New York University (NYU) professor emeritus Gary Marcus warned that OpenAI could become the WeWork of AI. He tied the comparison to a delayed initial public offering for a company last valued at $852 billion.
The warning matters for Nvidia (NVDA) shareholders and AI infrastructure lenders. A markdown at OpenAI could spread through chip demand and financing costs. Marcus said sentiment, not fundamentals, would drive the damage.
Does a Delayed IPO Make OpenAI the ‘WeWork of AI’?
OpenAI filed confidentially for a listing in June, then CEO Sam Altman ruled out 2026. Its last private valuation came in a $122 billion round in March.
Reports of a new private round cite valuations from $1.2 trillion to $1.4 trillion. Onchain markets have pushed the Anthropic pre-IPO valuation above $1 trillion, again without a public price to test it.
WeWork is the precedent Marcus invoked. Its 2019 filing showed almost $700 million in first-half losses. A $47 billion private mark then fell toward $10 billion within weeks.
Marcus noted that WeWork never completed its IPO. It did list through a special purpose acquisition company (SPAC) merger in 2021, then filed for bankruptcy in 2023.
Who Sets the Price if OpenAI Waits?
Anthropic is lining up a November listing ahead of OpenAI. Marcus said it still wants about $1.5 trillion, though he cited reports nearer $1.1 trillion. That compares with the $2 trillion target circulating a month earlier.
Reuters reported OpenAI’s revenue reached $13 billion in 2025. Marcus has put its operating losses near $2 billion a month.
Neo clouds, newer firms that rent out AI computing capacity, face the sharpest exposure, Marcus said. Their financing is getting costlier.
Anthropic’s debut could set the yardstick for roughly $600 billion of projected OpenAI hardware spending through 2030. The same yardstick could apply to the chipmakers and lenders behind it. A strong Anthropic debut would cut against the thesis.
