Analyst Gary Marcus argues frontier AI labs face a WeWork-style valuation reckoning

Gary Marcus, writing on his Marcus on AI blog, argues that OpenAI and Anthropic's trillion-dollar valuations are hard to justify.

He cites falling token demand, Chinese open-weight models driving down prices, and a revised OpenAI revenue forecast cut by $20 billion. He also points to rising liability risk from poorly controlled AI agents and a U.S. appeals court copyright ruling against generative AI companies.

The post reports that OpenAI has pushed its IPO to 2027 and that Anthropic has not yet publicly released its S-1 filing. Marcus's argument is his own assessment; the post does not report a financing failure or a collapse in either company's business.