The AI buildout has hit the stage that anyone who's ever lived through a renovation recognizes: you started out replacing a vanity, and now there's a structural engineer in the kitchen and everyone keeps saying "while we're in here..."
In buildout terms: the spending is accelerating, the borrowing has just begun, and the part where it pays off is still theoretical.
The companies doing it—the handful of tech giants building AI's data centers (hyperscalers)—were long known for throwing off cash. Now, their expected free cash flow, the money left over after operating costs and construction bills, has gone negative, and they're borrowing so they can keep building. Meanwhile, the chipmakers (the primary sellers of what all that money buys) have watched their expected free cash flow climb sharply. The homeowners are borrowing; the contractor's truck keeps getting nicer.
Insuring hyperscaler debt against default over the next five years has gotten more expensive. And earnings reactions suggest investors are reading past the headline. Companies beating estimates have underperformed the S&P 500, which is not how beats have historically worked, while those missing estimates have underperformed by less than usual. This pattern suggests the market cares less about whether the quarter beat the estimates and more about whether the spending behind it pays off.
The thing to watch from here is what the companies funding the buildout tell investors to expect about revenues and margins, because the borrowing and the construction budgets are both expected to keep growing in the near term.
Our view hasn't moved. We continue to favor high-quality businesses that can raise prices without losing customers and adjust when conditions do; the kind of company that doesn't need this particular bet to pay off.
Advisor Angles
Anyone who's lived through a renovation knows how this goes. You replace one thing, discover three others, and suddenly the budget looks very different. That's where the AI buildout is today. The issue isn't whether the project matters. It's whether the eventual payoff justifies the growing cost.
Clients asking whether AI has become "too big to fail" may be asking the wrong question. The bigger story is that some of the world's strongest companies are borrowing to keep building. That's not a warning sign by itself, but it does raise the bar. Spending is no longer the story; results are.
Strong companies don't always get rewarded for strong earnings. Markets care about the future, not the report card. That's why a company can beat expectations and still see its stock fall if investors come away less confident about what happens next.
A new technology can change the world without rewarding every investor who chases it. Railroads transformed transportation and the internet transformed commerce. Both created tremendous economic value, but not every company involved created lasting shareholder value. That's one reason diversification still matters, even when a trend looks unstoppable.
The headlines focus on who is spending the most. We're paying just as much attention to businesses that don't need a single theme to work perfectly. Companies with pricing power, resilient demand, and management teams that adapt tend to have more ways to win.
One thing we're watching closely is whether all this spending starts showing up in revenues, margins, and cash flow. Building the infrastructure is the easy part. Proving it was worth doing is what comes next.