The AI trade is morphing into something different. After a year of watching mega-cap tech stocks ride the infrastructure wave, investors are starting to wonder if the real money might be somewhere else entirely. Goldman Sachs thinks so—and they're pointing to a cohort of companies that aren't building AI, but are quietly deploying it to slash costs and boost productivity.
It's not that the infrastructure story is dead. It's just that valuations have gotten stretched, volatility is creeping back in, and smart money is starting to look for the next chapter. Third-quarter earnings for the S&P 500 crushed expectations, but the underlying question remains: how long can this last if the returns stay concentrated in a handful of chipmakers and cloud providers?
Everyone's Talking About AI, But Who's Actually Making Money From It?
If you tuned into earnings calls this quarter, you probably heard the letters "AI" more times than you care to remember. Nearly 47% of S&P 500 companies brought it up during their Q3 calls—a noticeable jump from the previous quarter. Communication Services led the pack at 74%, with Financials close behind at 66%.
The use cases sound familiar by now: automated coding, AI-powered customer support, marketing optimization, expense management, supply chain tweaks. About 30% of the companies mentioning AI cited coding automation or customer service bots as their main applications.
But here's where it gets interesting—or frustrating, depending on your perspective. Only a tiny fraction of these companies actually put a number on what AI is doing for them. Most are still in the "we're exploring this" or "we see potential" phase. The ones that did quantify impact, like ServiceNow Inc. (NOW) and C.H. Robinson Worldwide Inc. (CHRW), reported genuine margin improvements. But they're outliers, not the norm.
Meanwhile, the stock market returns from AI exposure in 2025 have remained stubbornly concentrated in the infrastructure layer—chips, cloud services, hardware. The builders are still winning, even if the users are starting to multiply.
The Hunt for Phase Two
That concentration is making investors nervous. Goldman Sachs analyst David J. Kostin put it bluntly in a recent report: "The combination of continued corporate AI adoption and concerns about the risks to the AI infrastructure complex has increased recent investor focus on the next beneficiaries of the ever-expanding AI trade."
Translation: people are worried about overbuilding in the infrastructure space, and they're looking for the next place to park their capital before everyone else figures it out.
Goldman is highlighting what they call "AI productivity beneficiaries"—companies with high labor intensity and significant exposure to automation. The thesis is straightforward: while the hype has pumped up valuations for the companies building AI tools, the actual returns might come from businesses that simply use those tools effectively to cut costs and operate more efficiently.
It's less sexy than building the next breakthrough model, but it might be more profitable. And as the infrastructure frenzy starts to cool, this quieter story is just beginning to unfold.
Who Makes the Cut? Goldman's Productivity Screen
So who are these mystery companies that might ride the next AI wave? Goldman built a screen to find them, and the criteria were pretty specific. To make the list, a company had to:
- Rank in the top 25% of the Russell 1000 based on labor costs as a percentage of revenue
- Have a high proportion of wages exposed to AI automation
- Have mentioned AI on earnings calls this year—specifically in connection with efficiency improvements
They also excluded infrastructure companies and firms primarily selling AI products. The goal was to isolate the users, not the builders.
What emerged was a surprisingly diverse mix across finance, consulting, software, and HR services. Here are some of the names that stood out:
These aren't the companies making headlines with splashy AI product launches or massive capital expenditures. But Goldman notes they're already showing something more valuable: better earnings revisions, quietly outpacing the broader market.
"Despite the recent underperformance, clients have increasingly expressed an interest in identifying high-probability productivity beneficiaries of AI," Kostin said. "These companies may represent the next leg of the AI trade."
The argument makes sense. If you're a bank or a consulting firm with thousands of employees doing repetitive tasks, even modest automation gains can translate into serious margin expansion. You don't need to invent the next ChatGPT—you just need to use the existing tools well.
And while the infrastructure names have dominated returns so far, that doesn't mean they're the only game in town going forward. The real test of AI's economic value might not be who builds the best models, but who deploys them most effectively. That's the bet Goldman is highlighting, and it's one that more investors seem ready to explore.
