There is one argument about the AI bubble that keeps bothering me.
People say AI can’t be a bubble because the demand is obviously real.
And they’re right.
We use it every day.
We ask questions.
Write emails.
Summarize things.
Code.
Research.
Study.
Make spreadsheets.
Plan trips.
Companies are putting it into customer service, software, cybersecurity, sales and basically every job where people sit in front of a computer.
But then I saw a simple response to this argument that I can’t get out of my head.
I use the Internet every day but the dotcom bubble still happened.
That’s the whole problem.
The Internet was real.
People needed it.
People used it.
Businesses were moving online.
The technology changed the economy.
And investors still managed to massively overpay for the companies building it.
The Internet surviving did not save the dotcom stocks.
In fact, many of the companies that eventually became huge were nowhere near the prices investors were paying for them during the bubble.
That is why “everyone uses AI” doesn’t really answer the investment question.
It answers whether AI is useful.
Yes.
It does not answer whether AI stocks are priced correctly.
And the amount of money already being attached to this future is getting enormous.
AI related companies have added roughly $27 trillion in market value since late 2022, according to Goldman Sachs.
Goldman estimates the present value of potential additional profits created by AI productivity at roughly $9 trillion.
That doesn’t mean the market is definitely $18 trillion too high. There are a lot of assumptions in both numbers.
But it shows the size of the bet.
The market has already assigned an enormous amount of value to something that still has to produce the profits.
And the spending underneath the story keeps getting bigger.
The largest cloud and computing companies are expected to spend around $754 billion on capital expenditures in 2026, according to Goldman Sachs.
That is up 83% from 2025.
Goldman also says AI infrastructure beneficiaries could account for roughly half of S&P 500 earnings growth this year and next year.
So now we have a strange situation.
AI is real.
Demand is real.
Usage is growing.
Corporate adoption is growing.
But the stock market has already moved very far ahead of the technology itself.
Goldman estimates AI related companies gained about $27 trillion in market value while the potential AI profit pool it calculates is around $9 trillion.
That gap has to be closed somehow.
Either the profits become much bigger.
Or the prices eventually come down.
There is another number I find even more interesting.
The S&P 500 forward P/E ratio was around 22 times earnings earlier this year, approaching the 24 times level reached in 2000.
That doesn’t mean 2026 is 2000.
The businesses are much stronger.
Many of today’s companies actually make enormous profits.
That is an important difference.
But it also makes the comparison more uncomfortable.
Because the dotcom lesson was never that the Internet was fake.
The lesson was that a real technological revolution can still become a terrible investment when expectations get too far ahead of reality.
AI doesn’t have to fail.
It doesn’t have to disappear.
People don’t have to stop using it.
The technology can become one of the most important technologies in history.
And investors can still lose a fortune.
That’s the part I think gets lost in the AI bubble debate.
The question isn’t whether we will use AI.
We obviously will.
The question is how much of that future has already been stuffed into today’s stock prices.
The Internet became essential.
The dotcom bubble still happened.
Those two things were both true.
And maybe that’s the better way to think about AI too.