The conventional wisdom in tech is that it’s better to overspend than miss out on a blockbuster hit. Of course, Big Tech’s investors hate when their companies overdo it. They showed that through their sell-off of Alphabet Inc. last week when it said it would lavish more cash than expected on artificial intelligence, and again with their drubbing of Meta Platforms Inc. shares when it stated much the same.
It didn’t help Mark Zuckerberg’s company that it reported its second-quarter earnings on the same day as Microsoft Corp., whose AI spending tends to deliver immediate returns via its Azure cloud business. Meta can’t translate its spending into revenue so easily. AI has improved its ad targeting, but there are limits to how much more it can squeeze from that. Even so, the Facebook founder and amateur mixed martial arts fighter has articulated his AI game plan more clearly than before, and it sounds more plausible than anything he ever said about his virtual-reality misadventure, the Metaverse.
Meta’s plans encompass three key areas: personal AI agents, business agents and renting out data centers to other companies. All three are in their nascent stages or non-existent in terms of sales, but Zuckerberg has solid foundations to make them viable. The Metaverse, by contrast, forced him to build a new computing platform from scratch, a bet that led to more than $60 billion in losses and not many users.
AI is a much bigger wager, of course. Meta has plowed about $110 billion into building data centers and buying Nvidia Corp. chips over the past two years — and it plans capital spending of $130 billion-$145 billion in 2026. It’s not slowing down, either. Chief Financial Officer Susan Li says Meta is “maximizing 2026 and 2027 capacity.”
So it’s a mild relief that Zuckerberg has at least started to articulate more coherently how he aims to make money from all this capex, indicating on this week’s earnings call that his all-star team of AI researchers and product builders may eventually earn its keep.
Meta last year spent $14.3 billion for a stake in Scale AI, bringing over its founder Alexandr Wang to lead Zuckerberg’s new Superintelligence Labs team, tasked with turning it into a leader in “personal superintelligence.” And despite a series of management exits and rumblings of Game of Thrones-style politicking within the team, it appears to be upgrading some Meta products in a meaningful way. One new model, Muse Spark, now underpins the Meta AI assistant, and Zuckerberg said the upgrade has coincided with a 60% jump in the number of people using the chatbot.
The team is also developing Muse Image and Muse Video, aimed at “content creators.” However repulsed you may be by AI slop, Zuckerberg contends it will spark “a whole new and nearly infinite universe of personalized content,” which could keep people hooked even longer on their phones.
That dystopian-yet-lucrative future is a credible prospect. When AI was used to improve the recommendation engines on Meta’s social-media feeds, there was double-digit growth in the time people spent looking at its apps, according to Zuckerberg. In other words, he can already turn dials with AI to get people to spend more time scrolling stuff they didn’t ask for.
If businesses can significantly improve their ads using Muse Image and Muse Video, that will be an early sign that Meta is translating its heavy AI investment into revenue. The other route will be through personal AI assistants for consumers, and an upgrade to Meta AI. Around a billion people talk to the chatbot on WhatsApp, Facebook Messenger and Instagram each month, but Zuckerberg wants them to do more.
He hinted about the launch soon of a “great consumer product that just works out of the box and is easy enough for billions of people” to use. His challenge — easier in theory than asking millions of people to wear a heavy virtual-reality headset for the Metaverse — will be to encourage users to shift from just talking to Meta AI to asking it to do things for them, like sending birthday greetings and checking in with friends.
There’s a test case for that among technologists and computer programmers, many of whom already talk to AI agents through chat apps such as Telegram, and get them to do things like write code, troubleshoot problems and automate repetitive tasks.
While it’s still an open question whether any of these revenue streams for AI will work, Zuckerberg has laid out a commercial roadmap more clearly than he did for virtual reality. It might not have the cleaner economics of Microsoft’s vast cloud business, but for the first time in a while Meta’s spending looks more like a calculated investment than a leap of blind faith.
A message from Advisor Perspectives and VettaFi: Discover something new! Click here to register for our upcoming webcasts.
Bloomberg News provided this article. For more articles like this please visit
bloomberg.com.
Read more articles by Parmy Olson
