Alexandr Wang leads Meta’s AI effort as spending expands

Meta says Wang founded Scale AI in 2016, when he was a 19-year-old MIT student, and that the company reached a valuation of nearly $29 billion under his leadership.[1] Meta paid $14.3 billion for a 49% stake in Scale AI as part of bringing Wang aboard.[3] Zuckerberg said he selected Wang after speaking with leaders of AI research labs; Wang sent Zuckerberg a long memo two days after their call.[3] Zuckerberg recalled that Wang’s first principle was to prioritize talent density over hiring many people.[3] The Financial Times reported that Wang later presented Meta executives with a detail-light “vibe report,” which did little to reassure finance chief Susan Li.[2]

Meta introduced Muse Spark as the first model in the Muse family, developed by Meta Superintelligence Labs.[4] The company describes it as a natively multimodal reasoning model with tool use, visual chain of thought and multi-agent orchestration.[4] It is available at meta.ai and in the Meta AI app, with a private API preview opening to select users.[4] Meta reported that Muse Spark’s Contemplating mode scored 58% on Humanity’s Last Exam and 38% on FrontierScience Research.[4] The company also said Spark could reach the same capabilities as Llama 4 Maverick with more than an order of magnitude less compute.[4] These are company-reported measures, not an independent evaluation in the cited evidence.

Meta later introduced Muse Wonder, which it called its first frontier-class model.[3] Wang said Wonder was state-of-the-art on personal and professional agent tasks according to Meta’s measurements.[3] An independent comparison put Muse Wonder Full roughly level with leading OpenAI and Anthropic models, but a couple of points behind the leader.[3] The Muse app launched in the United States on September 8 and exceeded 2.5 million downloads in about two weeks.[3] The Wall Street Journal reported that Meta’s stock had risen nearly 20% since Muse’s early-September debut and that the app had led Apple’s App Store for two weeks.[5] These figures describe early reception; they do not establish long-term model performance or product adoption.

Meta’s 2026 capital-expenditure forecast is $130 billion to $145 billion, nearly twice the $72.2 billion it spent the previous year.[3] Zuckerberg said Meta would write a check for about $100 billion or $200 billion.[3] A warning in the reporting said that failure of the AI effort would be very bad financially.[3] Zuckerberg also said Meta was unhappy with Llama 4’s trajectory and prepared to rebuild its AI team.[3] That context makes the cost of the effort relevant to business leaders weighing Meta’s AI products: the announced models are one part of a much larger commitment, and the available evidence does not show whether that investment will pay off.