Meta is spending heavily to rebuild its AI effort, and Mark Zuckerberg has described the stakes in unusually blunt terms: if its new model fails, the financial consequences could be “really, really bad”. The company’s 2026 capital spending forecast is $130bn to $145bn, while its new Muse Wonder model is the first from Meta Superintelligence Labs.
Meta AI Watch analysis
What happened
In a Fortune interview, Zuckerberg and Meta AI chief Alexandr Wang described how they began working together after Meta’s Llama 4 models received a lukewarm response. Wang sent Zuckerberg a memo recommending a focus on talent density and a willingness to start again; Zuckerberg said he chose Wang to lead the effort despite his lack of experience training frontier models.
The resulting research team has fewer than 200 people, closer to 100, Wang said. Fortune reports that Muse Wonder is the lab’s first model and that Wang calls it state of the art for personal and professional agent tasks, based on Meta’s own measurement. The outlet also says independent rankings put it roughly level with leading OpenAI and Anthropic models, but a couple of points behind the leader. Read Fortune’s account.
Why it matters
Meta has put a specific price tag on the push: its expected 2026 capital expenditure is nearly double the $72.2bn it spent last year. That makes the performance of its AI effort more than a prestige contest. The scale of the investment, and the model’s standing against competitors, will matter to investors as well as people deciding which AI tools to use.
Our read
A compact research team and a model that can hold its own against established rivals would be a notable result. But “state of the art” is Wang’s description, not a settled verdict, and the available comparison puts Muse Wonder behind the leader. Meta has made the stakes clear. Now it needs results that justify the cheque.
What to watch
- Whether Muse Wonder’s performance claims hold up in further independent comparisons.
- How Meta’s AI products develop alongside the research team’s work.
- Whether the company’s spending forecast changes as the results come in.
Discussion spark: Should Meta keep making frontier-scale bets if its models remain close to, but not ahead of, the leading competitors?
Sources and evidence
- Source update (17 October 2026, 11:55 UTC)
not affiliated with, endorsed by, or operated by Meta