Meta’s bold $14.3 billion investment in a 49 per cent stake in data‑labeling firm Scale AI—designed to accelerate its AI advancement—has quickly encountered turbulence. Only months into the venture, key technical leaders from Scale AI and veteran Meta researchers have exited, while concerns over data quality and internal culture clashes have prompted some Meta teams to turn to rival providers.
Meta appointed Scale AI’s founder, Alexandr Wang, to oversee its Superintelligence Labs, signalling high ambition. Yet several AI experts—including Avi Verma, Ethan Knight and Rishabh Agarwal—left the lab shortly after joining, with some returning to firms like OpenAI; Agarwal cited a desire for different challenges in line with leadership’s own advice. The stream of departures extends to Meta’s established AI personnel, such as Chaya Nayak, further exposing internal friction.
Culture clashes have intensified, as Wang’s closed‑door approach and aggressive restructuring are at odds with Meta’s former open‑source ethos. Longtime engineers are resisting demands to reevaluate or abandon previous work in response to the new direction. The company’s AI division has been reorganised into four units—research, superintelligence development, products, and infrastructure—with major staffing reviews underway, fuelling discontent.
Alongside internal upheaval, the deal’s consequences are reverberating across the AI data industry. Major clients—Google, OpenAI, and xAI among them—have paused or halted projects with Scale AI, citing concerns about conflicts of interest and data privacy following Meta’s involvement. Scale AI responded by emphasising its operational independence and continued neutrality despite the investment.
Market disruption has followed. With key clients pulling back, Scale AI laid off about 14 per cent of its workforce—including staff and contractors—just a month after the deal closed. The company cited internal inefficiencies and overexpansion, even as it restructured into core functional teams and pivoted toward enterprise and government sectors.
Meta’s leadership remains undeterred in its AI drive. The company has offered extravagant compensation to recruit top AI talents and is reportedly planning capital expenditures upward of US $70–72 billion. Yet analysts warn the aggressive “buy‑and‑build” approach may lack strategic cohesion and synergy, especially given past AI misfires.
Within Meta’s labs, researchers are increasingly sourcing training data from competitors, bypassing Scale AI’s platform amid trust concerns. Reports suggest that models trained on Scale’s datasets show more errors, hallucinations and shortcomings in advanced reasoning tasks, prompting internal reliance on providers such as Surge or Mercor.