AI Doubts Are Growing. Investors Could Be Making A Mistake.

One explanation for the downturn has been an investor rotation from Big Tech to small-caps. The tech-led Nasdaq Composite began sinking in relation to the Russell 2000 index around July 10. Analysts across Wall Street seized on the Russell’s breakout to dust off neglected small-cap names.

But no single factor explains market trends, and those little Russell stocks can’t absorb all the megacap money. The diverging fortunes of the two indexes only retraces a few months of the Nasdaq’s outperformance of the Russell and the S&P 500 this year.

Another explanation for the sentiment shift is that investors now fear spending on artificial intelligence is about to slow.

Big Tech has wowed the world with new generative-AI platforms. For most tech companies, that’s meant huge spending on AI chips. Nvidia has been the primary beneficiary, as the primary vendor of those chips. Everyone has been cheering as Amazon, Apple, Alphabet’s Google, Meta Platforms, and Microsoft spend tens of billions of dollars apiece on AI data centers.

The gen-AI models can write computer code and compose songs. One thing they haven’t yet done is generate a lot of revenue for anyone other than Nvidia and a few other hardware vendors.

It’s entirely fair to ask what returns investors can expect from those billions in capital expenditures. Back in May, I raised that question and found no clear answers.

When I queried the AI models themselves, they forebodingly mentioned workforce reductions as the first outcome.

As a journalist, I’m loath to take credit for ideas. Others have also wondered about the return on investing in AI. In June, David Cahn, a venture capitalist at Sequoia Capital, updated an influential analysis he first put out last year, in which he figures that Nvidia’s customers will need to add $600 billion in AI revenue to get an acceptable payback on their investments. As of now, he thinks they’ll be lucky to collectively get $100 billion in new revenue.

Cahn agrees that AI will transform the economy, as the internet and the railroads once did. And Sequoia has backed AI start-ups. But Cahn adds that many speculators in those previous technologies got torched when initial valuations crashed—to the benefit of everyone subsequently using rail or internet service.

As Meta stock was faltering two weeks ago, Morgan Stanley analyst Brian Nowak also considered what returns on invested capital the Facebook parent might expect from the $45 billion it plans to invest in capex this year and again in 2025. Concluding that AI will drive higher engagement at Facebook and Instagram, Nowak stuck with his Buy recommendation for Meta.

The point of this column isn’t to say that the Mag 7 selloff represents a better pricing of the returns the companies will get from their AI investments. No one can put a fine point on that forecast, any more than early investors could predict how cities would develop along rail lines or how social media would emerge from the internet.

I’m actually making the opposite argument. If anyone sold stocks in the belief that AI investment will slow, they haven’t been listening to the people doing the spending. Big Tech companies will continue to buy AI hardware from Nvidia and the other vendors.

When Alphabet’s June-quarter report this past week showed that capital spending had nearly doubled from the year-ago period, CEO Sundar Pichai told investors that the company had to invest aggressively up front in AI, which promises to transform all of the company’s core businesses—from search to YouTube to cloud computing.

“When you go through a curve like this, the risk of underinvesting is dramatically greater than the risk of overinvesting for us,” he said.

It isn’t just internet firms. Moderna co-founder Noubar Afeyan recently told me about plans by his biotech venture firm Flagship Pioneering to spend some of the billions it recently raised on AI models for drug discovery. “We don’t know if it’s going to work,” he said. “Do you sit there and let somebody else figure out if it works? No.”

No one can predict the winners of the AI sweepstakes in the long run, but in the near term you had better believe the money will keep flowing.

Write to Bill Alpert at william.alpert@barrons.com