a railway network out of railway mania. The 1990s gave us the fiber that carried everything we built in the 2000s, laid by companies that mostly went bankrupt laying it. The investors lost their shirts; the rails and the fiber stayed.

We are living through the same process now, except the rails of today are compute, data centers, models, and the habit of millions of people learning to think alongside a machine. The chips will depreciate, sure, but the chips were never the rail; the power, the grid hookups, the data-center shells, and a generation that learned to work with machines are, and those outlive the hardware the way the fiber outlived the routers.

Then the crash does three jobs that nothing else can do.

It finishes the build. By the time the music stops, the new infrastructure is in the ground and the new way of working has become ordinary. The changeover is complete precisely because people overspent on it.

It sobers the money. A frenzy makes capital arrogant and stupid. A crash makes it humble again, and humble capital is the kind that builds real companies slowly instead of chasing paper.

And it forces the rebuild. In the boom, no one will touch the hard questions of how a technology should be governed, who it should serve, or what gets protected from it. There’s too much money in not asking. The crash creates the only moment of urgency in which a society will actually sit down and rewire its institutions. Antitrust, interoperability, and standards can do some of this, and we should use them now. But in the boom, they’re outspent by the people who profit from not asking. The crash clears the resistance. By 1935 we had deposit insurance, securities law, the whole architecture that made the postwar decades possible. None of that was conceivable in 1928.

The depression is optional

A crash and a depression are not the same thing. The crash is the moment the paper values snap back to real ones. The depression comes after, and its length is not fixed in advance.

A depression, in Perez’s telling, is just a crash whose rebuild failed or came too late. 1929 turned into the 1930s because the monetary orthodoxy held for four more years before anyone was willing to change the rules. That option belongs only to a society still whole enough to exercise it. That wholeness is state capacity: whether a government can still turn a decision into a result (Whether our state qualifies is the question I take up next).

The size of the bubble sets how deep the hole goes; how fast and how well we rebuild sets how long we stay in it. And depth cuts both ways. A downturn that’s too mild is its own failure: it never breaks the old habits, and you get what Perez calls a gilded age instead of a golden one: growth resumes, finance stays in charge, and the tensions never resolve. Half a Renaissance at best. But a depression that runs too deep and too long does the opposite damage by radicalizing people. Severity raises the ceiling and lowers the floor at the same time.

The bubble decides how hard the crash hits, left to right. We decide how well and how fast we rebuild, bottom to top. The same deep crash sits one move from Florence and one move from Weimar. And weak does not mean slow: Germany’s rebuild was fast, forceful, and aimed at the wrong portrait. A rebuild captured by the cheapest answer available is a failed rebuild, however quickly it moves.

The people praying for a soft landing—namely the Fed, the “this time is different” optimists, and the investors & founders hoping the music never stops—are all unwittingly praying for a slow death. We should not want the softest possible landing but a crash hard enough to break the current orthodoxy, met by a rebuild fast enough to keep the recession from curdling into something worse. Which means the answer has to be ready in advance.

This is the most important point, and the reason I wrote “The Fourth Pillar.” The crash does not guarantee you the golden age. It only gives you the chance at one.

The same 1929 collapse that helped produce the American postwar boom also produced Nazi Germany and entrenched Stalin’s Soviet Union. One society got the suburbs and the GI Bill; the others got the camps and the famine. What determines which side we land on is not the crash. It’s whether we’ve built the fourth pillar—a shared answer to what a human being is for—and built it correctly, before the crash arrives.

The first three pillars are showing up on their own, carried by the same economic gravity: the technology, the recovered knowledge, the concentrated capital. The fourth, the portrait of what a human being is for, has only ever been built in advance and installed in the narrow window when everyone else was finally paying attention because the easy money was gone.

The crash is coming, and it's necessary. A crash is the crudest kind of error correction—what reality resorts to when criticism didn’t get there first. A crash destroys more than savings; it destroys the answer people were living by—the career, the equity, the number going up—and demand for a replacement spikes within months. Supply moves slower. A portrait of what a human being is for takes a generation to build: the books, the schools, the people worth imitating. So the crash never chooses the future; it elects for whatever is already on the shelf.

In 1933, America's shelf held a portrait thirty years in the making—the progressives, the churches, the land-grant universities—and America got the New Deal. Germany's shelf held the völkisch man; Russia's, the New Soviet Man. Same decade, three shelves, three futures. That is what the fourth pillar determines. It is the answer when the question arrives; it’s the spec the new rules get written to, the ladder the freed talent climbs. Skip the work and the crash decides nothing on its own; the cheapest answer on the shelf wins by default.

Florence is built, not given

So what goes in those hands? The 1930s are the right place to look, because the Depression was never the crash itself. It was the years of things we failed to do after it. Recovery came only once the old orthodoxy broke and we built what should have existed already. The next crash will ask for the same kind of action.

