Meta’s Scale AI Partnership Shows Signs of Strain
▼ Summary
– Meta invested $14.3 billion in Scale AI in June, bringing CEO Alexandr Wang and executives to run Meta Superintelligence Labs, but the relationship is already showing signs of strain.
– At least one executive from Scale AI, Ruben Mayer, departed Meta after just two months, with conflicting accounts about his role and reporting structure.
– Meta’s TBD Labs is working with Scale AI’s competitors, Mercor and Surge, as some researchers reportedly view Scale AI’s data as low quality despite the investment.
– Scale AI faced setbacks after Meta’s investment, including losing OpenAI and Google as customers and laying off 200 employees in July, while shifting focus to government contracts.
– Meta’s AI unit has experienced internal chaos, talent departures, and bureaucratic frustrations, raising questions about the stability and success of its AI strategy and investment.
Meta’s recent $14.3 billion investment in Scale AI, intended to supercharge its artificial intelligence ambitions, is already showing signs of turbulence just months after the high-profile partnership began. The collaboration, which brought Scale AI CEO Alexandr Wang and several top executives to lead the newly formed Meta Superintelligence Labs (MSL), appears to be facing internal friction and strategic shifts.
One of the key executives who moved from Scale AI to Meta, former Senior Vice President of GenAI Product and Operations Ruben Mayer, left the company after only two months. Sources indicate Mayer oversaw AI data operations during his brief tenure, though he later clarified that his role was focused on helping establish the lab and that he was part of the core TBD Labs unit from the beginning. He also stated he did not report directly to Wang and described his experience at Meta positively.
Beyond personnel changes, Meta is broadening its data labeling partnerships beyond Scale AI. The company is now working with competitors like Mercor and Surge to train its next-generation AI models. While it’s common for AI labs to use multiple vendors, Meta’s significant investment in Scale AI makes this move particularly noteworthy. Several sources familiar with the matter say researchers within TBD Labs have expressed concerns about the quality of Scale AI’s data and prefer working with rival providers.
Scale AI originally built its reputation using a crowdsourced, low-cost workforce for basic data labeling tasks. However, as AI models grow more advanced, the demand has shifted toward highly specialized experts in fields like medicine, law, and science. Although Scale AI has launched its Outlier platform to attract such talent, competitors like Surge and Mercor entered the market with a focus on high-quality, expert-driven data from the start.
A Meta spokesperson denied there are quality issues with Scale AI’s offerings. Scale AI declined to comment on Meta’s use of other vendors, instead pointing to the initial investment announcement that highlighted an expanded commercial relationship between the companies.
The situation is more critical for Scale AI, which lost major clients like OpenAI and Google shortly after announcing the Meta deal. In July, the company laid off 200 employees in its data labeling division, citing “shifts in market demand.” Scale AI is now pivoting toward government contracts, recently securing a $99 million deal with the U.S. Army.
Some industry observers believe Meta’s investment was primarily a talent acquisition strategy aimed at bringing Alexandr Wang onboard. Wang, though not a researcher himself, has significant experience in the AI sector and is helping Meta attract top talent. However, questions remain about how integral Scale AI’s technology and personnel are to Meta’s AI goals. Several executives who joined from Scale AI are not working on the core TBD Labs team, and the broader AI unit has experienced internal chaos since the reorganization.
Meta’s AI efforts have been under pressure since the disappointing launch of Llama 4 in April, which frustrated CEO Mark Zuckerberg and prompted a aggressive hiring and acquisition spree. The company has recruited researchers from OpenAI, Google DeepMind, and Anthropic, and acquired AI voice startups like Play AI and WaveForms AI. It also announced a partnership with Midjourney and plans for massive new data centers, including a $50 billion facility in Louisiana named Hyperion.
Still, retention remains a challenge. Several high-profile hires from OpenAI have already left, and longtime members of Meta’s original GenAI team have departed amid the restructuring. Recent exits include AI researcher Rishabh Agarwal, who cited the compelling vision of the Superintelligence team but decided to pursue new risks, echoing Zuckerberg’s own advice. Product lead Chaya Nayak and research engineer Rohan Varma have also announced their departures.
As MSL pushes forward with development of its next AI model, aiming for a late 2024 release, the central question is whether Meta can stabilize its AI division, retain essential talent, and effectively leverage its partnership with Scale AI, or if the multibillion-dollar bet will continue to face headwinds.
(Source: TechCrunch)
