Former OpenAI Employee Urges Colleagues to Cash Out Before IPO

Andrew Ho, who joined OpenAI's technical team in late 2025 and left after about eight months, has been vocal on social media, telling former colleagues to participate in any tender offers and cash out immediately instead of waiting for an IPO. He believes the valuation logic differs fundamentally between private and public markets: private markets allow for storytelling and optimistic expectations, while public markets demand cash flow, profits, and real growth. He also notes practical issues like lock-up periods and potential selling pressure after the lock-up expires. Ho himself owns about $700,000 in OpenAI equity but missed the recent tender offer in October, when over 600 employees sold shares in a record deal. Despite this, he remains bearish on the valuation of frontier AI labs, arguing that they have high costs for training next-generation models, uncertain revenue, and are locked in a Red Queen race where they must continuously invest to stay competitive. He calculates that to support a $1 trillion market cap, a lab would need annual revenue of $100-200 billion, which he deems unrealistic. He also suggests that model improvements are mostly in verifiable tasks like math and code, while complex real-world tasks progress slowly. Nevertheless, he remains optimistic about AI's future but believes the next bottleneck will be high-quality training data, not just compute, and he has founded a startup to provide reinforcement learning datasets to frontier labs.