Cracks are forming in Meta’s partnership with Scale AI

Since June, Meta's substantial investment of $14.3 billion in Scale AI has brought CEO Alexandr Wang and key executives from the data vendor to spearhead the Meta Superintelligence Labs (MSL). However, signs of strain are emerging in this partnership. Notably, Ruben Mayer, formerly Scale AI's Senior Vice President of GenAI Product and Operations, left Meta just two months after his arrival, according to sources familiar with the situation. Mayer's tenure at Meta involved overseeing AI data operations teams and reporting directly to Wang. However, he was not included in the core TBD Labs team, which is dedicated to developing AI superintelligence and has attracted top researchers from OpenAI. Despite requests for comment, Mayer did not respond. Furthermore, it's reported that TBD Labs is collaborating with other data vendors, including Mercor and Surge—both significant competitors of Scale AI—to train its upcoming AI models. In the competitive landscape of AI, it is common for labs to work with multiple vendors. However, Meta's significant investment in Scale AI makes the situation particularly intriguing. Despite this financial commitment, some researchers within TBD Labs have reportedly expressed concerns about the quality of Scale AI's data, showing a preference for the offerings from Surge and Mercor. Scale AI initially thrived on a crowdsourcing model, utilizing a large, low-cost workforce for straightforward data annotation tasks. As AI advancements require increasingly sophisticated data, Scale AI is now attempting to attract domain experts through its Outlier platform. Meanwhile, competitors like Surge and Mercor have rapidly expanded, focusing on high-paid talent from the start. A Meta spokesperson has argued against claims of quality issues with Scale AI's data. In response to inquiries about Meta's reliance on rival data providers, a Scale AI representative referred back to the original investment announcement, highlighting the expansion of their commercial relationship. The dynamics shifted further when Scale AI lost partnerships with OpenAI and Google shortly after Meta's investment, resulting in 200 layoffs in July due to market demand changes, according to CEO Jason Droege. Despite the setbacks, Scale AI has secured a $99 million contract with the U.S. Army, indicating a pivot towards government sales. Speculation has also arisen that Meta's investment aimed to secure Wang's expertise in attracting top talent in AI, but doubts linger regarding Scale AI's overall value to Meta. Reports indicate that many of the executives brought over are not integrated into the core TBD Labs team. Moreover, Meta's AI unit has experienced increased chaos since the influx of new talent, with frustrations reported regarding the bureaucracy of the organization. Meta's significant investment in Scale AI, intended to bolster its AI development, may be facing critical challenges. Following the lukewarm reception of Llama 4 in April, CEO Mark Zuckerberg's impatience with the AI team has been noted. In a bid to catch up with rivals, Zuckerberg has aggressively pursued partnerships and recruited top-tier AI researchers from various leading organizations. As part of its AI strategy, Meta is expanding its data center capabilities, including a landmark $50 billion facility in Louisiana, named Hyperion. Wang's unconventional appointment to lead the AI lab has raised eyebrows, given his non-technical background. Attempts to recruit more traditional candidates for leadership roles were reportedly unsuccessful, with many new hires from OpenAI leaving shortly after joining. The departure of long-standing members from Meta’s GenAI unit continues, with notable exits including MSL researcher Rishabh Agarwal, who cited the compelling vision presented by Meta but ultimately decided to pursue new opportunities. The pressure is mounting for Meta to stabilize its AI operations and retain essential talent as MSL pushes forward with its next-generation AI model, slated for launch by year-end.