Version 1

Meta released Muse Spark in April 2026, its most capable AI model to date, approximately one year after Mark Zuckerberg appointed Alexandr Wang — then 28, formerly Scale AI's founder — to lead a 'wartime-mode' AI reorganization [1][2]. A Financial Times investigation published June 3 examines whether Wang's unconventional appointment is working, finding genuine model progress alongside internal friction [3][1]. In parallel, a debate has opened over Meta's market valuation: some investors argue that Wall Street misreads Meta's AI capital expenditure as a drag, when it constitutes a durable advantage [11], while skeptics argue that aggressive talent spending does not guarantee Meta closes the gap with frontier rivals [6].

Meta is attempting a significant AI capability upgrade while operating at $1.5 trillion in market cap and facing established frontier competitors in OpenAI and Google. The Wang experiment tests whether an outsider's urgency can succeed where an incumbent's AI organization had struggled — and whether the answer affects how markets should price Meta's AI infrastructure spending.

Around April 2026, Meta announced Muse Spark, which the company described as its most powerful model yet and which current and former employees characterized as Meta's most credible AI effort to date [1][2]. The launch came roughly a year after Mark Zuckerberg made an unconventional organizational bet: handing responsibility for Meta's AI revival to Alexandr Wang, who was 28 at the time and best known as the founder of data-labeling company Scale AI, rather than to a veteran researcher from within Meta's established AI organization [1].

A Financial Times investigation published June 3 provides the most detailed account of this arrangement [3][1]. It finds genuine progress — Muse Spark is described as credible in a way prior Meta models were not — alongside persistent challenges. Wang reportedly faced criticism internally over his relative inexperience and the political difficulty of driving change inside a large incumbent organization. A Medium piece went further, framing the restructuring in skeptical terms and describing Wang as a potential early casualty of a broader internal overhaul [4]. Separately, reporting noted that Meta made substantial compensation offers in the AI talent competition but that money alone may not be sufficient to close capability gaps with rival labs [5][6].

The Muse Spark release has been received unevenly. CNBC covered it as a significant development [7], Simon Willison noted the model's associated chat tools as genuinely interesting [8], and CNBC separately reported that Wall Street sees promise in the model but still wants Zuckerberg to articulate a clearer long-term AI strategy [9]. A Reddit discussion captured a common outside read: Muse Spark may not be the best model of 2026 by benchmark measures, but it could be among the most practically deployed [10].

The sharpest debate concerns Meta's valuation. A widely circulated Milk Road AI thread argues that Wall Street is fundamentally misreading Meta's AI capital expenditure — treating it as a liability when it is, in their framing, the competitive moat [11]. The thread cites Jensen Huang's public statement that 'nobody uses AI better than Meta' as supporting evidence [12]. Fortune counters that spending on AI talent is not sufficient to guarantee Meta catches up to rivals who have built more sustained research compounding [6].