Did Meta just make a multi-billion dollar bet that’s already crumbling? Fresh off a massive investment, the tech titan is now turning to Scale AI’s rivals for critical AI model training. Insider reports reveal surprising quality concerns and executive departures. What does this mean for Meta’s ambitious AI future?

Just two months after a colossal $14.3 billion investment by Meta into the data-labeling titan Scale AI, the foundational tech partnerships appear to be unexpectedly fracturing. This significant outlay was intended to fortify Meta’s ambitious push into advanced AI development, bringing Scale AI CEO Alexandr Wang and several key executives into the fold to lead Meta Superintelligence Labs (MSL). However, recent developments suggest a turbulent beginning for this high-stakes collaboration.

Early signs of strain emerged with the swift departure of Ruben Mayer, Scale AI’s former Senior Vice President of GenAI Product and Operations, from Meta after a mere two months. Mayer, who previously spent five years with Scale AI, was reportedly overseeing AI data labeling operations at Meta but was notably absent from TBD Labs, the crucial unit spearheading the company’s core artificial intelligence superintelligence initiatives, which has also attracted top researchers from industry leaders like OpenAI.

Beyond personnel shifts, Meta’s strategic approach to its AI development seems to be diversifying. Insiders reveal that TBD Labs is actively engaging with third-party data labeling vendors, distinct from Scale AI, for the training of its next-generation Meta AI models. These include Mercor and Surge, both significant competitors in the artificial intelligence data space.

While it is customary for artificial intelligence labs to collaborate with multiple data labeling providers, the current situation is particularly striking given Meta’s substantial financial commitment to Scale AI. Despite the multi-billion-dollar investment, various sources indicate that researchers within TBD Labs have expressed concerns regarding the perceived quality of Scale AI’s data, showing a preference for the offerings of Surge and Mercor.

Historically, Scale AI built its business on a crowdsourcing model, utilizing a vast, cost-effective workforce for basic data labeling. Yet, as Meta AI models and the broader artificial intelligence landscape demand greater sophistication, the need for highly specialized domain experts—such as medical professionals, legal experts, and scientists—to generate and refine premium data has become paramount. This shift highlights a critical challenge for tech partnerships in a rapidly evolving sector.

Scale AI has endeavored to adapt by introducing its Outlier platform to attract these subject matter experts. Nevertheless, competitors like Surge and Mercor have experienced rapid growth, largely because their business models were established from the outset on a foundation of highly remunerated, skilled talent, offering a perceived advantage in high-quality AI data labeling for complex AI development.

A spokesperson for Meta AI has publicly disputed any claims of quality issues with Scale AI’s product, while Surge and Mercor declined to comment on the matter. When questioned about Meta’s increasing reliance on rival data labeling providers, a Scale AI representative referred TechCrunch back to their initial announcement of Meta’s investment, which highlighted an expansion of their commercial tech partnerships.

Meta’s diversification into deals with additional third-party data labeling vendors strongly suggests that the company is not solely dependent on Scale AI, even after its immense investment. This stands in stark contrast to Scale AI, which, shortly after Meta’s announcement, reportedly saw other significant clients disengage from their tech partnerships. The implications for Scale AI’s market position are considerable as Meta AI charts its own course.

The broader AI development unit at Meta has reportedly faced increasing internal complexities since Wang’s arrival and the influx of new researchers. Both new talent from OpenAI and Scale AI, along with existing Meta AI team members, have voiced frustrations with navigating the company’s extensive bureaucracy, indicating a rocky start for what was intended to be a stabilizing and accelerating force in Meta’s pursuit of artificial intelligence superintelligence. The question remains whether Meta can stabilize its operations and retain the crucial talent needed for future success in this highly competitive domain.