Key Takeaways:
- Meta is building a new internal AI lab focused on achieving superintelligence, a form of AI that could eventually exceed human capabilities across general tasks.
- Mark Zuckerberg is personally overseeing the initiative, reorganizing teams and recruiting elite talent from OpenAI, Google, and Scale AI.
- The move signals Meta’s dissatisfaction with recent AI performance and reflects a broader pivot toward long-term, high-stakes AI development.
Meta is forming a new AI research unit aimed at building superintelligence—an ambitious form of artificial general intelligence (AGI) that, if achieved, could outperform humans in a wide range of cognitive tasks. The group is being quietly assembled under the personal direction of Mark Zuckerberg, who is reportedly dissatisfied with the company’s current trajectory in large language model development.
According to internal sources, Zuckerberg has begun personally recruiting top engineers and researchers for a new, standalone AI group, separate from the broader Meta AI division. This new team is being physically centralized at Meta’s Menlo Park headquarters, where Zuckerberg has rearranged office space to accommodate the roughly 50-person group.
The focus of the team will be the development of a next-generation model architecture—one that goes beyond Meta’s existing Llama family of models. While Meta’s Llama 3 has been well-received, insiders suggest that efforts around the more ambitious “Behemoth” model have stalled. The new group aims to reboot that effort with a sharper focus, broader autonomy, and more aggressive resource allocation.
To that end, Meta is offering compensation packages reportedly in the high seven-figure range—and in some cases, even higher—to attract elite AI talent. Several hires have already been poached from competitors like Google DeepMind, OpenAI, and most notably, Scale AI. Meta has also held discussions about investing over $10 billion into Scale AI as part of a longer-term strategic alignment. Though the investment has not been finalized, it reflects Meta’s intent to secure not only infrastructure and data resources but also executive-level talent from the company.
This superintelligence effort is notable for several reasons. First, it reflects a significant shift in Meta’s AI strategy—away from broad, incremental development and toward centralized, high-risk, high-reward innovation. It also marks a more direct role for Zuckerberg in technical execution, signaling a hands-on approach reminiscent of his earlier involvement in product design and engineering during Facebook’s formative years.
Zuckerberg’s initiative runs parallel to ongoing work by Meta’s existing AI research division, FAIR (Facebook AI Research), led by Chief AI Scientist Yann LeCun. While LeCun remains influential, the emergence of this new group introduces a parallel track within Meta—one that could either complement or eventually compete with FAIR’s research agenda. How the two units align their long-term goals remains an open question.
The timing of Meta’s move is also strategic. While much of the industry has focused on rapid iteration—pushing out updates to foundational models with minimal differentiation—Meta appears to be betting on a reset. Industry observers have noted that many frontier model efforts have plateaued in performance relative to compute cost. Meta’s new lab may represent a deeper, more experimental approach, aimed at leapfrogging current limitations rather than optimizing within them.
Despite the scale of the undertaking, the company has made no formal public announcement about the group. Sources say Meta intends to keep the operation quiet for now, possibly to avoid external pressure or scrutiny during early development phases. Internally, however, the lab is seen as a critical bet for the company’s long-term future—not unlike the metaverse initiative, but with clearer near-term commercial implications.
What’s still unclear is the governance and safety framework under which the team will operate. As competition in frontier AI intensifies, scrutiny around model alignment, data provenance, and control structures is mounting. Meta has not disclosed whether this group will follow existing internal safety protocols or adopt new ones. Given the lab’s stated goal of working toward superintelligence, questions around oversight are likely to grow.
The broader market has yet to respond decisively to Meta’s AI reset. As of now, Meta stock has remained relatively stable following news of the lab’s formation. However, investors and analysts will be watching closely for signs of tangible progress—such as hiring momentum, model releases, or partnership announcements—in the months ahead.
For now, the new team represents a calculated escalation in the AI arms race. Whether it leads to breakthroughs in general intelligence or becomes another iteration of big tech’s fragmented AI strategies remains to be seen. But what’s clear is that Zuckerberg is no longer content to play catch-up.
Learn how AI Agents can supercharge your company’s profits and productivity at TMC’s AI Agent Event in Sept 29-30, 2025 in DC.
Rich Tehrani serves as CEO of TMC and chairman of ITEXPO #TECHSUPERSHOW Feb 10-12, 2026 and is CEO of RT Advisors and is a Registered Representative (investment banker) with and offering securities through Four Points Capital Partners LLC (Four Points) (Member FINRA/SIPC). He handles capital/debt raises as well as M&A. RT Advisors is not owned by Four Points.
The above is not an endorsement or recommendation to buy/sell any security or sector mentioned. No companies mentioned above are current or past clients of RT Advisors.
The views and opinions expressed above are those of the participants. While believed to be reliable, the information has not been independently verified for accuracy. Any broad, general statements made herein are provided for context only and should not be construed as exhaustive or universally applicable.
Portions of this article may have been developed with the assistance of artificial intelligence, which may have contributed to ideation, content generation, factual review, or editing.
