Moonshots EP 179: Meta’s 15 Billion Dollar AI Bet and The Race To ASI

Summary

A WTF episode with Peter Diamandis, Dave Blundin, and Salim Ismail covering the intensifying AI talent war and the race to ASI. The central drama: Meta acquired 49% of Scale AI (Alexander Wang, youngest MIT billionaire dropout) for $14.8B in a non-voting stake structured to dodge Hart-Scott-Rodino review and the 20% financial consolidation threshold — a deal template Dave predicts will become standard. Meta also tried to buy Ilia Sutskever’s SSI ($32B valuation on $6B raised with no product) and is now attempting to hire SSI’s CEO Daniel Gross and Nat Friedman. Sam Altman claims Zuckerberg is offering $100M no-cliff signing bonuses to poach OpenAI talent, though Dave is skeptical of the no-vesting claim. The panel explores a provocative theory: Ilia’s $32B valuation was justified because he pitched investors that he knows how to build a safe ASI that will be “the first and last ASI” — making it impossible for investors not to participate. Jeff Clune (DeepMind adviser) frames the nationalization question: any government discovering an ASI within its borders will inevitably nationalize it. Dave argues the natural dynamic is winner-take-all because AI self-improvement creates compounding advantages. Salim pushes back, arguing ASI will quickly exceed any national worldview’s limitations. Elon predicts digital super intelligence “this year or by end of next year” and proposes using Grok 3.5 to rewrite the corpus of human knowledge, deleting errors and adding missing information, then retraining on the corrected data. SoftBank’s Masayoshi Son pitches a trillion-dollar plan to replicate Shenzhen-scale manufacturing within the US. The IPO and M&A markets are described as “wide open” for AI companies as of the last month.

Key Segments

• [00:00-10:00] Elon’s super intelligence timeline, rewriting human knowledge corpus, AGI vs ASI definitions, swim-lane progress model
• [10:00-18:00] “First ASI is last ASI” thesis (Jeff Clune), nationalization inevitability, winner-take-all dynamics
• [18:00-27:00] Ilia’s $32B SSI valuation theory, Meta’s $100M signing bonuses, Llama 4 quality issues, Scale AI acquisition structure
• [27:00-35:00] 49% non-voting stake as future deal template, SoftBank’s $1T Shenzhen replication, Alexander Wang replacing Yann LeCun era

Notable Claims

• Meta’s Scale AI deal: 49% non-voting stake dodges Hart-Scott-Rodino review and 20% consolidation threshold — investors received distributions (effectively an acquisition)
• Sam Altman claims Meta offering $100M no-cliff signing bonuses to poach OpenAI talent
• Ilia Sutskever raised $6B at $32B valuation with no product — theory is he pitched “first and last ASI” thesis
• Elon: digital super intelligence “this year, or next year for sure”
• Dave: IPO and M&A markets “wide open” for AI as of June 2025 — “just like 1997-98 all over again, but much bigger”
• Multiple sources confirmed during SF tour that Llama 4 “really does suck”