- Meta is forming a “superintelligence”/AGI team: Zuckerberg is assembling a group of roughly 50 AI experts, including a new head of AI research, located near him at Meta’s Menlo Park HQ. This move stems from growing frustration with the performance of Meta’s Llama 4 and delays in its upcoming upgrade, codenamed “Behemoth”.
- Major investment in Scale AI: Meta is reportedly acquiring a 49% stake in Scale AI, a leading data-labeling company, for approximately $14.8–15 billion. Scale’s CEO, Alexandr Wang, is expected to join the new AGI team once the deal closes.
- Building a “lab” for superintelligence: The initiative is being portrayed as Meta’s most audacious AI effort yet—a dedicated lab for building systems that surpass human cognitive abilities. Zuckerberg is offering lucrative compensation (even “seven to nine‑figure” packages) to attract top-tier researchers.
- Strategic capex boost for AGI: The move is part of Meta’s broader plan to ramp up capital expenditures, boosting its AI infrastructure budget to about $64–72 billion in 2025—aimed at scaling up computing capacity and fueling its AGI ambitions.
🔍 Why It Matters
- This aggressive strategy signals Meta’s intent to directly challenge the AGI leadership of OpenAI, Google, Microsoft, and others.
- By partnering deeply with Scale AI, Meta gains a powerful edge in training-data quality—crucial for building more capable models.
- Zuckerberg’s hands-on recruitment suggests the initiative is a top-level priority, with real urgency behind it.
TL;DR
| Element | What’s Happening |
|---|---|
| Team | ~50 AI experts joining a “superintelligence” unit. |
| Scale AI deal | Meta to buy ~49% for ~$15B; Scale’s CEO to join AGI team. |
| Funding | $64–72B capex planned; heavily investing in infrastructure and talent. |
Today’s Zuckerberg‑AI superintelligence news
Here are more detailed insights about Meta’s AI superintelligence initiative led by Mark Zuckerberg:
- Strategic Talent Shift: Zuckerberg is pulling top AI researchers from Meta’s FAIR (Fundamental AI Research) group into a new elite team focused solely on achieving artificial general intelligence (AGI). This reflects a shift from academic-style research to practical AGI development.
- AGI Lab Structure: The team is being physically centralized near Zuckerberg at Meta HQ, with strong coordination and fast iteration. This contrasts with previously decentralized AI efforts across Meta’s various divisions (e.g., Facebook, Instagram, WhatsApp).
- Scale AI Partnership: Meta’s huge investment in Scale AI aims to improve its access to massive, high-quality datasets critical for training AGI models. Scale is known for working with the U.S. government and OpenAI, which gives Meta competitive leverage.
- Behemoth and Llama 5: Internally, Meta is betting on its next model—code-named Behemoth (likely to be Llama 5)—to rival or surpass GPT-5 and Gemini. But delays and performance issues with Llama 4 have triggered this more aggressive push.
- Compute Scale-Up: Meta is expected to dramatically expand its use of NVIDIA GPUs, including the H100s and upcoming Blackwell chips, to support this AI ramp-up. Datacenters in the U.S. are being optimized for large-scale model training.
- Meta AI’s Broader Vision: Zuckerberg wants Meta to be a dominant player in consumer-facing AI. This includes integrating AGI into Meta’s AR/VR platforms, smart glasses, and assistants like Meta AI across WhatsApp and Instagram.
About the Author
