Mark Zuckerberg says it is better to 'misspend' billions of dollars than risk losing out on 'superintelligence'

Mark Zuckerberg

Meta CEO Mark Zuckerberg says he’s willing to spend staggering sums to make sure the company doesn’t miss out on artificial intelligence — even if it means overshooting. Speaking on the Access podcast, Zuckerberg admitted an AI bubble is “quite possible,” but argued that moving too slowly is the greater danger.

“If we end up misspending a couple of hundred billion dollars, that is going to be very unfortunate,” he said. “But the risk is higher on the other side.” For Zuckerberg, the nightmare scenario isn’t wasting money — it’s being caught flat-footed if artificial superintelligence arrives sooner than expected.

Meta has already committed at least $600 billion through 2028 to US data centres, infrastructure, and talent, a figure CFO Susan Li later clarified as covering both AI buildouts and broader business operations. The spending spree underscores Zuckerberg’s view that “compute per researcher” can be a competitive advantage, as Meta snaps up GPUs and builds custom systems to outmuscle rivals.

The push comes amid mounting talk of an AI bubble, with parallels drawn to the dot-com crash. Meta itself has trimmed costs elsewhere, pausing aggressive hiring after paying hefty signing bonuses in the AI talent wars. Wall Street has warned that stock-based pay could dilute investors without clear returns.

Zuckerberg also drew a contrast with rivals OpenAI and Anthropic, which rely on outside fundraising to fuel their massive compute bills. A downturn, he noted, could cut them off. Meta, by comparison, has the balance sheet to power through.

To prepare for superintelligence, Meta has assembled a small, elite “superintelligence lab” with flat structures and no strict deadlines, reflecting its research-heavy mission. The stock, meanwhile, is up nearly 40% over the past year.