You can believe AI will change everything and still get hurt by how everyone else is betting on it.
That is the message from Gabriela Santos, chief strategist for the Americas at JPMorgan Asset Management. "You can be very bullish on all things AI and still need to think very carefully on portfolio construction," she said.
Her point is simple. The trade is everywhere. That means the usual labels investors use to spread out their risk no longer work the way they used to.
"That's where it's getting complicated," Santos said. "Because you can't just think of traditional factors, or sectors or regions. Or even asset classes, because that AI tentacle is everywhere now."
The proof came in July, when tech stocks suddenly turned south.
That kind of move does not happen by accident. It happens when too many investors pile into the same trade, and then a few start heading for the exit at once.
Santos said the episode should make investors check three things: how much they own, how much borrowed money they are using, and whether their holdings reach beyond AI at all.
The catch is that most portfolios already have AI exposure baked in, even when it is not obvious. A fund that owns "defensive" stocks might still hold companies selling chips to data centers. An "international" fund might lean heavily on the Korean and Taiwanese names that power the AI supply chain.
The bond market is where the warning gets sharper.
Companies have been selling investment-grade debt at record pace for four straight months. Alphabet is among the issuers selling bonds that do not mature for 100 years. That means bond investors are now lending money to the same AI giants whose stocks they already own.
So the person holding a broad stock fund and a corporate bond fund has AI risk on both sides of the ledger. When one side wobbles, the other may not catch it the way it used to.
Santos also flagged that debt from the big cloud companies, known as hyperscalers, should be reviewed one issue at a time rather than treated as a single group. Vehicles backed by leases on data centers have become more complex, she noted, which makes them harder to judge at a glance.
Only a small set of assets still appears to offer genuinely different return patterns, in her view. That list is short: U.S. Treasuries, gold, and core real estate.
Here is the scale of what is being built.
What makes this buildout unusual, she said, is that it is already showing up in corporate profits. Past technology booms often required years of losses before the money started flowing back.
Analysts at Goldman Sachs projected in June that spending on AI data centers will likely top $900 billion in 2026, with 2027 forecasts reaching as high as $1.4 trillion. Those are staggering numbers, and they help explain why investors keep piling in.
But here is the tension. No matter how fast AI stays the leading investment theme, its growth will eventually decelerate. Santos called the buildout "very unique," and she thinks starting to diversify now is the prudent move rather than waiting for the slowdown to arrive.
The bottom line: You do not have to bet against AI to protect yourself from its popularity. The question is whether your portfolio is built for the moment when everyone tries to leave the same trade at once.