Mark Zuckerberg Reportedly Building a ‘Superintelligence’ AI Team at Meta

TL;DR

• Meta CEO Mark Zuckerberg is assembling a team to build superintelligence, AI that surpasses human capabilities.
  
• The effort stems from frustrations over the Llama 4 model’s progress, despite Meta’s existing AI tools.
  
• Zuckerberg aims to personally oversee the initiative, tapping experts and shifting company structure.
  
• Meta plans to fund the effort through its ad business, while exploring partnerships like one with Scale AI.
  
• The AI race includes major players like OpenAI, Google, xAI, and now a more aggressive Meta.
  

Zuckerberg’s Superintelligence Mission Accelerates

Meta’s CEO Mark Zuckerberg is making an aggressive new move in the AI race by forming a dedicated team to build “superintelligence,” a form of artificial intelligence that could exceed human cognitive ability. According to a report from Bloomberg, the initiative is being spearheaded directly by Zuckerberg, who has grown increasingly frustrated with the internal pace of Meta’s AI efforts—especially surrounding the underperformance of its Llama 4 large language model.

The revelation highlights how seriously Meta views the intensifying AI competition, particularly with OpenAI, the creator of ChatGPT, continuing to lead the space with significant investments and user adoption.

Meetings at Home, Hiring on the Horizon

Rather than relying solely on Meta’s existing AI infrastructure, Zuckerberg has begun hosting private meetings with AI experts at his homes in Lake Tahoe and Palo Alto, according to sources cited by Bloomberg. The goal is to build a team of around 50 engineers and researchers who can push Meta’s capabilities far beyond what Llama has achieved so far.

Per the New York Times, the new team may also include Alexandr Wang, CEO of Scale AI, a well-funded AI infrastructure startup. Meta is reportedly considering billions in investments in Wang’s company to support the superintelligence project.

Data Callout Box

Why Meta Is Ramping Up Its AI Bet

Despite its dominance in social networking, Meta has struggled to differentiate its AI strategy in a crowded and fast-moving field. Tools like Ray-Ban smart glasses and AI chatbots across WhatsApp and Instagram have been public-facing, but Meta’s open-source Llama models have not yet achieved the level of sophistication or adoption seen with OpenAI’s GPT models.

Meta’s long-standing strategy has been to push open-source AI, making Llama accessible to the broader developer community. The intention is to become a foundational layer in global AI development—similar to what Android did for mobile devices.

However, internal frustrations with Llama 4’s capabilities, along with major advances from rivals like xAI (Elon Musk), Anthropic, and Google DeepMind, are believed to have triggered Zuckerberg’s new approach.

Competitive Pressure from Tech Giants

The broader AI landscape is in a state of hyper-competition:

• OpenAI, backed by Microsoft, continues to lead with ChatGPT, now integrated across Office products and Azure.
  
• Google is aggressively enhancing its Gemini models, designed to challenge OpenAI across search, productivity, and cloud.
  
• xAI, led by Musk, is pushing the Grok chatbot while integrating tightly with the X platform.
  
• Anthropic and other new entrants are drawing major investments and academic partnerships.
  

Meta, in contrast, has seen mixed results. While its open-source approach appeals to developers, its models lag in performance benchmarks.

Superintelligence vs. AGI: What’s the Difference?

Zuckerberg’s stated goal of building “superintelligence” suggests ambitions that go beyond AGI (Artificial General Intelligence). AGI refers to AI that can perform any intellectual task a human can. Superintelligence, however, implies surpassing human intelligence in nearly all domains, from creativity and logic to social skills and innovation.

AI researchers remain divided on the timeline to AGI—some say it’s achievable in this decade; others argue it’s still speculative and lacks a defined path. Building superintelligence adds another layer of complexity and uncertainty.

Nonetheless, companies like Meta are increasingly treating it as inevitable—or at least essential to pursue before rivals dominate.

Office Layout and Funding Strategy

According to Bloomberg, Zuckerberg has also physically reorganized Meta’s Menlo Park headquarters, relocating the superintelligence team closer to his office. The intent is to monitor progress closely and eliminate bureaucratic slowdowns.

Funding for the team is expected to come from Meta’s advertising business, which brought in over $130 billion in revenue last year. This internal funding model gives Meta flexibility and independence compared to peers that rely on venture capital or cloud contracts.

Implications for Meta’s Future

Meta’s pivot toward superintelligence could dramatically reshape its product lineup, internal culture, and competitive positioning. While the company has already been integrating AI into user-facing platforms, the leap to superintelligence would demand new infrastructure, ethical guardrails, and long-term planning—something not easily achieved even by trillion-dollar firms.

It also raises questions about how this effort will interface with existing AI teams. Will it be a parallel research lab like Google’s DeepMind? Or will it eventually absorb and reshape Meta’s current AI operations?

Final Thoughts

Meta’s superintelligence initiative is bold, risky, and emblematic of an AI arms race that shows no signs of slowing. With major tech CEOs now personally involving themselves in model training, team recruiting, and even AI ethics, the stakes are as high as the ambitions.

If Zuckerberg’s initiative succeeds, Meta could leap from an open-source also-ran to a central AI powerhouse. If it fails, it risks further lagging behind rivals who are moving faster and smarter in the race toward the future of intelligence.