Former OpenAI Researcher Questions Frontiers AI Lab's Trillion-Dollar Valuation: The More Cutting-Edge, the More Cash Burn

According to Beatling monitoring, former OpenAI researcher Andrew Ho questioned the trillion-dollar valuation of the cutting-edge AI lab.

Ho's back-of-the-envelope calculation suggests that even with an inference business gross margin of 80% and a generous 20x P/E ratio, a company would need to achieve annual revenues of $100 billion to $200 billion to potentially support a $1 trillion valuation. This calculation also assumes that the company no longer spends money on training new models.

In reality, companies like OpenAI and Anthropic must continue to train the next generation of models. Once they stop, users may shift to cheaper alternatives like Qwen or Kimi. As revenues rise, so do the training costs for the next generation of models. The frontrunners find themselves needing to continually increase their investments.

Ho also does not believe that AI capabilities will see a rapid full-scale breakthrough. He thinks that large models are still biased and progress may be slower, further constrained by high-quality data. Even if model capabilities plateau at today's levels, the real integration of AI across industries may still take over 20 years.

He believes that programming agents may be just a few high-value products that have been discovered. Deploying more engineers to help companies adopt AI may not necessarily lead to quickly discovering the next profitable scenario.

Ho remains bullish on the long-term growth of the AI industry but is not willing to buy into OpenAI or Anthropic at their current valuations.