Meta is betting the house on AI — CFO warns the stakes are 'really bad' if it flops

Oh, to be Meta's ( META ) very well-regarded CFO, Susan Li right now.

Li — who has worked at Meta for 18 years, mostly in financial roles — is the person tasked with overseeing the company's financial statements and ensuring Meta invests in projects with the highest potential return for investors. It's also her job to assess risk and push back against the growth-at-all-costs duo of Mark Zuckerberg and Alexandr Wang.

Right now, the job sure seems like a more Herculean task compared to years past.

Wang revealed in a new episode of Stripe ( STRI.PVT ) co-founder John Collison's Cheeky Pints podcast that Li pulled him aside "many times" during an executive offsite in Hawaii to warn that if the company's newest model didn't work, it would be bad news financially.

"And I think this is, and I kind of viscerally remember this. We were in Hawaii at the end of last year. It was kind of like the whole exec team got together," Wang said. "And yeah, Susan pulled me aside many times and was like, you know, I just want to tell you if this doesn't work, this is going to be really, really bad. It's going to be very, very bad in financial terms. It's like, it has to be good. If it's good, it's all good. We're great. But if it's bad, it's really, really, really bad."

Meta's new model appears to work quite well. Now people just need to pay for it to justify the gargantuan investment Li signed off on.

The company officially debuted its next-generation frontier AI models, Muse Wonder and Muse Wonder Full , this week from its Meta Superintelligence Labs.

These frontier models represent a dramatic surge in intelligence by featuring deep multistep reasoning, dynamic memory retention, and native multitool execution capabilities across web and desktop environments.

"Yesterday's launch of Muse Wonder and Wonder Full is Meta's first frontier-class release and, in our view, is at or near parity with leading frontier labs," said Pivotal Research analyst Jeff Wlodarczak. "On Meta's benchmark set Wonder Full leads on roughly 7 of 14 tests in Meta's benchmark set (including internal ones), with strong results on long horizon coding, legal agents, and memory, while scoring second on the independent Artificial Analysis Index (56 vs 58 for Claude Opus 5.5. Wonder is priced at $4/$20 per million tokens, matching Opus 5.5, while the stronger Wonder Full is sold only through Muse Code subscriptions ($50-$200 a month) rather than the API, which supports our view that Meta will monetize through paid tiers and enterprise rather than open weights."