Somewhere in the Nevada desert, a fiber optic cable runs along a right-of-way that was deeded to a railroad company that ceased to exist before anyone alive was born. The railroad is gone. The corridor isn't. This is how it always works: the crash wipes out the investors and leaves everything else exactly where it was, waiting for whoever shows up next with cash and patience.
1. The Railroads
The great American railroad boom of the 1860s and 70s was built on land grants — the federal government handing over territory to companies that promised to lay track. They laid it. They also borrowed against it, overcapitalized it, and collapsed spectacularly in the Panic of 1873. Dozens of railroad companies failed in the first years alone; by the end of the decade the crash had swept through a significant fraction of the industry.
The physical infrastructure didn't go anywhere. Successor companies, often the bondholders themselves, acquired the routes at distressed prices and kept running them. The right-of-way — the legal corridor the track sat in — proved more durable than any of the companies that had claimed it. When telegraph lines needed to go coast to coast, they followed the railroads. When telephone trunk lines came, they followed the telegraph. When fiber came, it followed the telephone. The bankruptcy was an ownership event. The corridor was permanent.
2. The Broadcast Freeze
In 1948, the FCC froze all new television station licenses while it sorted out the technical standards for the new medium. The freeze was supposed to last a few months. It lasted four years. When it lifted in 1952, the Commission issued a plan that allocated channels to markets and markets to operators with the quiet finality of a land grant.
The companies that held licenses when the freeze lifted held them for decades. The carriage relationships — who could reach which households — became the underlying asset of American media. Cable didn't displace this structure; it parasitized it, using the same household relationships and negotiating against the same license holders. Streaming inherited the content libraries of companies whose core asset was spectrum allocation decided in 1952. The technology changed every twenty years. The power map changed very little.
3. The Dark Fiber
Between 1996 and 2001, American telecom companies laid more than 80 million miles of fiber optic cable. Most of it was never lit. The business model — that internet traffic would grow fast enough to fill it — was correct. The timeline was wrong by about five years, which was long enough to bankrupt most of the companies that laid it. Global Crossing. WorldCom. 360networks. The fiber stayed in the ground.
Google began acquiring dark fiber rights in 2005. The purchases were quiet enough that it took years for the scale to register. What Google understood was that the cable itself was only part of what it was acquiring — the other part was the right-of-way, the conduit permits, the relationships with municipalities that had already been negotiated by the companies that no longer existed. When YouTube arrived in 2006 and Google bought it for $1.65 billion, the infrastructure to actually serve video at scale was already in place. The crash had been a transfer, not a destruction.
4. The Datacenter Real Estate
The dot-com boom needed somewhere to put the servers. Between 1998 and 2001, purpose-built datacenter facilities went up across the country — raised floors, redundant power, fiber access points, the specific requirements of commercial computing infrastructure. The companies that built them, and many of the companies that leased them, went under in the bust.
The buildings remained. Amazon began building AWS on a foundation of cheap colocation space and existing facilities in the mid-2000s — infrastructure priced for a market that no longer existed. Microsoft and Google followed. The hyperscaler model — centralized compute leased on demand — was made economically viable in part because the physical substrate had already been paid for by investors who lost everything. The companies that defined the next decade of computing did not build from scratch. They bought the aftermath.
5. The AI Datacenters
The current buildout is the largest technology infrastructure investment in history. In the first half of 2025, AI-related capital expenditure was the dominant driver of U.S. business investment growth, outpacing every other sector combined. Microsoft, Google, Amazon, and Meta have collectively committed hundreds of billions to datacenters, chips, and power infrastructure. The question analysts are asking is whether the demand will materialize fast enough to justify it.
It may not matter. The datacenters will exist regardless of which AI companies survive. The power contracts, the land, the fiber connections, the cooling systems — these are physical assets with long lives and alternative uses. The companies best positioned for a correction are the ones that already operate the infrastructure: the cloud providers who run the models for the startups, who will still be running infrastructure when the startups are gone. And then there is Apple, sitting on over $140 billion in cash and marketable securities as of early 2026, building its own silicon quietly in Houston, with a datacenter expansion underway across six states and an AI server chip entering mass production in 2026. Apple has done this before — arrived late, acquired selectively, and ended up owning the category. The railroad barons who bought the bankrupt lines in 1874 did not build the railroads. They just showed up when the receivers needed to sell.
The crash doesn't destroy the infrastructure. It clarifies who owns it. The fiber in the Nevada desert didn't know it was laid for a company that would cease to exist. It didn't know Google would use it to serve video at planetary scale. It doesn't know what it's carrying now, or who will own the right-of-way when the current boom finishes its correction.