'Big Short' investor Steve Eisman on AI: The companies are trying to manufacture a crisis

Overview

Steve Eisman (host of the Real Eisman Playbook podcast; former senior portfolio manager at Neuberger Berman) argues that much of the current “AI is about to trigger a Terminator-style crisis” rhetoric is exaggerated and not supported by evidence that AI has achieved—or is on track to achieve—artificial general intelligence (AGI).

He claims the situation is being misrepresented: companies are not primarily driven by genuine fear of AI’s capabilities. Instead, he suggests they are working to manufacture a crisis.

Core Claims and Analysis

• “Terminator” fears are not credible

  • Eisman argues there’s no evidence AI has reached AGI.
  • If AGI ever arrives, he suggests it would likely be far in the future.

• A narrative is being used to reshape the market

  • He contends the industry is “nervous” not because AI is genuinely unstoppable, but because there are no real protective barriers around key parts of their business.
  • He also claims that open models have taken significant market share.

• Manufactured urgency to enable a duopoly

  • Eisman suggests the likely end goal is regulatory capture and the creation of a dominant two-company structure (a duopoly).
  • He frames this as consistent with broader “bubble” dynamics in which AI shows signals of speculative or destabilizing conditions.

Regulatory Leverage and Concentration Risk

• The AI supply chain depends on OpenAI/Anthropic

  • Eisman argues that much of the AI ecosystem (including Nvidia, hyperscalers, and major AI labs) depends heavily on the health of Anthropic and OpenAI.
  • He calls OpenAI the weaker link and suggests this may relate to why its IPO was postponed.

• Accountability laws vs real-world enforcement

  • He disputes the need for brand-new laws to ensure agent accountability.
  • However, he acknowledges that incidents (including hacking involving agents and similar cases) suggest enforcement gaps.
  • He argues that if those incidents involved “normal actors,” they would likely lead to lawsuits or jail.

Nvidia / Hugging Face as “Ecosystem Insurance”

Eisman proposes that Nvidia’s acquisition of Hugging Face may function partly as insurance:

• By protecting Hugging Face (which could otherwise be sued following security events), the deal may indirectly help protect OpenAI and the broader ecosystem.

He also considers a more nuanced alternative:

• Companies may genuinely worry about security, and regulatory capture could be an unintended consequence.
• Even if that’s the case, he argues the outcome could still be the regulatory/market consolidation he worries about.

Challenge to the “Slow Down” Message

Eisman pushes back on calls to pause AI development by highlighting inconsistency between rhetoric and action:

• If leaders truly believe AI is too dangerous, he argues they should back up the argument by postponing major fundraising and IPO plans.
• He says he hasn’t seen that happen.

Presenters or Contributors

• Steve Eisman
  • Host of Real Eisman Playbook; former senior portfolio manager at Neuberger Berman
• David Sachs
  • Referenced as a contributor/associated figure in the discussion
• Jensen
  • Referenced via “Jensen,” context suggests Jensen Huang of Nvidia
• CEO of Hugging Face
  • Referenced (not named in the subtitles)
• Nvidia
  • Referenced as an organization (not a direct person in the subtitles)