Every financial mania eventually meets an accountant. Artificial intelligence will be no exception. The question isn’t whether investors are going to lose enormous amounts of money. They most certainly will. The question is whether the coming AI bust will expose another technological fantasy or simply a colossal mispricing of the most important economic development in human history.
I’m betting on the latter. I’ve witnessed the computer revolution, the personal computer revolution, the internet boom and the spectacular dot-com collapse. I’ve watched supposedly revolutionary technologies amount to almost nothing, while others that initially appeared insignificant went on to transform the world. Along the way, I’ve learned that the stock market’s assessment of a technology’s value and its actual economic value are often two entirely different things. AI is going to prove that point on a scale we’ve never witnessed.
The Bust Is Practically Guaranteed
We’re witnessing one of the greatest capital spending binges in history. Hundreds of billions are pouring into semiconductor factories, data centers, electrical infrastructure and companies whose valuations depend on extraordinary assumptions about future demand. The IMF estimates that worldwide private-sector AI investment could exceed $2 trillion in 2026.
And the financing is getting increasingly creative. Some companies investing in AI developers are also among the principal beneficiaries of those developers’ enormous infrastructure spending. If that sounds familiar to anyone who remembers previous financial manias, it should.
There’s nothing wrong with investing heavily in a promising technology. There’s something very wrong with assuming that every dollar invested will earn an acceptable return simply because the underlying technology is revolutionary.
The problem is particularly acute because AI is advancing so rapidly that today’s technological marvel can become tomorrow’s obsolete product. A company might spend billions developing a model, only to discover that a competitor has produced something substantially better at a fraction of the cost.
Think about that contradiction. The technology can become exponentially more useful while the companies producing it become progressively less profitable. The same competition that makes AI increasingly valuable to its users can make investing in AI providers a financial nightmare.
AI could become the greatest bargain in economic history and one of the worst investments in financial history. At the very same time.
Why the Dot-Com Comparison Misses the Point
Whenever the subject of an AI bubble comes up, someone inevitably trots out the dot-com collapse. The internet was revolutionary, they argue, but that didn’t prevent investors from losing trillions. Amazon survived, Pets.com didn’t, and the world kept turning.
Fair enough, but the comparison misses something fundamental. The internet revolutionized communication, information distribution and commerce. It made information easier to obtain and transactions easier to complete. But it didn’t actually perform most of the intellectual work required to run a business. It gave you access to information. You still had to figure out what to do with it.
AI is fundamentally different. It doesn’t simply deliver information. It can analyze information, draw conclusions, write computer programs, develop marketing campaigns, examine complex documents and help execute decisions. It’s increasingly performing tasks that previously required trained, experienced and expensive human labor.
Consider the implications for a small business owner. Work that once required a research department, marketing agency, programmer and several consultants can increasingly be handled by one person using inexpensive AI tools. The results aren’t always perfect, and professional judgment remains essential, but the economics have changed dramatically.
A competent individual can now undertake projects that would have been economically impossible just a few years ago. And we’re barely getting started.
The Productivity Gains Are Already Real
AI skeptics love dismissing the technology as another overhyped Silicon Valley fantasy. Unfortunately for that argument, we already have substantial evidence that the productivity benefits are real.
Researchers studying more than 5,000 customer service workers found that AI assistance increased productivity approximately 14 percent on average. Among less experienced workers, the improvement was roughly 34 percent. A separate Harvard Business School study found that consultants using AI completed certain tasks more than 25 percent faster while producing work rated substantially higher in quality.
These aren’t theoretical projections dreamed up by venture capital promoters. They’re measured improvements in actual work. Now imagine millions of accountants, lawyers, engineers, physicians, designers, programmers and business owners increasing their effective output by 20, 30 or 50 percent. Then consider what happens when those improvements spread through entire supply chains, with one company’s efficiencies reinforcing another’s.
The potential is staggering, especially when AI itself becomes an increasingly effective tool for developing better AI. Of course, not every application will succeed. Some will waste money, others will produce unreliable results and some businesses will discover that buying sophisticated technology doesn’t automatically make management competent. But you don’t need every business to succeed for the overall economic effect to be enormous. You just need enough of them.
Why There Really Is No Historical Comparison
The Industrial Revolution replaced human and animal muscle with machinery. The computer revolution brought unprecedented calculating power to business and science. Personal computers distributed that power to individuals, while the internet connected billions of people to a global information network.
Each advancement was extraordinary. But every one of them ultimately depended on human beings to supply the intelligence necessary to put the technology to work. AI is changing that equation.
For the first time, we’re developing a broadly applicable technology capable of assisting with the intellectual processes required to improve virtually every other technology. It can help design better machinery, develop more efficient manufacturing systems, accelerate scientific research, optimize transportation networks and improve the software controlling the modern world.
And unlike previous revolutions, AI doesn’t require entirely new distribution networks to reach billions of people. Much of the infrastructure already exists. Smartphones, computers and the internet provide the delivery system. Every major improvement can potentially reach millions of users almost immediately. The Industrial Revolution took generations to transform the economy. AI improvements can spread globally in weeks.
