Michael Burry, the hedge fund manager who became a household name by betting against the US housing market before the 2008 financial crisis, has a new target: the AI industry’s safety narrative.
On September 14, Burry posted on X that recent calls from leaders at OpenAI and Anthropic to slow down AI development are “self-serving,” designed to benefit incumbents, generate IPO hype, and paper over genuine growth problems. His core argument is blunt: large language models are not true artificial intelligence and will never become artificial general intelligence, so there’s nothing dangerous enough to justify pumping the brakes.
The contrarian’s case against AI safety theater
Burry’s skepticism lands at a particularly charged moment. Anthropic CEO Dario Amodei recently published an essay advocating for measured advancement of AI capabilities, a position that drew public support from OpenAI CEO Sam Altman and Elon Musk. Several high-profile AI researchers have also resigned from major labs in recent months, citing concerns about existential risks.
The market apparently took the safety warnings seriously, though. AI-related stocks declined on September 14, with the sell-off rippling through chipmakers and adjacent sectors.
Two companies, two very different IPO strategies
OpenAI has confirmed it will not pursue an IPO in 2026, with Altman calling such a move “ill-advised” given ongoing work on safety and alignment.
Anthropic, meanwhile, is reportedly eyeing a Nasdaq listing as early as October 2026, with a target valuation approaching $2 trillion and hopes of raising nearly $100 billion.
Burry’s read is that these two approaches represent different flavors of the same game. OpenAI uses safety concerns to justify staying private and avoiding valuation pressure. Anthropic uses safety credibility to justify an astronomical public valuation. In both cases, the “we must be careful because this technology is so powerful” framing serves corporate interests rather than public ones.
The LLM-is-not-AGI argument
The philosophical core of Burry’s position, that LLMs are not AI and will never become AGI, is not particularly controversial among computer scientists but remains deeply contested in the broader tech discourse.
LLMs are, at their foundation, statistical prediction engines. They generate text by predicting the most likely next token based on patterns in their training data. What’s clear is that these systems lack the kind of general reasoning, goal-setting, and world-modeling capabilities that most AGI definitions require.
What investors should actually watch
Anthropic’s potential $2 trillion valuation will be an important test case. That number implies a level of future revenue generation that would require AI adoption to accelerate dramatically across nearly every industry. If the safety-conscious approach actually slows deployment, as the rhetoric suggests it should, that valuation becomes very difficult to justify on fundamentals.