You can’t blame investors for being skeptical of Meta’s enormous AI outlay

There’s blood in the water. After a period of huge bullishness over generative AI, investors are starting to freak out about the massive costs involved, as evidenced by the reaction to Meta’s earnings yesterday.

To be fair, there’s more than one reason why the Facebook and Instagram parent’s share price took a 15% dip in after-the-bell trading, once its quarterly results were out. While Q1 revenues were strong, with no pullback in Chinese ad-buying, Meta’s Q2 guidance was a bit lower than Wall Street hoped to see. But the scale of its AI investment—with Meta’s estimate for the year now up 12% to $35 billion to $40 billion in spending and with next year’s outlay set to be even greater—is what proved truly sobering, especially with so little to show for it just yet.

CEO Mark Zuckerberg asked investors for patience and pointed out that Meta had “historically seen a lot of volatility in our stock during this phase of our product playbook where we’re investing and scaling a new product but aren’t yet monetizing it.” He also talked up the possibility of AI features in Meta’s apps attracting more users and therefore more ad revenue, and AI-powered smart glasses taking off, but that wasn’t enough to calm investors’ nerves.

“The disappointment on the revenue side is overshadowing any optimism about AI,” Cresset Wealth Advisors chief investment officer Jack Ablin told Bloomberg. “It’s hard to tell what the benefit will be to users, and while AI could ultimately mean some cost savings down the line, that isn’t visible yet.”

It is certainly true that big ideas need big investments, but if we’re talking this big, it’s very understandable that investors want to see some kind of solid strategy for profitability.

This is particularly true in the case of Meta, which—lest we forget—was until recently going all-in on the metaverse, another very expensive endeavor that has so far turned out to be a money pit (Meta’s Reality Labs division lost over $16 billion last year, and $3.85 billion in Q1). Yes, AI has way more immediate, demonstrable uses than virtual reality does, but seeing it as a profitable venture still requires a hefty dollop of imagination.

Another likely factor here is Meta’s open-source approach to releasing its Llama large language models. There are many real advantages to this, such as ensuring that there’s a powerful and attractive alternative out there to models from the likes of OpenAI and Google, and getting the wider community to help improve Meta’s AI. It may well pay off, but many investors will look at this strategy and ask where the moat is, and how Llama will make them tons of money.

Is the reaction to Meta’s outlay a damning indictment of Big Tech AI investments in general? We’ll see after Microsoft and Alphabet’s results later today, but it’s worth remembering that those companies are also major cloud providers whose AI capex will likely pay off one way or another. Meta may find itself in a similar position after its spending spree—Zuckerberg hinted at the possibility—but right now that’s not the business it’s in.

At this point it’s worth remembering that, like Elon Musk at Tesla, Zuckerberg is in the unusual position of being a law unto himself. Meta’s board can’t fire him, because of his voting control, and he has previously said that this allows him to take unpopular decisions that end up paying off. Meta’s vast AI expenditure could prove to be a serious test of that structure’s real utility, but either way, there’s nothing to stop investors from punishing the company if they don’t like what they see.

More news below.

David Meyer

Want to send thoughts or suggestions to Data Sheet? Drop a line here.

This story was originally featured on Fortune.com