Mark Zuckerberg comes before the inauguration of Donald Trump as the 47th President of the United States, Washington, DC, is inside the Capital Rotunda of the US Capital Building on January 20, 2025.
Kenny Holston | Via
Mark Zuckerberg So disappointed with Meta Standing in artificial intelligence that he is ready to spend billions of dollars to convince Scale AI CEO Alexandra Wang to join his company, people familiar with the case told CNBC.
Meta is finalizing a deal to invest in a 14 billion dollar scale AI, according to the person familiar with the case that did not name because the conditions are confidential. Bloomberg Earlier this week, it was reported that an investment could be $ 10 billion and a story at the top. Information On Tuesday, Meta said that Meta would pay close to $ 15 billion.
As the founder of one of the most prominent AI startups, Wang has created a reputation as an ambitious leader who understands the technical complications of both AI and how to build a business that according to only two former Meta AI employees, who agreed to speak on oblivion status, do not focus on research. After the lukewarm launch of the latest Lama AI model of Zuckerberg Company, Meta’s AI will be counted on Wang to better execute the ambitions.
By not achieving a direct scale, Meta seems to take a strategy similar to companies. Google And MicrosoftWhich has brought key leaders in AI from the character of startups. Meta is currently Test against Federal Trade Commission Many people are familiar with the matter, for the antitrust claims, and the company Scale AI, and many familiar with this matter.
As part of the deal, meta will take 49% stake in data-labeling and anotation startups, information InformedWhile Wang will help lead a new AI Research Lab to the social networking company and join by some of his colleagues. the new York Times Was to report earlier About the new AI lab.
Alexandra Wang, CEO of Scalia, speaks on CNBC’s squalk box outside the World Economic Forum in Davos, Switzerland, Switzerland on 23 January 2025.
Garry Miller | CNBC
Scale AI, established in 2016, has made a splash in the era of generic AI by helping prominent technical companies like OpenAI, Google and Microsoft, which they prepare data used to train the cutting-edge AI model. According to two people familiar with the case, the meta scale is one of the largest customers of AI.
Startup, a funding round about a funding round Year ago In $ 14 billion, number 28 CNBC disintegration 50 listIn the middle of 2024, the company Put signature on One of the recent largest commercial leases in San Francisco, one city building has a place of about 180,000 square feet space, which was captured Airbnb,
Scale AI has rapid roads in the defense industry, and announced a multimilian dollar deal in March Department of DefenseIn November, it collaborated with the Meta on Defense Lama, a custom version of the Meta’s Open-SUS LAMA Foundation Model, especially designed to “support the American National Security Mission”, the company said. blog post,
Meta and Scale AI refused to comment.
AI challenges of meta
Title in 2025AI was one of the top priorities of Meta. But Zuckerberg has been excited that rivals such as openiI appear in both the built-in AI model and consumer-supporting apps, the current and former meta employees said.
Zuckerberg is already already already separating the CNBC to help the Meta make headways in AI, in favor of its fundamental artificial intelligence research unit, or fair, in favor of its more product-oriented Jeanai team, and AI model, CNBC’s already separating CNBC to help CNBC’s improvement in its LLAMA family. Informed,
Meta release Lama 4AI Model In April, the developers were not well received, further disappointment Zuckerberg, people said. At that time, Meta released only two small versions of Lama 4 and stated that it would eventually release a large and more powerful “Bhamoth” model.
People said that this model has so far been made available due to concerns about its abilities relative to Zuckerberg’s competition model. In particular, there is a concern about how Bhamath stack against the latest from companies such as Openi and China’s Deepsac, whose models are widely preferred by the wide developer community.
After LLAMA 4 deficiency, Meta restructured its Genai unit, dividing it into two. Connor Hayes, a long-time meta employee, was laid as in charge of AI products, while the AGI foundation was given to Amir Frenkel, which was earlier given to Meta’s reality labs hardware units, vice-president of engineering and product and previous chief of Jenai Ahmed Al-Dahle.
Al-Dahle’s new situation as a co-leader was seen as a sign that Zuckerberg lost confidence in him, people said.
Ahmed Al-Dahle, VP and head of Jenai in Meta.
Courtesy: Meta
Zuckerberg praises Wang and considers him to be capable of a major role in Meta, who are as AI leaders, people said. A dropout from the Massachusetts Institute of Technology, Wang, has created a big business and is familiar with AI’s technical complexities. People described Wang as a “War -time CEO”, which corresponds to Zuckerberg’s position that the US faces increasing competition from China, thus requiring help from the technical industry.
Wang told CNBC In January, he believes that there is a “AI war” between the US and China, and the US will need more computing power to compete.
Wang said at that time, “The United States needs a large amount of computational capacity, large amounts of infrastructure.” “We need to get energy to enable this AI boom.”
This is an unusual step for Zuckerberg, which has traditionally placed loyalists in high-ranked positions. But this reflects the magnitude of the moment and reflects Zuckerberg’s confidence that a major outsider like Wang could be better than any existing meta employee to increase the position of the company in AI, people said.
Wang also brings many external knowledge of how contestants like OpeniI are producing their consumer chatbott and AI models. In recent years, data labeling and training have become more complex as the AI model has increased capabilities, one of the competitors of Scale AI, said Petrosian, CEO of Supernotete.
Petrosian said, “I would say that Scale has probably covered 70% of all models.” Along with Wang and Scale AI others, Meta “can get collective intelligence information on how to build a better chat.”
“When Meta is buying them, they are buying their intelligence,” said Petrosian.
Watch, Mark Zuckerberg Lobs Trump to avoid meta antitrust trial