AI Spending Slowdown? Goldman Sachs Warns of a 20% Market Drop

The rally in U.S. equities over the past two years has one clear driver: artificial intelligence. From Nvidia’s meteoric rise to Microsoft’s massive cloud-AI push, Wall Street has been betting that AI will not just transform businesses but also justify sky-high valuations. But Goldman Sachs recently cautioned that if AI spending growth cools, the S&P 500 could face a 15–20% pullback.

AI Spending Could Slow Down

Ryan Hammond, an analyst at Goldman, laid out the concern in a recent note. He argues that hyperscalers—companies like Microsoft, Meta, Alphabet, and Amazon—will eventually have to dial back their AI investments. Current expectations assume that capital expenditure can keep climbing at an extraordinary pace. But history tells us that no company, no matter how deep its pockets, can spend indefinitely at such levels.

Hammond puts it bluntly: if AI spending expectations revert to where they stood at the start of 2023, the S&P 500’s valuation could shrink by as much as one-fifth. And given how concentrated the index has become, that warning matters. Nvidia alone now accounts for roughly 7% of the S&P 500. The eight biggest tech names—each heavily invested in AI—make up more than one-third of the index’s total market cap. In short, the AI trade isn’t just a sector story; it is the market story.

Still, the Money Keeps Flowing

What makes this tricky is that we’re still seeing headline-grabbing commitments. Meta recently announced plans to spend $600 billion on AI over the next three years, with Zuckerberg even hinting that the figure could grow. Microsoft just inked a $17.4 billion deal with Nebius for AI infrastructure. These aren’t signals of companies easing up.

But here’s the catch: Wall Street is forward-looking. Investors aren’t asking whether 2024’s AI budgets are big—they’re asking whether 2026’s will be bigger. Goldman thinks that’s where the growth curve could flatten, and once markets start to price in slower growth, the valuation impact could be severe.

Déjà Vu from the Dot-Com Era?

Layered on top of Goldman’s warning is another debate: are we in an AI bubble? Bret Taylor, OpenAI’s board chair, thinks the answer is yes. In a recent interview, he echoed Sam Altman’s view that we’re in the middle of a hype cycle. He didn’t mince words: a lot of people are going to lose a lot of money.

But Taylor also added nuance. The fact that we’re in a bubble doesn’t mean AI is overhyped as a technology. He compared it directly to the late 1990s dot-com boom. Back then, countless startups failed spectacularly—Webvan, Pets.com, and others—but the internet itself proved transformative. Amazon and Google, born in that era, are now among the world’s most valuable companies.

Taylor’s key point: bubbles often get the direction right but the timing wrong. The capital poured in during the internet bubble funded infrastructure that was later critical to the digital economy. Similarly, today’s wave of AI investment may leave behind bankruptcies, but also the hardware, models, and data pipelines that future AI applications will depend on.

AI Market Is Still Immature

Another reason this feels like a bubble: the spending-to-value gap is wide. Many companies are still in what Taylor calls “AI tourism”—dabbling with pilots and experiments that don’t deliver much real ROI. MIT research backs this up, finding that many AI investments haven’t yet boosted productivity.

Taylor believes the problem is maturity. Just as in the early days of the internet, business models aren’t clear, and companies often try to build solutions themselves rather than buy ready-made ones. He predicts the next wave of AI will be defined by domain-specific applications: legal AI (like Harvey), customer service AI (like Sierra), and other specialized agents that solve clear problems. When that shift happens, AI investment will likely look more efficient—and more sustainable.

My Take

To me, the Goldman and Taylor views actually fit together. Yes, the market is over-reliant on AI right now, which means valuations are vulnerable if spending slows. But yes, AI is also going to change the economy in a way that justifies much of today’s excitement—just not in a straight line.

If we are replaying the dot-com script, then volatility is part of the bargain. The S&P may indeed face a 15–20% pullback if AI capex plateaus, but over the long run, today’s spending spree could build the foundations of tomorrow’s AI-driven economy.

The challenge for investors is separating the winners from the tourists. Betting on every AI name is like buying every dot-com in 1999—you’ll end up with more Pets.coms than Amazons. But dismissing the whole sector is just as risky.