News

Meta Assembles AI Powerhouse Amid Fierce Talent War

• By John K. Waters
• 07/08/2025

Meta Platforms has formally launched Meta Superintelligence Labs (MSL), a new division intended to centralize and accelerate its artificial intelligence development efforts, as it moves to compete more directly with top-tier AI labs including OpenAI, Google DeepMind, and Anthropic.

Announced by CEO Mark Zuckerberg in an internal memo on June 30, MSL consolidates all of Meta's foundational AI efforts, including large language models (LLMs), applied research, and the Fundamental AI Research (FAIR) unit. It also introduces a new lab focused on developing frontier models, Bloomberg and CNBC reported.

Leadership Overhaul and Strategic Hires
The new division is led by Alexandr Wang, the former CEO of Scale AI, who joins Meta as its chief AI officer. Zuckerberg described Wang as "the most impressive founder of his generation" and emphasized his long-standing relationship with the data-labeling entrepreneur. The appointment followed Meta's $14.3 billion investment in Scale AI, which also brought over several of Wang's former colleagues.

Wang will co-lead the initiative with Nat Friedman, the former GitHub CEO and co-founder of VC firm NFDG. Friedman will head applied research and product development within MSL. According to CNBC, Meta had previously attempted to acquire Safe Superintelligence, a startup founded by OpenAI co-founder Ilya Sutskever and backed by Friedman and Daniel Gross, but the deal fell through.

Intensified Recruitment Push
Zuckerberg's aggressive recruitment campaign, described as a personal effort involving in-home meetings and high-dollar compensation offers, has added fuel to the competitive race for top AI talent. OpenAI CEO Sam Altman claimed in June that Meta offered his researchers signing bonuses as high as $100 million in an effort to lure them away, Reuters reported.

New additions to Meta's AI roster include researchers from OpenAI, DeepMind, Anthropic, and Google, many of whom led development on major AI models such as GPT-4o, Gemini, and Llama. Notable hires include:

Trapit Bansal, co-creator of OpenAI's O-series models

Shuchao Bi, lead for GPT-4o's voice mode

Jack Rae, Gemini 2.5 reasoning tech lead at DeepMind

Pei Sun, responsible for post-training and coding at Gemini

Huiwen Chang, previously at Google Research

Joel Pobar, Anthropic and ex-Meta engineer

Hongyu Ren, co-creator of GPT-4o

Jiahui Yu, who led perception work at OpenAI and Gemini

Shengjia Zhao, a key contributor to ChatGPT and GPT-4

Johan Schalkwyk, former Google Fellow

Ji Lin, involved in several GPT-4 variants

Infrastructure, Model Strategy, and Market Positioning
Meta's reorganization follows criticism over the lukewarm reception of its Llama 4 model and a string of senior staff exits, which reportedly stalled momentum while competitors advanced. Zuckerberg acknowledged these challenges and emphasized the company's intent to develop "personal superintelligence for everyone," according to the memo.

The company has committed to spending "hundreds of billions" of dollars on AI infrastructure, including compute resources and data centers, over the next several years. Zuckerberg cautioned that many firms could be "overbuilding" AI capacity, but said that falling behind would be a more costly error, given the strategic significance of the technology.

Meta's AI assistant, powered by Llama models, currently reaches more than 1 billion users across its apps, and the company is preparing for iterative updates to the models (Llama 4.1 and 4.2), according to the internal memo. In parallel, MSL is beginning work on a new generation of models that aim to set the industry benchmark within a year.

Broader Implications
The launch of MSL marks Meta's most definitive step yet in its pursuit of AI leadership, amid escalating competition and consolidation in the sector. With deep financial backing, consumer reach, and recent success in wearables such as smart glasses, Meta is positioning itself as a major player not only in model development but in AI deployment at scale.

Meta shares reached a record high of $747.90 on June 30 but were little changed by the close of trading that day.