The AI slowdown story is getting treated like one giant trade.
I don’t think it is.
The people building the models are saying slow down.
The people building the data centers are still spending hundreds of billions.
And the market is suddenly trying to figure out which one matters more.
Dario Amodei called for a slower pace of frontier AI development.
Sam Altman backed him.
Elon Musk backed him.
Trump basically said the opposite. Whoever wins AI wins.
Then the stocks got crushed.
The semiconductor index fell 5.4% Monday. Nvidia dropped around 3%. Micron, AMD, Marvell and other AI hardware names got hit even harder.
But look at what happened underneath the headline.
Alphabet was up.
Microsoft was up.
Meta was up.
The companies buying enormous amounts of AI infrastructure held up much better than the companies selling the infrastructure.
That tells you what Wall Street is actually worried about.
Not that AI disappears.
Not that everyone stops using ChatGPT.
Not that data centers suddenly become useless.
The fear is that the insane rate of spending eventually slows.
And that matters because the numbers are insane.
The biggest hyperscalers are on track to spend roughly $700 billion this year on capital expenditures, with AI infrastructure taking a huge share.
Oracle alone spent $28.5 billion in its latest quarter and kept its $90 billion to $95 billion annual capex forecast. Its cloud infrastructure revenue is growing 121% year over year.
So the “AI is over” crowd is getting ahead of itself.
But the “nothing changed, buy everything” crowd is doing the same thing.
There is another problem.
S&P Global says hyperscaler credit quality is gradually weakening because capex keeps rising, financing is getting more complicated and the returns on all this spending may take years to arrive.
That is where this gets uncomfortable.
Imagine AI keeps improving.
Great.
Now imagine AI keeps improving but a little slower than Wall Street expected.
The data centers still get built.
The chips still get bought.
The power demand still grows.
But maybe the next $100 billion of spending takes longer to show up.
That’s enough to hurt a stock.
Especially when investors have already priced in years of explosive growth.
There is also a weird point buried in the AI debate.
Some people are saying the newest models barely feel better than the last ones.
Other people who actually use the newest models with huge amounts of compute say the opposite.
Both can be true.
AI progress can be real while the improvement becomes harder and more expensive to achieve.
And that might be the real question now.
How much more money does it take to get the next 10% of AI capability?
If the answer keeps going up, the economics get harder.
If the answer suddenly goes down because models become far more efficient, the chip trade could get hit even while AI itself gets better.
That’s why I don’t think “AI slowdown” automatically means buy the dip.
And I don’t think it means AI is finished either.
The market is finally starting to separate the AI winners from the companies that simply benefited from the assumption that AI spending would keep accelerating forever.
That distinction could get ugly.