The mechanical problem the 1930s exposed was simple: when a boom pulls all the income to the top, there is no one left to buy what the new machines can make. The reflex answer is redistribution—tax the winners, hand the proceeds back—but that will not work this time. The new wealth is mobile and the broad base owns none of it; when capital, code, and the most valuable work can move to a new state or country overnight, the money leaves faster than the law can reach it. Even a tax that landed would not rebuild what the 1930s rebuilt: a population with skin in the productive game.

So the fix does not run through the tax code. It is to spread the means, not the proceeds: put direct ownership of and cheap access to the productive layer into many hands: open models running locally instead of rented, compute and tools and equity at the source. Rented capability sends the value to the center; owned capability compounds it where people live. That, and a portrait of what the new life is for, is what the rebuild needs. Neither is built by decree, and the work falls to three sets of hands.

Government owns the guardrails and the floor. The guardrails are the rules it can still enforce—finance, antitrust, an open and competitive model layer—and its one real job is keeping exit cheap, so the layer stays open on its own. The floor is the fall it can still break: unemployment insurance, deposit insurance, retraining, the lights kept on. Catching the people who absorb the crash is not the same as taxing the people who do not, and a state too weak for the second still owes the first. What it should not do is write the portrait: the Soviets built the literacy, the recovered knowledge, and the state capital, then let the government define the human being, and got the New Soviet Man.

Companies own the breadth. Every day they choose between wide, cheap access and winner-take-all rent-seeking. Right now the most open frontier weights in the world ship out of China under an MIT license, and OpenAI (the one with “open” in the name) shipped no open-weight model between GPT-2 in 2019 and gpt-oss in 2025, moving only once cheap Chinese models made staying closed untenable. Do not expect virtue; expect arithmetic. A tollbooth is an invitation to build a road around it: the firms that widen the base capture the market, the ones that wall it off get bypassed. Keep exit cheap and breadth becomes the rational strategy on its own.

Everyone else owns the portrait and the stake, the two things no treasury will hand out. The portrait is the book, the school, the model person, the aesthetic, built and funded by many hands rather than legislated by one, not a single official answer but many in competition, kept honest by exit and criticism. The stake is ownership that outlasts the opening: getting computational equity, locally-run models and cheap compute, into many hands while the layer is still cheap, before it re-concentrates.

None of it can wait for the crash. The most expensive thing we did in the 1930s was exactly that—four years of denial before the rebuilding began, and in those same four years a continent radicalized. The crash hands us the urgency. It does not hand us the plan.

Florence ran on perhaps two thousand patrons. The Medicis built the schools, the workshops, the institutions that outlived them; today’s landlord buys back his own stock and collects rent on a walled platform that builds nothing new.

Be ready

I opened in fear of the bubble, and fear is just preparation that hasn’t been spent yet. The crash is coming whether we welcome it or not, and the only choice left is whether we meet it with a plan or with our hands open and empty.

A crash can scar instead of cleanse; the deepest ones leave lost output and a generation of lower earnings that never fully come back, and plenty of good firms die alongside the bad. The golden age isn't promised either. And even when it has arrived before, the people inside it rarely called it golden: Renaissance Florence ran on plague and war, with life expectancy that fell by some counts to eighteen. Let me be blunt about what I'm calling for: people will lose jobs, savings, companies.

But slow rot is much worse. I know I’ll get pushback for where I'm standing: I'm insulated from the worst of this, and this argument should be judged on whether it helps the people who aren't. Most people are unprepared, and the work is to make the fall survivable, not to savor it.

That work doesn't fit a quarter. It fits the calm before the crash, because the window to build slams shut the moment it arrives. The full menu is in "The Fourth Pillar": the book, a school, the open model that keeps the layer from closing, the net that catches the people who take the hit, the portrait of a life worth wanting on the far side. You don't have to carry all of it. You have to carry one piece, on purpose, while the money is still easy and the room is still calm. The bubble decides when the question gets asked; we decide what's already built when it does.

Let it crash. Just don't be empty-handed when it does.

Vijay Pande is the co-founder of VZVC, a venture firm investing at the frontier of AI and the life sciences. He previously spent a decade as a general partner at Andreessen Horowitz, where he founded and led its $3B+ Bio + Health fund, and two decades as a professor at Stanford in chemistry, structural biology, and computer science. He created Folding@home, one of the first distributed-computing projects to simulate protein folding at scale, and has authored more than 300 scientific papers. He writes about how technology, markets, and culture move in cycles, and what we can build at their turning points.

Behind this: Carlota Perez, Technological Revolutions and Financial Capital. Davidson and Rees-Mogg, The Sovereign Individual. And my work “The Fourth Pillar.”

Acknowledgements: Thanks to Greg Bowman, Joel Dudley, and Lara Pande for their critical reading of this manuscript.

Image credit: Thomas Cole's "The Course of Empire: Destruction”