Facebook parent company Meta—which lost over $83 billion since 2021 trying to convince people the metaverse was the next internet frontier—has finally burned enough cash to catch up with the AI frontier, CEO Mark Zuckerberg said on Monday.
Zuckerberg told the FT on Monday that Meta’s new Muse Wonder AI model would “allow much more sophisticated behaviors that are at the frontier of <a href="https://bitcomme.com/what-is-an-analytics-engineer/” title=”What Is an Analytics Engineer?”>what is being done in the AI field.” He added, “Getting to this point both required a large amount of compute and investment, and a lot of very rigorous scientific work by the group working on it over an extended period of time.”
The Muse app has indeed taken off. According to Barron’s, after debuting in early September, Muse hit the No. 1 spot on Human Security’s list of top agent-based traffic sources on the internet, accounting for 72% of the total. Analysts at Citigroup are bullish on Muse, recently projecting $27 billion in annual revenue by 2030 largely based on transaction and subscription revenue before advertising. Muse’s agentic abilities have sparked concern in other sectors (like travel agencies, tax preparers, and car valuation websites, Reuters reported) that Meta could be about to eat their lunch.
Yet there are also the jaw-dropping expenditures on compute that make Muse possible in the first place and the chance that this product could be joining the market just as the AI bubble begins to strain. The tens of billions spent on the metaverse push, for example, pale in comparison to data center buildout costs. As of October 2025, Meta projected that it would spend <a href="https://www.reuters.com/business/meta-plans-600-billion-us-spend-ai-data-centers-expand-2025-11-07/” rel=”nofollow noopener” target=”_blank”>at least $600 billion on data center infrastructure and jobs by the end of 2028, a push that is possibly unsustainable if debt investors start to balk. There have also been reports that Meta only found serious security flaws in the product right before launch, including one that could have let malicious users escape Meta’s sandbox and access sensitive, internal databases and systems. These safety questions are particularly relevant because Muse allows Meta to scoop up more user data than ever.
Zuckerberg told the FT that he doesn’t support slowing down AI development, as Anthropic’s Dario Amodei proposed and OpenAI’s Sam Altman endorsed, and that Muse’s differentiator is its focus on individual consumers rather than large enterprises. (It’s worth noting that Meta has also added features to Muse focused on small businesses, and Monday’s announcement included plans for Meta to directly compete with Anthropic and OpenAI for the enterprise market by selling access to the latest Muse Wonder models.) He seemed less concerned about AI superintelligence than about a lack of it.
“Everyone else is focused on more knowledge work, automation-type cases, and we’re much more focused on personal empowerment,” Zuckerberg said. He argued that Meta wants to build “things that are good in the world” and that the company is investing so heavily in AI because agents “still don’t have enough common sense to be the ideal personal assistant.”
“I don’t know what that means, like, sociologically that you know that we got AI systems that could do these advanced mathematics stuff before we got something that could handle customer support reliably for everyone. But I think that there’s obviously a lot more to do there,” Zuckerberg added.
He also addressed skepticism that Meta, which has countless privacy scandals under its belt, can be trusted with even more personal data. Zuckerberg characterized the distrust as a spur to innovation, not a flaw.
“Most of the ways in which people might not trust the company can lead to technological innovations that could be quite productive,” the CEO told the FT, pointing to the upcoming Muse “Confidential VM” feature that purportedly walls off private user data from being seen by Meta.
