OpenAI's early backer and prominent Silicon Valley venture capitalist Vinod Khosla has long been one of the most optimistic investors in the robotics sector, predicting the field will experience its "ChatGPT moment" within the next two years. At the same time, he cautions that excessive investment and inflated valuations surrounding the hottest robotics startups signal that a large-scale industry shakeout is on the horizon.

In a recent interview, he stated, "I believe that by 2030, more than half of these companies will see their valuations fall below current levels." He was quick to add, however, that the companies that avoid valuation declines will command exceptionally high multiples.

Admittedly, it is normal for most venture capital investments to ultimately fail, but Khosla's remarks carry particular weight given that China has issued similar warnings about its own robotics industry, where it holds a globally leading position. Last year, China's economic planning agency warned that more than 150 companies were developing humanoid robots, a staggering number that fueled concerns of a potential bubble in the sector. According to The Information, leading humanoid robot company Unitree had a disappointing listing on the Shanghai Stock Exchange last month, which prompted Chinese regulators to tighten approval for humanoid robot startups seeking to go public.

Khosla's venture fund, Khosla Ventures, has invested in several robotics startups, including Rhoda AI, which focuses on training foundation models for robots, and Aim, which enables autonomous operation of earthmoving machinery.

Khosla remains convinced that the robotics industry's "ChatGPT moment" — the critical turning point he foresees — will give rise to several highly competitive humanoid robot leaders. He predicts that by then, humanoid robots with fine manipulation capabilities will be able to learn a new industrial task with just one or two hours of data.

A key question remains: whose AI model will ultimately serve as the "brain" for these robots? Until recently, "physical AI models" developed by startups such as Physical Intelligence and Skild were considered frontrunners in this race, but AI companies that train language models have recently made significant progress in robot control. For instance, in recent tests, OpenAI's GPT-6 Astra outperformed certain specialized robot AI models when controlling a robotic arm. Some even had a robot equipped with the Astra system pick up a sketchpad and brush, and the model successfully drew a scene of the Golden Gate Bridge.

If any language model company can bring about a "ChatGPT moment" for robotics, Khosla's bet is on OpenAI — hardly surprising given that his investment firm was one of OpenAI's earliest backers. "If I were to rank them in order, I would pick OpenAI first, then perhaps DeepMind, and then maybe Anthropic," he said. "Meta is also doing some specialized work on its home robotics project."

On Khosla's list, Anthropic ranks lower, for this reason: "In robotics, Anthropic has only just gotten started. They entered this field too late. Because of that, I don't think they are likely to become a leader in this transformation."

Leading AI companies are attempting to slow the pace of model development, which could ultimately decelerate progress in robotics technology. Such proposals have recently gained favor among these companies as a means of buying time to advance AI safety efforts; however, current discussions mainly focus on the digital capabilities of AI models. Khosla noted that if AI models gain the ability to drive military robots, robotics technology could also be brought into future discussions about the pace of development. At that point, "national security will become a key consideration." Once robotics experiences a breakthrough moment akin to ChatGPT, "I think this will become a central issue."

The robotics sector is not the only industry with absurdly high valuations. "There are many companies I was interested in investing in but ultimately passed on because of valuation concerns," Khosla said. "This situation is quite common now. You know, this momentum of raising a round every single month is simply crazy."

Khosla said he keeps a list of companies he turned away due to excessive valuations despite admiring their business models. On the other hand, Khosla pointed out that some valuations may appear high at first glance, but given the exceptional quality of the business, such "high valuations are well worth it." For example, Khosla invested in AI coding startup Cognition three years ago. At the time, the company's valuation seemed quite high, but its subsequent growth has proven that investment to be "very worthwhile."