Meta is pouring nearly $15 billion into data labeling startup Scale AI, snapping up a 49% stake and bringing CEO Alexandr Wang onboard to lead a new “superintelligence” lab.
The move echoes Meta’s past big bets like the $19B WhatsApp buy and $1B Instagram deal. Many investors are already skeptical this weekend, wondering if Meta’s overpaying again—or if this will cement another major win for Mark Zuckerberg.
Scale AI powers data for top AI labs like OpenAI. Frontline AI research depends heavily on labeled data. Recently, Scale AI started hiring PhDs and senior engineers for high-quality frontier AI data. 
Meta’s AI efforts have faltered. Their GenAI lineup Llama 4 lagged behind competitors like DeepSeek. Meta also faces talent drain—losing 4.3% of its top AI staff in 2024, according to SignalFire.
Meta’s betting on Wang. The 28-year-old Scale CEO is savvy, well-connected, and meeting world leaders on AI’s future. But he lacks a deep AI research background seen in other lab heads like Ilya Sutskever or Arthur Mensch.
That’s why Meta is also recruiting big names like DeepMind’s Jack Rae for its new lab.
The future of Scale AI post-deal is unclear. Some AI groups bring data labeling in-house or use more synthetic data. In April, Scale missed revenue targets ahead of its share sale, per The Information.
Anyscale co-founder Robert Nishihara told TechCrunch this is a dynamic challenge:
“Data is a moving target,” Nishihara said. “It’s not just a finite effort to catch up — you have to innovate.”
Meta’s exclusive deal could push other AI labs toward Scale competitors like Turing, Surge AI, or new players like LM Arena.
Turing CEO Jonathan Siddharth said interest is climbing amid Meta’s move:
“I think there will be some clients who will prefer to work with a partner that’s more neutral,” Siddharth stated.
Meta faces stiff competition. OpenAI preps GPT-5 and its first publicly accessible model in years, setting up a direct battle with Meta’s Llama series.
Meta’s $15B bet on Scale AI and Wang is a massive gamble to regain ground in AI. Whether it pays off remains to be seen.
