The AI trade has become a religion. And like any religion, it demands faith — faith that the trillion-dollar valuations, the endless hype cycles, and the promises of a productivity revolution will all pay off. But on Tuesday, Goldman Sachs threw cold water on the altar.
In a note that landed like a brick through a stained-glass window, Goldman's equity strategists warned that investor assumptions about AI are stretching reality to the breaking point. The message: you're betting on miracles, not margins.
The Numbers Don't Lie — But Investors Do
Goldman's analysts ran the math. They looked at the revenue growth needed to justify the current valuations of AI stocks — the Nvidias, the Microsofts, the Alphabet's of the world. And they found a gap. A big gap.
"Current pricing implies that AI will deliver a level of economic impact that is historically unprecedented," the note read. Translation: you're betting on something that has never happened before. Not once. In any industry. At any scale.
"Current pricing implies that AI will deliver a level of economic impact that is historically unprecedented."
Let's be clear: AI is real. It's not a fad like crypto or a gimmick like the metaverse. Large language models are already changing how people work, search, and create. But the market isn't pricing in gradual change. It's pricing in a revolution — and revolutions, historically, have a nasty habit of disappointing.
The Productivity Mirage
The core assumption fueling AI mania is that the technology will unlock massive productivity gains across the economy. Every company will cut costs, fire workers, and watch profits soar. But Goldman's note points out a dirty secret: we've heard this before.
In the 1990s, the internet was going to revolutionize everything. It did — eventually. But the first wave of internet stocks crashed hard. In the 2000s, robots were going to steal all the jobs. They didn't. Productivity gains from technology have been lumpy, unpredictable, and often overestimated.
Goldman's strategists argue that AI will likely follow the same pattern. There will be winners — big ones. But the market has already decided that everyone is a winner. That's not investing. That's gambling.
The Cost Problem Nobody Wants to Talk About
Here's the part that gets glossed over in every breathless earnings call: AI is expensive. Really expensive. Training a single large language model costs tens of millions of dollars. Running inference — the act of actually using the model — burns through server capacity like a wildfire through dry brush.
Goldman's note flags the capital expenditure required to sustain the AI boom. Data centers, chips, electricity, cooling. It's a staggering bill. And right now, the companies footing that bill are doing so on the promise that future revenues will cover it. That's a bet that looks a lot like WeWork's "community-adjusted EBITDA" — creative accounting designed to make terrible economics look sane.
The companies footing that bill are doing so on the promise that future revenues will cover it. That's a bet that looks a lot like WeWork's "community-adjusted EBITDA."
The markets have been conditioned to ignore costs. But costs have a way of becoming real when the easy money dries up.
The Echo Chamber Problem
There's a reason Goldman's note matters: it comes from inside the machine. This isn't a random blogger screaming that the sky is falling. This is Goldman Sachs — the same firm that helped inflate the dot-com bubble and then profited from its collapse. They know a mania when they see one, because they've been selling tickets to the circus for decades.
And what they're seeing now is an echo chamber. Every analyst, every conference, every earnings call — it's all AI, all the time. The hype has become self-reinforcing. Companies mention AI in their earnings calls, and their stocks go up. So they mention it more. And it goes up more. Eventually, the connection between reality and stock price becomes tenuous at best.
Goldman's note is a reality check. It's not predicting a crash — they're too smart for that. But they're saying the assumptions are stretched. And when assumptions stretch, they eventually snap.
What This Means for Your Portfolio
If you're heavily invested in AI stocks, here's the uncomfortable question: are you betting on the technology or the narrative? Because the technology is real, but the narrative has gotten ahead of itself. The gap between what AI can do today and what the market is pricing in is wider than the Grand Canyon.
Does that mean you should sell everything and hide in cash? No. But it does mean you should be asking harder questions. Which companies have actual revenue growth from AI? Which ones are just riding the wave? The difference will become brutally apparent when the tide goes out.
Goldman's note doesn't name names, but you can do the math yourself. Look at the companies that have doubled in the last year. Look at their earnings. Look at their cash flows. If the story is better than the numbers, you're not investing — you're speculating.
The Bottom Line
The AI trade has been a gift for anyone who got in early. But gifts have a way of turning into traps. Goldman Sachs is saying what many analysts are thinking but few have the courage to write: the assumptions are stretched, the costs are real, and the history of revolutionary technologies is littered with the corpses of companies that promised more than they could deliver.
The question isn't whether AI will change the world. It will. The question is whether the market is pricing in a miracle that won't arrive on schedule. And if history is any guide, the answer is yes.
So keep your eyes open. Question everything. And remember: when Goldman starts warning about stretched assumptions, it's usually because they've already placed their own bets on the other side.
