Everybody keeps telling me AI is a bubble.
And maybe they’re right.
Companies are spending an almost ridiculous amount of money building data centers, buying Nvidia (NVDA) chips and finding enough electricity to power all of it.
The numbers have gotten so large that I don’t think most of us can even comprehend them anymore.
A billion dollars used to sound like a lot of money.
Now we’re talking about trillions.
So naturally, people look at all this spending and say the same thing: This can’t possibly continue.
Maybe it can’t.
But Barron’s recently pointed out something that I think makes this whole conversation a lot more interesting.
America has done this before.
When railroads were being built across the country in the 1800s, spending eventually reached roughly 25% of the size of the entire U.S. economy.
Something similar happened during the internet boom.
Barron’s calls this the “rule of 25.”
It’s not actually a rule. There isn’t some alarm that goes off when spending reaches exactly 25% of the economy.
But it’s an interesting way to compare today’s AI boom with some of the biggest investment booms in American history.
And if we use that same math today, we get a pretty crazy number.
The U.S. economy is now roughly $32 trillion. Twenty-five percent of that is about $8 trillion.
Eight trillion dollars.
So where are we now?
Barron’s estimates that more than $1 trillion has already been poured into AI infrastructure since 2024. And the spending is accelerating. Goldman Sachs estimates roughly $581 billion will be invested in AI in the U.S. this year alone.
That sounds enormous. Because it is.
But now compare it with that $8 trillion historical measuring stick.
Even after everything we’ve already spent, we could still be talking about trillions of dollars of additional investment before we get anywhere close to the levels that accompanied some of America’s previous great infrastructure booms.
That doesn’t mean we’re definitely going to $8 trillion. It doesn’t mean stocks can’t crash before we get there. And it certainly doesn’t mean every dollar being spent today is going to produce a good return.
But it does raise an interesting question.
What if all this spending that looks so crazy today is still just the beginning?
The Bubble Can Be Right
Here’s the part I think people get confused about.
A bubble doesn’t necessarily mean the thing everyone is excited about is stupid.
Railroads really did change America. Before them, moving people and products across the country was slow and expensive. Railroads connected cities, created new businesses and completely changed the way the economy worked.
Investors went nuts for them. Eventually, they went too nuts.
Railroad companies borrowed enormous amounts of money to keep building. Investors kept throwing money at new projects.
Soon, the money stopped working, financing dried up, and the whole thing came crashing down.
But here’s the funny part: America still needed the railroads.
The technology wasn’t the problem. The economics were.
We saw something similar with the internet.
People love making fun of the dot-com bubble because some ridiculous companies were worth ridiculous amounts of money.
But investors weren’t wrong that the internet was going to change the world.
They weren’t bullish enough!
Think about what happened after the dot-com bubble burst. The internet became even more important than almost anyone imagined.
Today, we shop on it, watch television on it, talk to our families on it, work on it, and carry it around in our pockets everywhere we go.
The internet was very real.
That didn’t stop the Nasdaq from falling almost 80%.
That’s the lesson I keep coming back to when people tell me AI has to be a bubble because we’re spending too much money on it.
Those two things can both be true.
AI could change the world.
And investors could still lose a ton of money along the way.
Watch the Money
This is why I don’t think $8 trillion is some magical number we should put on the calendar.
Maybe the problems start at $4 trillion. Maybe we get to $10 trillion without a crisis.
Nobody knows.
And these comparisons aren’t perfect anyway.
Some of the AI spending estimates we hear are global, while that $32 trillion number is the size of the U.S. economy. We’re not trying to solve an equation here.
The point is to understand the scale.
What really matters is what we’re getting for the money.
If a company spends $100 billion building AI infrastructure and that investment eventually produces $200 billion in profits, spending $100 billion wasn’t crazy.
If it produces $10 billion, we’ve got a problem.
That’s really what this entire AI debate is about now.
The largest technology companies in the world are spending hundreds of billions of dollars building the infrastructure they believe will power the next generation of computing.
Barron’s estimates that hyperscalers could spend trillions globally through the end of this decade.
And they aren’t alone.
Everybody wants in.
That’s where things get interesting for investors.
As long as customers keep showing up, revenues keep growing, and companies can justify spending more money to make even more money, this thing can keep going much longer than people think.
History suggests these enormous technology booms don’t usually end because somebody on Twitter notices that spending looks high.
They end when the money stops working.
That’s what happened with the railroads. That’s what eventually happened during the internet boom. And if AI turns into the next great investment bust, I suspect we’ll see something similar.
The technology won’t suddenly become useless.
We won’t wake up one morning and decide artificial intelligence was all a big mistake.
More likely, companies will eventually build too much, borrow too much, or discover that all those expensive data centers aren’t producing enough money to justify building the next one.
Maybe we’re already closer to that point than we realize.
But this Barron’s comparison suggests another possibility that I find much more interesting.
What if we’re nowhere near it?
Stay sharp,
JC Parets, CMT
Founder, TrendLabs