Meta CEO Mark Zuckerberg fielded a barrage of investor questions on Wednesday about where exactly the billions of dollars Meta is pouring into AI are going.
Meta expects capital expenditures to climb sharply through 2026 as it races to build computing infrastructure and hire top AI researchers, the company revealed in its third-quarter earnings report. In a call with analysts, Zuckerberg defended this strategy as essential to Meta’s push to compete with AI rivals.
Meta CFO Susan Li said the company now expects to spend $70 billion to $72 billion on building AI infrastructure this year, up slightly from the $66 billion to $72 billion projected last quarter.
She added that spending will climb even higher in 2026 as Meta pours money into data centers, cloud services, and salaries for the growing ranks of AI researchers and engineers hired for Meta Superintelligence Labs, Zuckerberg’s new division focused on developing AI that can eventually surpass human capabilities.
Li said those hires, along with rising infrastructure and cloud costs, will make employee compensation the company’s second-largest driver of expense growth next year, after infrastructure itself.
“Employee compensation costs will be the second largest contributor to growth as we recognize a full year of compensation for employees hired throughout 2025, particularly AI talent, and add technical talent in priority areas,” Li said on the earnings call.
Naturally, Wall Street wanted to know where all that money is going — and how it will eventually translate into profits.
When JPMorgan analyst Doug Anmuth pressed him on the ballooning costs, Zuckerberg responded that spending big on AI was the smart play, even if Meta ends up overshooting.
In what Zuckerberg called the “very worst case,” the company would simply have “pre-built for a couple of years,” eating some depreciation costs and eventually growing into the extra capacity, he said.
Better that, Zuckerberg said, than being caught short on compute when the next wave of AI breakthroughs hits.
“Rather than continuing to be constrained on capex and feeling in the core business like we have significant investments that we could make that we’re not able to make,” he said, “the right thing to do is to try to accelerate this to make sure that we have the compute that we need, both for the AI research and new things that we’re doing.”
Analysts from Goldman Sachs, Bernstein, and Bank of America also pressed Zuckerberg on what Meta gets in return for all this investment and how soon it might pay off. Zuckerberg said Meta’s AI tools already reach more than a billion people each month across its apps, and that engagement continues to grow as the company rolls out more advanced models.
Related stories
Business Insider tells the innovative stories you want to know
Business Insider tells the innovative stories you want to know
He added that the next wave will come from Meta Superintelligence Labs, which is building what he called “truly frontier models” that could power new kinds of assistants, content-creation tools, and business products.
While it’s still early, Zuckerberg said, smarter models are already improving Meta’s core business: powering better recommendations and more effective ads that help the company make money.
Barclays analyst Ross Sandler had a basic question: how is Meta going to make money from all this talk of superintelligence and AGI? (Artificial general intelligence, or AGI, refers to a milestone where AI can match or surpass human capabilities.)
Zuckerberg’s answer boiled down to this: every leap in AI capability will feed directly into Meta’s apps and ad business.
“The research is going to enable new technological capabilities to exist,” he said, “and then those capabilities can get built into all kinds of different products.”
Smarter reasoning models could power assistants, business tools, and advertising agents, he continued, while better video generation tech could help creators make more content and help advertisers sell more stuff.
“Being the best in a given area will drive great returns,” Zuckerberg said. “This is not like a check-the-box exercise.”
Meta’s stock plunged nearly 9% after the earnings report landed.
Have a tip? Contact Pranav Dixit via email at pranavdixit@protonmail.com or Signal at 1-408-905-9124. Use a personal email address, a nonwork WiFi network, and a nonwork device; here’s our guide to sharing information securely.
