Mark Zuckerberg said he is shelling out big money to ensure Meta doesn’t miss the AI moment.

The Meta CEO said on an episode of the “Access” podcast published Thursday that an AI bubble is “quite possible.” He pointed to historical precedents where companies overbuilt, collapsed, and left valuable assets behind. But for Meta, he said the larger risk is hesitating.

“If we end up misspending a couple of hundred billion dollars, I think that that is going to be very unfortunate, obviously,” he said. “But what I’d say is I actually think the risk is higher on the other side.”

Zuckerberg said that if a company builds too slowly and artificial superintelligence arrives sooner than expected, it’ll be “out of position on what I think is going to be the most important technology that enables the most new products and innovation and value creation and history.”

“The risk, at least for a company like Meta, is probably in not being aggressive enough rather than being somewhat too aggressive,” he added.

The tech giant has poured money into AI infrastructure. Zuckerberg earlier this month said Meta would spend at least $600 billion on US data centers and infrastructure through 2028. Meta‘s chief financial officer, Susan Li, later clarified that the figure covers all of Meta’s US data center buildout in the US as well as “all the investments that go towards supporting our US business operations,” including new hires.

The spending pledge comes as talk of an AI bubble reaches a fever pitch. Some investors fear the industry is overheating, drawing parallels to the dot-com bubble burst in 2000.

Meta has also shown signs of pulling back in other areas. Last month, the company pumped the recruitment brakes after shelling out huge signing bonuses to lock down hires in the cutthroat AI race. The hiring freeze came as Wall Street scrutinized Meta’s labor costs and warned that stock-based compensation could dilute shareholder value without clear innovation gains.

Zuckerberg or Meta did not respond to a request for comment from Business Insider.

How Meta prepares for superintelligence

Zuckerberg contrasted Meta and other AI labs like OpenAI and Anthropic, which rely on fundraising to cover their enormous compute bills.

“We’re not at risk of going out of business,” he said on the podcast. Private companies, like OpenAI and Anthropic, face the question of whether they can keep raising capital. That depends not only on their performance and the trajectory of AI but also on broader economic conditions, he added.

A market downturn triggered by global events could quickly leave them unable to cover their massive computing costs.

“It might be a different situation if you’re in one of their shoes,” he said.

The company’s stock is up nearly 40% in the last year.

Zuckerberg said Meta is preparing for superintelligence by concentrating elite talent in a small, flat “superintelligence” lab — with no top-down deadlines, to reflect the research-heavy nature of frontier AI. He said the company is also making “compute per researcher” a competitive advantage, outspending rivals on GPUs and the custom infrastructure needed to power them.