Has AI Shrugged?

1873: A Financial Glitch, Not a Flawed Idea

The Panic of 1873 made it into textbooks as the collapse of a railroad bubble. In reality, it was a breakdown in the financing mechanism, not in the transportation concept itself. The failure of investment bank Jay Cooke & Company, overleveraged with Northern Pacific bonds, coincided with a shift in monetary regime— Coinage Act 1873, which removed silver from circulation, draining liquidity and triggering a cascade of debt defaults.

Yet looking at FRED statistics on total U.S. railroad mileage, the picture is clear: within just a few years after the panic, construction returned to its previous pace, and by 1890 the network reached approximately 200 thousand miles. Cross-referencing with data from John Stover's American Railroads shows growth from ~93 thousand miles in 1880 to ~163 thousand in 1890—meaning the financial "timeout" didn't alter the network's trajectory.

Data Centers as the New Rails

Today we're laying lines again – only not across prairies, but through the clouds. Server farms, power lines, undersea cables – this is the new track. According to Dell'Oro Group, global data center investment surged 51% in 2024, reaching $455 billion. And according to NVIDIA Investor Relations, data center revenues exceeded $115 billion – a figure that looks more like a mid-sized country's GDP.

But can we unequivocally call this a sign of maturity? Or are we witnessing a 21st-century "railway mania" phase, where infrastructure is being built faster than the economy that would justify it can emerge? So far, there are no answers. We're at a stage where even excess capacity doesn't look like a mistake, because it creates opportunity for something not yet invented.

"The Fifth Revolution": Human or Machine?

The World Economic Forum speaks of a fourth industrial revolution – an era of digital and biological convergence. But there's growing discussion of the next one – Industry 5.0: humans returned to the center of processes.

And yet – who's at the center? The human explaining to an algorithm what fairness means? Or the system itself, gradually learning to explain us? Perhaps we're already in the fifth revolution but don't realize it, just as railway builders didn't understand they were creating the "internet of the 19th century." Or perhaps, conversely, we're overestimating the depth of change, confusing computational speed with progress.

When Atlas Doesn't Shrug, But Builds

In "Atlas Shrugged" there's a scene where Dagny Taggart, exhausted by stupidity and regulations, flies over a ruined America and first glimpses the valley – a utopia where people who refused to "carry the world on their shoulders" built their own system. John Galt tells her:

"We are the motor and the mind of the world, Dagny. Stop us, and the world stops."

– Ayn Rand, Atlas Shrugged, Part III, Chapter I ("Atlantis")

This phrase sounds different today. Our "motors" aren't people but computing systems, and they operate not in defiance of reason but at its outer limits. The 21st century offers an inverted allegory: Atlas didn't retreat to the mountains—he stayed in megawatt data centers. He's not fleeing the world; he's servicing it.

And yet the question remains open: who is stopping the motor today? Governments regulating models? Algorithms autonomously optimizing themselves? Or perhaps man himself, having delegated thinking to what he once created as a tool?

The irony is that if Rand's heroes lived in our time, they wouldn't hide from the system—they'd write code for it, building the very machine that makes the world predictable. Their Atlantis valley isn't a hidden utopia but a server room flickering with green indicator lights.

Lessons from the Past and the Shadow of the Future

Financial cycles don't equal technological ones. In 1873, banks collapsed, but the rails remained. Today a valuation crisis may slightly damage the AI narrative, but it won't stop it.

Infrastructure is a public good. Back then, the state regulated metal and credit; today it's energy and data. And once again, the pace of progress depends on policy.

The Geography of the Future. Wherever AI hubs are being built, new industrial centers are being born.

And yet...

The railroads weren't a bubble—but they transformed society far more profoundly than investors ever imagined. AI infrastructure may well be following the same path. Only now the question isn't "when will all this pay off," but rather: who owns the future built by machines.

And if this era truly has its own Atlases—who are they? The people holding the world on their shoulders, or the algorithms upon which humanity now rests?