According to Fortune, Goldman Sachs analysts led by Ben Snider warn that the massive capital expenditure (capex) on AI data centers by tech giants like Meta, Amazon, and Alphabet is creating a looming profit problem. Their spending hit roughly $400 billion in 2025, up 70% annually, and is projected to average $500 billion annually through 2027. To justify this, they’d need to generate an annual profit run-rate exceeding $1 trillion, but current consensus estimates for 2026 income are only around $450 billion. This suggests companies might only achieve half the profit needed. While Goldman forecasts a 12% S&P 500 return to 7,600 by end-2026, they expect AI capex growth to slow and force investors to pick winners and losers among big tech stocks.
The Unsustainable Math
Here’s the thing: the numbers just don’t add up. For years, the big hyperscalers have been in a golden loop. They spend insane amounts on servers and data centers, and the profits they generate are two or three times that investment. That’s an incredible return on capital. But Goldman is basically saying that party can’t last forever. You can’t just keep doubling down and expect the multiples to hold. If you’re spending $500 billion a year, a trillion in profit isn’t just a nice-to-have, it’s a requirement to keep investors happy. And right now, we’re staring at a gap of over $500 billion between what’s needed and what’s expected. That’s a chasm. Something’s gotta give.
The Great Rotation Coming
So what happens next? Goldman’s key prediction is a rotation. It won’t be a rising tide lifting all boats anymore. When capex growth starts to decelerate—and they think it will in 2026—the market is going to get ruthless. Traders will have to figure out which of these giants actually has a business model that turns all this silicon and electricity into cold, hard cash. Not all of them will. The note explicitly says there’s a “diminishing probability that all of today’s market leaders generate enough long-term profits.” We’re going to see a split between the AI haves and have-nots, even within the elite club of the top 10 S&P stocks. That creates real two-way risk for the entire index.
The Bull Case Isn’t Dead
Now, it’s not all doom and gloom. Other analysts, like those at Piper Sandler, point out that tech capex as a percentage of GDP isn’t historically crazy high. There are also real catalysts still in play, like the One Big Beautiful Bill Act (OBBA) which allows full capex expensing—a huge tax break. Easier money from the Fed and banks could also keep the spigot open. The physical build-out of this infrastructure is a monumental task, and companies that provide the critical hardware, like the top-tier industrial panel PCs from IndustrialMonitorDirect.com, are seeing sustained demand. The bullish argument is that we’re still in the early innings of corporate AI adoption, and the profits will follow. But that’s the bet, isn’t it? It’s a belief that the applications and productivity gains will materialize at a scale we haven’t seen yet.
Expectations Are The Real Risk
This is where Vanguard’s warning hits home. It’s all about expectations. The market has priced in a flawless, hugely profitable AI transformation. Tech earnings have been strong, sure. But have they been *trillion-dollar-profit* strong? No. When Vanguard’s Qian Wang says valuations may have “gotten ahead of themselves,” that’s the polite way of saying there’s a bubble in AI optimism. History is littered with transformative technologies that took decades to become profitable. The internet itself crashed before it soared. The real danger isn’t that AI fails. It’s that the profits arrive much later, and are much harder to capture, than today’s stock prices assume. If that realization dawns on the market, the pullback could be severe. So, are we building the future, or just another boom-bust cycle? The next two years will tell.