This is why I suspect the Industrial Revolution, computer revolution and PC revolution may eventually look like mere blips on the long-term chart of human economic advancement. Not because they were unimportant, but because they created the conditions for something vastly greater.
Previous technological revolutions amplified human muscle, memory and communication. AI is beginning to amplify intelligence itself. And intelligence is the essential ingredient in virtually every form of economic advancement.
The Greatest Deflationary Force Ever Created?
There’s another aspect of this revolution that deserves far more attention, particularly from those of us who follow monetary policy, government debt and inflation. Productivity ultimately means producing more output from the same resources. If AI allows companies to double their useful output without doubling their costs, the effect on unit production costs could be enormous. Multiply that across thousands of industries and we’re potentially looking at one of the greatest disinflationary forces in history.
That doesn’t mean Washington will suddenly discover fiscal responsibility. Our political class has demonstrated an almost supernatural talent for spending money it doesn’t have. But AI introduces an extraordinary countervailing force. On one side, governments continue increasing debt, money and credit. On the other, advancing technology makes it possible to produce increasing quantities of goods and services at declining real costs.
In the short term, ironically, the AI infrastructure boom can actually be inflationary. Electricity, copper, land, cooling equipment and skilled construction labor aren’t available in unlimited quantities. AI might eventually make intelligence extraordinarily cheap, but it still operates in a physical world with very real resource constraints. And those constraints create both extraordinary investment opportunities and the potential for catastrophic capital misallocation.
The Real Winners May Not Be AI Companies
Here’s something investors should seriously consider. The companies creating AI won’t necessarily be the companies capturing its greatest financial benefits.
A manufacturer using AI to reduce operating costs by 25 percent might enjoy extraordinary returns without spending a nickel developing an AI model. An insurance company could improve underwriting efficiency. A small business might eliminate administrative bottlenecks and suddenly become competitive with companies ten times its size.
Meanwhile, the AI provider responsible for those improvements might be locked in a brutal price war with five competitors.
That’s capitalism. Customers frequently capture enormous benefits from innovations while the companies financing those innovations struggle to earn satisfactory returns.
The real investment winners may ultimately be determined less by who develops the most sophisticated intelligence than by who uses increasingly inexpensive intelligence most effectively.
And the ultimate beneficiaries could be consumers, entrepreneurs and businesses that never invested a dime in AI infrastructure.
The Bust Could Actually Accelerate the Revolution
Suppose the AI bubble bursts. Stock prices collapse, financing dries up, overleveraged data center operators go bankrupt and investors abandon companies that were worth billions on paper but never figured out how to generate sustainable profits. What happens to the technology? Absolutely nothing that would erase its underlying usefulness. The knowledge doesn’t evaporate, the software doesn’t suddenly stop working and productivity improvements don’t reverse themselves because Wall Street has decided to sell.
Instead, bankruptcies and consolidations could make computing capacity available at distressed prices. Competition could drive AI service costs substantially lower. Businesses that could never justify building their own systems might suddenly gain access to extraordinary capabilities for pennies on the dollar. The investors who financed the infrastructure could lose their shirts. The businesses using that infrastructure could become substantially more profitable.
The greatest economic benefits of AI might actually arrive after a financial collapse, when speculative excesses have been purged and the surviving technologies become cheaper, more reliable and more widely available. That’s the magnificent irony of the coming AI bust. It could destroy enormous amounts of financial wealth while simultaneously accelerating the underlying economic revolution.
The Market Can Be Wrong About the Price and Right About the Revolution
None of this means the transition will be painless. Businesses will fail, professions will change, competitive advantages will disappear and millions of workers will be forced to adapt. There will be enormous winners and equally enormous losers.
But the magnitude of that disruption tells us something about the power of the technology itself. For centuries, economic progress depended on increasing our supplies of capital, labor, energy and natural resources, while finding more productive ways to use them. AI is beginning to change one of the most fundamental limitations on all economic activity: the amount of useful intelligence available to get things done.
We’re moving toward a world where increasingly sophisticated intellectual capabilities can be replicated and distributed at a scale that was previously inconceivable. We have never experienced anything quite like it.
That’s why I have no difficulty believing two things that initially appear contradictory. First, that an enormous AI financial bust is coming, perhaps sooner than most investors expect. Second, that AI will ultimately produce economic benefits so vast that today’s valuations and financial speculation will seem almost insignificant in retrospect.
A century from now, historians may view the Industrial Revolution, the computer revolution and the internet revolution as the opening acts of a far greater transformation. They may also wonder how investors could have been so spectacularly wrong about the financial value of individual companies while being so profoundly right about the importance of the underlying technology.
The coming AI bust may be one of the greatest wealth-destroying events in financial history.
And it may occur right in the middle of the greatest wealth-creating technological revolution mankind has ever experienced.
The two aren’t mutually exclusive.
In fact, I suspect they’re going to happen together.