Meta hires Alexandr Wang to lead AI push as Zuckerberg takes charge of selling it

Meta is trying to prove that its massive AI spending can become a real business, not just an internal upgrade to its advertising machine. A year after spending more than $14 billion to recruit Alexandr Wang and a team of top Scale AI engineers, the company has regained some momentum in artificial intelligence, but it remains well behind OpenAI, Anthropic and Google in market perception and developer enthusiasm.

The biggest milestone so far was the release of Muse Spark in April, Meta’s first major proprietary foundation model. The launch marked a shift away from Meta’s long-running open-source strategy, which had centered on the Llama family of models. That earlier approach was meant to win favor with developers, but it lost momentum after Llama 4 failed to impress last year, prompting CEO Mark Zuckerberg to change course.

Wang now leads Meta Superintelligence Labs, the unit created to strengthen Meta’s position in the most competitive part of the tech industry. The new model is designed less as a standalone developer platform and more as a tool that can power Meta’s own ecosystem, including Facebook, Instagram, the Meta AI app and site, and devices such as Ray-Ban Meta glasses. Analysts say that internal focus makes sense, but it also raises a key question: can Meta turn AI into a direct revenue source?

Wall Street remains unconvinced. Meta’s stock has fallen 18% over the past 12 months, making it the weakest performer among megacap tech stocks, even though the company posted 33% revenue growth in the first quarter, its fastest pace since 2021. Investors appear to want evidence that Meta can attract paying users for AI products, rather than relying on AI only to improve ad targeting and efficiency.

That challenge is particularly important because advertising still generates 98% of Meta’s revenue. Since Muse Spark launched, Meta has introduced new AI and business subscription plans, signaling a broader push beyond ads. But analysts and industry figures say the company still needs more proof that these products can gain traction and eventually generate meaningful revenue.

Meta also faces skepticism from developers. Some say the company has lost trust after moving away from the open-weight strategy that initially made Llama appealing. Others note that Meta has released only one major model under Wang’s leadership so far, and that it has not been widely accessible to the broader AI community. Still, Meta says it plans to give outside developers access to Muse Spark’s underlying technology through an API, with early testing already underway.

Beyond the product strategy, Meta is also dealing with internal pressure, including recent layoffs and questions about morale and safety priorities. Wang and other new hires are expected to deliver not just technical progress, but business results. For Zuckerberg, the stakes are high: Meta’s AI push must now prove it can become a durable growth engine, not another expensive bet.