Noah Smith cuts through the noise of AI hype to expose a financial trap that most observers are blind to: the technology might succeed spectacularly while the companies building it go bankrupt. While the market fixates on whether AI is a bubble, Smith argues the real danger is a timing mismatch where massive debt comes due before the economic value of the technology can materialize.
The Illusion of Failure
Smith begins by dismantling the most common fear: that AI will turn out to be useless. He contrasts the current landscape with the Virtual Reality boom, where Meta spent billions on a technology that never found a mass market. "Meta spent $77 billion on developing the virtual reality 'Metaverse', but outside of gaming and some niche entertainment applications, nobody really wanted VR for anything," Smith writes. He argues this scenario is unlikely for AI because adoption is happening at an unprecedented pace. "Humans just know when a technology works. If AI weren't useful, we'd see people trying it for a while and then setting it aside. But we don't see that."
This distinction is crucial. Smith notes that while early skepticism is high, with MIT researchers finding that "95% of organizations saw zero return on their investment in AI initiatives," the actual usage data tells a different story. "As far back as a year ago, 40% of people were already using AI at work," he points out. The argument here is that the technology is real, even if the immediate financial returns are muddy. A counterargument worth considering is that high adoption rates do not guarantee profitability; users might rely on cheap or free tiers that never convert to the massive revenue streams needed to service corporate debt.
Humans just know when a technology works. If AI weren't useful, we'd see people trying it for a while and then setting it aside. But we don't see that.
The Railroad Parallel
The piece's most compelling insight reframes the potential bust not as a technological failure, but as a historical repeat of the Panic of 1873. Smith draws a direct line between the current data center spending spree and the railroad boom of the 19th century. "The railroad buildout in the 1800s was, in percentage terms, the greatest single feat of capital expenditure in U.S. history, dwarfing even what the AI industry is spending on data centers right now," he notes. Just as railroads transformed the American economy, they also triggered a decade-long depression because the infrastructure was built faster than the economy could generate the traffic to pay for it.
Smith explains that the railroad bust did not happen because America built too many tracks; it happened because the financing structure was fragile. "America financed its railroads faster than they could capture value," he writes. The full economic benefits, such as the Sears Catalog revolutionizing retail, only emerged fifteen years after the crash. "In other words, the great railroad bust did not happen because America built too many railroads. America didn't build too many railroads!" This historical context adds necessary depth to the current debate, suggesting that a financial crisis could occur even if the technology ultimately succeeds. Critics might argue that the modern financial system is more sophisticated than the 1870s, but Smith counters that the fundamental constraint remains: "There's basically no technological limit on how many... loans [the financial system] can disburse in a short period of time... And yet there is a limit on how fast businesses can create real value."
The Debt Trap
The core of Smith's warning lies in the balance sheets of the "hyperscalers" and the startups they fund. He outlines a simple but terrifying math problem: if companies spend more on data centers than they earn in profit, they are walking a tightrope. "If you're spending $70 billion a year things get dicey; you might have to take a couple of years of losses to pay back the loans if there's a bust," Smith explains. He highlights that while giants like Google and Microsoft can currently cover costs, the gap is widening. "Until recently, these 'hyperscalers' have been making enough cash to cover their AI spending. But spending is rising, so that may not be true for much longer."
The risk is even higher for companies that don't generate their own profits, such as OpenAI and Anthropic, which rely on borrowed money. "If AI takes 10 or more years to generate enough value to pay back all these debts, many of these companies could go bankrupt," Smith warns. This scenario implies that the financial system could seize up, not because AI failed, but because the timeline for value creation is too slow for the creditors. "Whatever financial institutions they've borrowed money from — private credit firms, banks, etc. — may fail or have to pull back significantly when their loans suddenly go bad."
If AI takes 10 or more years to generate enough value to pay back all these debts, many of these companies could go bankrupt.
The Airline Scenario
Finally, Smith introduces a third, often overlooked possibility: the "Airline Scenario." Even if AI generates immense value for the global economy, the companies building it might not capture that value. He suggests AI could become a commoditized, low-margin utility, similar to solar power or airlines. This would mean the technology transforms the world, but the investors who bet the farm on it see their returns evaporate. "Even if AI works and manages to create value very quickly, that value may not be captured by the AI companies themselves," Smith concludes. This reframing shifts the focus from the technology's potential to the business model's fragility.
Bottom Line
Smith's strongest contribution is decoupling technological success from financial stability, using the 1873 railroad crash to illustrate how a transformative era can still trigger a depression. His biggest vulnerability is the assumption that the timeline for value creation is fixed; if AI adoption accelerates faster than historical precedents suggest, the debt trap might be avoided. Readers should watch for the gap between capital expenditure and actual revenue in the coming quarters, as that spread will determine whether the next bust is a correction or a crisis.