Since June, Meta has invested $ 14.3 billion on the AI data suppliers scale, bringing CEO Alexandr Wang and many of the top start -up executives for conducting Meta Superintelligence Labs (MSL) workshops. But the relationship between the two companies already presents signs of shrinkage.
At least one of the Wang executives brought to help MSL run – former senior vice president of Genai Product and AI’s businesses Ruben Mayer – has departed after just two months with the company, two people familiar with the issue told TechCrunch.
Mayer spent about five years with AI scale at two stops. In a short time at Meta, according to these sources, Mayer supervised the AI Data Function Groups and reported to Wang but had not come to the company’s TBD Labs – the main unit tasked with the building of Superintelligence, where AI’s top researchers were landed.
However, Mayer disputes some details about his role, telling TechCrunch that his original position was “to help create the lab, with whatever it needed” instead of running data functions and that it was “part of the TBD workshops from day one” instead of excluding the Core AI unit. Mayer also clarified that “did not report directly [Wang]”And he was” very happy “with Meta’s experience.
In addition to staff changes, Meta’s relationship to the AI scale seems to be shifting. TBD Labs works with third -party data suppliers, in addition to the AI scale to train upcoming AI models, according to five people familiar with the subject. These third -party suppliers include Mercor and Surge, two of the largest competitors on the AI scale, people said.
While AI Labs usually works with several data suppliers – Meta works with Mercor and Surge before rotating TBD Labs – it is rare for an AI lab to invest so much in a data supplier. This makes this situation particularly remarkable: even with the investments of many billions of dollars of META, several sources have reported that researchers in TBD laboratories see AI scale data as low quality and have expressed the preference to work with Surge and Mercor.
Sale AI initially created its business in a crowdsourcing model that used a great low cost workforce to handle simple data commentary tasks. However, as AI models have increased more sophisticated, they are now demanding high-quality experts-such as doctors, lawyers and scientists-to create and improve the high quality data needed to improve their performance.
TechCrunch event
Francisco
|
27-29 October 2025
Although Scale AI has moved to attract these experts to its platform, competitors such as Surge and Mercor are rapidly increasing because their business models were built on the foundation of high paid talents from the beginning.
A Meta spokesman has questioned the fact that there are quality problems with AI’s AI product. Surge and Mercor refused to comment. Asked about the deepening of META by competing data providers, a AI scale spokesman addressed TechCrunch to his initial announcement Meta’s investment in the start, which mentions the expansion of the companies’ trade relationship.
Meta’s agreements with third -party data suppliers may mean that the company does not put all its eggs on a AI scale, even after the starting of billions. However, the same cannot be said about the AI scale. Shortly after the announcement of its mass investment on a AI, Openai and Google scale, they said they would stop working with the data provider.
Shortly after the loss of these customers, AI’s escalation was fired 200 employees in highlighting his data Businesses in July, with the new CEO of the company, Jason Droege, accusing the changes in part in the “market demand shifts”. Droege said the AI scale would personal to other parts of the business, including government sales – the company has just landed A $ 99 million contract with the US Army.
Some initially believe that Meta’s investment on AI has really been to entice Wang, a founder who worked in the AI area since Scale AI was founded in 2016 and seems to help META attract top AI talents.
In addition to Wang, there is an open question about how valuable scale the meta is.
A current MSL employee says many of the scale executives transferred to Meta do not work on the Core TBD Labs team.
Meanwhile, META’s AI unit has become more and more chaotic since it brought Wang and a wave of top researchers, according to two former employees and one MSL today. The new talent from Openai and Scale AI expressed their frustration for navigating the bureaucracy of a large company, while Meta’s previous Genai team has seen Limited the scope, they said.
Tensions show that META’s largest AI investment to date can launch a rocky start, despite the fact that it had to face the company’s growth challenges. Following Llama 4 in April, Meta Mark Zuckerberg’s chief executive was frustrated by the company AI team, a today’s and a former TechCrunch employee.
In an attempt to turn things around and cover Openai and Google, Zuckerberg rushed to hit the deals and launched an aggressive campaign to hire AI’s top talent.
In addition to Wang, Zuckerberg managed to pull AI’s top researchers from Openai, Google Deepmind and Anthropic. Meta has also acquired AI Voice starts, including Play AI and AI waveforms, and announced a collaboration with the launch of Generation Ai Image, Midjourney.
To feed AI’s ambitions, Meta has recently announced several massive data centers in all the US, one of the largest is one Data Center of 50 billion dollars In Louisiana he called Hyperion, who was named after a Titan to Greek mythology who acquired the god of the sun.
Wang, who is not a AI researcher from the background, was regarded as a somewhat unusual choice to drive an AI lab. Zuckerberg reportedly held talks to bring more traditional candidates to lead the effort, such as Openai’s lead researcher, Chen NoteAnd he tried to get the newly formed businesses of Ilya Sutskever and Mira Murati. All of them refused.
Some of the new AI researchers recently brought from Openai has already left metaWired previously reported. Meanwhile, many long -term members of Meta’s Genai unit have departed in the light of changes.
MSL AI RISHABH AGARWAL researcher is from the last, Post in x This week he will leave the company.
“The stadium by Mark and @alexandr_wang to build the Superintelligence team was incredibly imperative,” Agarwal said. “But in the end I choose to follow Mark’s advice:” In a world that changes so quickly, the greatest risk you can get is not taking any risk. “
He was later asked about his time at Meta and what led his decision to leave, Agarwal refused to comment.
Product Management Manager for Genetics AI, Chaya nayakand research engineer, Rohan VarmaThey have also announced their departure from Meta in recent weeks. The question now is whether Meta can stabilize the functions of AI and maintain the talent it needs for its future success.
MSL has already started working on the next generation AI. According to reports from TrustingIt aims to start it by the end of this year.
Update: This story has been informed by comments by Mayer, who arrived at TechCrunch after the publication.
