On May 31, Sam Altman posted a job ad on X:
"OpenAI Robotics is hiring, looking for exceptional full-stack hardware, ops, systems, and ML engineers to help us program and manufacture robots that are useful for society."
Three signals in one sentence: OpenAI has formally established a robotics division, plans to build its own hardware, and is scaling up hiring across hardware, operations, systems, and machine learning.
Not OpenAI's first foray into robotics
Longtime users may recall OpenAI's earlier robotics project, Dactyl, launched in 2017. In 2019, its single-arm robotic hand famously solved a Rubik's Cube. Then in 2020, the entire robotics team was disbanded — the reasoning was that reaching AGI didn't necessarily require robotics, and real-world training data was too difficult to obtain.
Six years later, they are back. According to Altman, the new team grew out of the "world simulation" research project led by Aditya Ramesh, and rebuilding began in January 2025. More tellingly, after the Sora video app was shut down, the original Sora team was folded into this effort.
Short-term: building infrastructure; long-term: entering homes
OpenAI has laid out a clear two-tier strategy. The near-term goal is pragmatic: build robots that can assist skilled workers in constructing infrastructure — essentially tackling labor shortages while generating large amounts of real-world data. The long-term vision is more ambitious. Altman's own words: "everyone having a personal robot doing anything they need."
What OpenAI really wants may not be robots
There is an interesting angle here. What OpenAI needs most right now is not money, but real-world data and an alternative path beyond scaling large language models. Robots can provide both: they operate in the physical world, collecting embodied data that text and video cannot supply; and they force the team to explore embodied intelligence, a different direction from pure language models.
So rather than rushing to sell you a robot, OpenAI is likely using robotics as a vehicle to gather data and test new approaches — and, incidentally, to fit into the larger AGI narrative. The absorption of the Sora team is telling: video generation is essentially about "simulating the world," which aligns with robotics' need to "understand the world." As for when everyone will have a personal robot, Altman offered no timeline.
Sources: CocoLoop, OpenAI starts with infrastructure robots but aims for "everyone having a personal robot doing anything they need" (The Decoder); OpenAI Begins Hiring Engineers for Robotics Division (Analytics India Magazine); Sam Altman's OpenAI just made robotics its next frontier (Tech Funding News)