Meta Bets Big on Scale to Boost AI Supremacy
Meta is pouring $14.3 billion into artificial intelligence company Scale AI, while also bringing Scale’s CEO Alexandr Wang into its team dedicated to building “superintelligence”. The move, announced Thursday, underscores Meta CEO Mark Zuckerberg’s renewed commitment to advancing AI at the parent company of Facebook and Instagram, as it faces escalating competition from Google, OpenAI, and Microsoft.
In a joint statement late Thursday, Meta and Scale described the arrangement as a “strategic partnership and investment,” valuing Scale at over $29 billion. Although Scale will maintain its independence, the deal deepens its commercial ties with Meta, which will acquire a 49% equity stake.
While Wang will transition to Meta alongside a select group of Scale employees, he will remain on Scale’s board of directors. Jason Droege, previously Scale’s chief strategy officer and a former executive at Uber Eats and Axon, steps in as interim CEO.
Zuckerberg’s Pivot from Metaverse to Superintelligence
Zuckerberg’s intensified focus on “superintelligence” — often referred to as artificial general intelligence (AGI) by industry peers — marks a significant shift from his previous metaverse-centric vision, which saw Meta rebrand and invest heavily in virtual reality technologies in 2021.
This strategy mirrors recent moves by other tech giants seeking top AI talent and capabilities without full acquisitions. Microsoft, for instance, hired much of Inflection AI’s leadership, including co-founder Mustafa Suleyman, who now leads its AI division. Google recruited leaders from Character.AI, while Amazon struck a deal with San Francisco-based Adept, acquiring key staff and licensing its AI models and datasets.
The Rise of Scale and Alexandr Wang
Wang founded Scale in 2016 while studying at the Massachusetts Institute of Technology, alongside co-founder Lucy Guo. Their fledgling startup gained early backing from Y Combinator, then led by Sam Altman, who now helms OpenAI. Emulating Zuckerberg’s own academic exit, Wang left MIT to pursue Scale full time.
Initially, Scale provided crucial data-labeling services to refine AI systems, employing workers to annotate images for applications like autonomous driving. Clients included industry heavyweights like General Motors and Toyota. The company offered a more customized alternative to Amazon’s Mechanical Turk, which historically paired freelance workers with short-term online tasks.
A New Frontier: Training Large Language Models
As large language models (LLMs) like OpenAI’s ChatGPT, Google’s Gemini, and Meta’s Llama rose to prominence, Scale evolved its services. Its annotation teams now help fine-tune training datasets and evaluate model performance for nearly every major LLM developer, including Anthropic, OpenAI, Meta, and Microsoft. However, the implications of Meta’s investment for Scale’s existing client base remain uncertain.
Bridging Silicon Valley and Washington
Wang has also cultivated ties with the U.S. government, securing Pentagon contracts and attending President Donald Trump’s 2017 inauguration. Notably, Michael Kratsios, Trump’s former chief technology officer, served as an executive at Scale between Trump’s presidential terms. Meanwhile, Meta itself has expanded its AI services to federal agencies.
Meta’s Contrarian AI Approach
Unlike many competitors, Meta has opted to open-source its flagship Llama models, granting public access to key components for modification and use. Though Meta claims its AI products reach over a billion users monthly, it still trails behind OpenAI and Google in promoting consumer applications of LLMs.
Despite previewing its most advanced model, Llama 4 Behemoth, in April — touting it as “one of the smartest LLMs in the world and our most powerful yet” — Meta has yet to publicly release it.
Skepticism and Vision Inside Meta’s AI Division
Yann LeCun, Meta’s chief AI scientist and 2019 recipient of the Turing Award, has voiced doubts about the prevailing emphasis on LLMs.
"How do we build AI systems that understand the physical world, that have persistent memory, that can reason and can plan?" — Yann LeCun, Meta
According to LeCun, these traits represent essential aspects of intelligence that current LLMs either fail to achieve or only approximate superficially. Instead, Meta remains committed to a longer-term vision of creating AI systems that not only match but surpass human intelligence. Speaking at France’s VivaTech conference on Wednesday, LeCun alluded to the Scale deal without providing specifics, stating that Meta now has “a clearer vision for how to accomplish this.”
Leadership Changes Reinforce Meta’s AI Ambitions
LeCun co-founded Meta’s AI research division over a decade ago with Rob Fergus, a former New York University professor. After a five-year stint at Google, Fergus returned to Meta last month to lead the research lab, replacing Joelle Pineau as director.
Writing on LinkedIn, Fergus reaffirmed Meta’s enduring dedication to groundbreaking AI research, describing its mission as “building human-level experiences that transform the way we interact with technology.”
