On June 1, a recruitment post shared by OpenAI CEO Sam Altman on social media officially announced the company's entry into the physical robotics track. Altman stated that the company is forming a new team named "OpenAI Robotics" and is publicly recruiting full-stack hardware, operations, systems, and machine learning engineers. The goal is to "co-program and build robots that are truly useful to society."
According to Altman's explanation, OpenAI's robotics strategy has both short-term and long-term goals. In the short term, OpenAI focuses on developing robots that can assist technical workers in building future infrastructure. In the long run, the company envisions a future where every person can own a personal robot capable of fulfilling various needs.
Altman revealed that the move into robotics is based on the rapid development of an internal OpenAI research project called "Worldsim." This project evolved over the past year into OpenAI Robotics, led by the company's Research Vice President, Aditya Ramesh, a core developer of the text-to-image model DALL·E and the video generation model Sora. The project's foundation lies in the deep integration and co-design of robotics hardware research and machine learning research.
OpenAI's return to the robotics field is, in fact, a "comeback." As early as the company's founding days, robotics technology was a crucial direction in its exploration of Artificial General Intelligence (AGI). Between 2016 and 2019, OpenAI successively launched the reinforcement learning benchmark environment OpenAI Gym, the open-source robot simulation platform Roboschool, and successfully developed a dexterous robotic hand named Dactyl.
In 2019, OpenAI, using reinforcement learning and "Automatic Domain Randomization" (ADR) technology, trained an AI system that enabled a humanoid robotic hand to successfully solve a Rubik's Cube. This research proved the feasibility of the technical path of training in a simulation environment and then transferring the capabilities to a real robot. However, due to the scarcity of robotics training data and slow iteration at that time, contrasted with the vast and easily accessible text and image data on the internet, OpenAI made a strategic decision around 2020: to disband the robotics team and concentrate resources on the development of large language models represented by the GPT series. This decision ultimately led to the creation of ChatGPT.
In the following years, OpenAI ignited the global large model boom with its ChatGPT series of products, becoming the world's highest-valued AI unicorn. According to multiple media reports, OpenAI secretly submitted a draft IPO prospectus on May 22, planning to go public as early as September 2026. In its latest round of financing completed in March of this year, OpenAI's valuation reached $852 billion. Institutions like Deutsche Bank predict its listing valuation could exceed $1 trillion, with a fundraising scale potentially reaching $60 billion, potentially making it one of the largest tech IPOs in U.S. public market history.
Nevertheless, OpenAI also faces significant pressure from substantial losses. The company is projected to incur a full-year loss of approximately $14 billion in 2026, with cash burn expected to further expand. It might not achieve cash flow break-even until 2030 at the earliest. Its gross margin is only about 33%, with the high inference costs of AI models being the primary reason for eroding profits.
During the years after disbanding its in-house robotics team, OpenAI did not completely abandon the robotics track. Instead, it adopted a "multi-point investment" strategy through its venture fund, successively investing in several robotics startups, including the Norwegian humanoid robot company 1X Technologies, the American humanoid robot star company Figure AI, and Physical Intelligence.
The most notable collaboration was with Figure AI in February 2024. At that time, OpenAI not only participated in Figure AI's $675 million Series B funding round but also announced the development of a dedicated multimodal AI model for Figure's humanoid robots. Just 13 days after the partnership was announced, the Figure 01 humanoid robot equipped with the OpenAI model demonstrated fluent natural language interaction, object recognition, and autonomous operation capabilities.
However, this collaboration lasted less than a year. In February 2025, Figure AI founder Brett Adcock officially announced the termination of the partnership with OpenAI, opting to independently develop an end-to-end robot AI model. The primary reason for the breakdown was a divergence in technical approaches. Figure believed that general-purpose large models could not adapt to the hardware requirements of robots; it was essential to create a vertically integrated end-to-end model. This also prompted OpenAI, after a six-year hiatus, to choose to "revive" its robotics team, personally entering the field and upgrading robotics from an "investment" to an "internal strategic business."
Simultaneously, this is also OpenAI's move to outline a new growth narrative for the capital market ahead of its IPO. It aims to show investors a grand vision of transitioning from pure software to a software-hardware combination and expanding from the virtual world to the physical world. The company hopes to use the "embodied intelligence" story to hedge against market concerns about the sustainability of its business model and its massive losses.
OpenAI's advantage in entering the robotics field lies in its globally leading AI large model capabilities, particularly its "world model" for understanding and simulating the physical world. Its technical path may differ from many companies starting with hardware platforms, instead following a logic of "first build the brain, then grow the body." This means first enabling the AI to understand physical laws through a powerful world model, then infusing that capability into physical robots. If successful, this software- and algorithm-defined hardware approach could reshape the R&D model of the robotics industry.
This article is from "Jiemian News," author: Li Kefeng
