When will AI's stock-market dominance fade? Goldman says watch this key signal.

• Amazon, Meta, Microsoft, and Alphabet have spent hundreds of billions on AI in the past four quarters.
• Goldman Sachs says ongoing AI-stock outperformance could eventually be challenged.

Artificial intelligence stocks are ripping higher, with the likes of Nvidia and Microsoft propelling the market to new records. But Goldman Sachs says that if these companies want to maintain their lofty valuations, they'll eventually need to have something to show after spending big.

The firm notes that Amazon, Meta, Microsoft, and Alphabet have poured a combined $357 billion into capital expenditures and research and development over the past four quarters. Of that nine-figure sum, a "significant portion" has been allocated to to AI, Goldman says.

With that in mind, the authors of the note — led by Ryan Hammond — are closely watching downward revenue revisions for signs the AI spending isn't paying off. If those results don't materialize, they anticipate stock losses ahead for the market's biggest leaders.

"Today's hyperscalers will eventually be required to prove that revenues and earnings will be generated from their investments," analysts wrote. "Early signs that those revenues and earnings may not be generated, as measured by negative sales revisions, could lead to valuation de-rating."

Goldman notes that worries have started to creep into their conversations with market participants.

"Investors have expressed uncertainty about the return on those investments for the mega-cap tech stocks, but these stocks remain extremely popular," the analysts wrote.

They continued: "Even among investors that are long-term bullish on the potential gains from AI adoption, there appears to be considerable uncertainty about the timeline."

Goldman also points out that, according to a measure it tracks, just 5% of companies are currently using AI to produce goods and services.

Still, the disconnect between capital spending and actual sales and profit is not as pronounced now as it was during past periods of perceived excess.

"Adjusting for profits of these companies, the AI capex cycle still pales in comparison to the tech bubble," the analysts wrote.