OpenAI is building a humanoid robot, and Sam Altman says the brain is the hard part

The OpenAI humanoid robot is no longer a rumor. Sam Altman has said plainly that the company will build one, and he expects a demo in 2027. A fresh look at the effort by Engadget this week, along with a growing list of robotics jobs on OpenAI's careers page, has pushed the project back into the spotlight. The bigger story is not the body of the robot. Altman's own framing is that the hard part is the brain, the AI that lets a machine see, plan and act in a messy human world. Here is what we know, what is still secret and why it matters.

What Sam Altman said

Altman's clearest comments came on the Sources podcast, which Forbes reported on September 3. Asked whether OpenAI would build a humanoid, he answered, "We will." He went further, according to a recap by tbreak: "We will definitely do a humanoid," adding that OpenAI would "do other form factors as well."

His reasoning is practical. The world, he said, is "very much designed for people." Doors, keyboards and tools all assume a human body. A robot that needs to work in the same spaces will likely need a similar shape, so "matching that form factor seems good."

On timing, Altman was more careful. In a Fortune interview from September 15, cited by Engadget, he said: "I think we'll have a cool demo in 2027 ... I don't think there will be robots walking the streets for a few years longer."

That is the full public timeline. A demo next year. Real world use later, with no date attached.

Why the OpenAI humanoid robot is trending now

Nothing official changed this week. What changed is the evidence. Engadget's report, published October 17, walked through OpenAI's robotics hiring and found 27 open robotics roles on the careers page. TokenPost and tbreak followed on October 18, both stressing the same point: the 2027 date is a demonstration, not a product launch.

The job listings tell a more detailed story than Altman's interviews. According to Engadget, the postings cover:

• Motion control and state estimation, the systems that let a robot know where its limbs are and move them smoothly
• Actuator control and whole robot behavior
• Machine learning infrastructure for training physical AI systems
• Custom electromechanical actuators, meaning motors, transmissions, sensing and thermal systems
• Manufacturing and supply chain roles for producing hardware at scale

That last group is the most telling. Companies hire supply chain staff when they plan to make more than a handful of prototypes.

One posting, as described by Engadget, lays out the goals in two stages. In the short term, OpenAI wants general purpose robots that help skilled workers build infrastructure. In the long term, the aim is a personal robot that can handle a much wider range of tasks.

The brain matters more than the body

Altman said the robot's design is "less important than really figuring out ... the brain that makes the robot work." That line explains why OpenAI thinks it can compete in a field where others have years of hardware experience.

Think about a simple task, like picking up a coffee mug. Engadget uses this example, and it is a good one. The robot has to see the mug, understand the instruction, figure out where to put its hand, adjust its grip and correct itself if the mug slips. Each step sounds easy for a person. For a machine, each one is a separate hard problem, and they all have to work together in real time.

This is where OpenAI's strength in large AI models could help. The same kind of systems that power chatbots and AI agents are increasingly being trained to understand images, video and physical space. A robot that can follow loose spoken instructions needs exactly that kind of general understanding.

OpenAI is not starting from zero. Its earlier Dactyl research project trained a robotic hand to handle real objects using only simulation, and the team later used the same approach to have the hand work a Rubik's Cube. That was a research demo, not a product, but it showed the company's interest in training robots in virtual worlds before letting them loose in the real one.

The competition is already walking

OpenAI is entering a crowded field. Tesla has worked on its Optimus humanoid for years. XPENG, the Chinese automaker, recently had its IRON humanoid walk off the production line on its own after assembly, according to Engadget.

Those companies have a head start on hardware: factories, motors, batteries and the experience of building physical products at scale. OpenAI's bet seems to be that the software will decide the winner. If a robot body becomes a commodity, the company with the smartest control system could end up with the most valuable piece.

There is also a race among AI labs themselves. Meta has been pushing its own frontier models, including the launch of Muse Wonder, and every major lab is looking for ways to move AI off the screen and into the physical world. Robotics is the most obvious next step.

What OpenAI has not said

For all the attention, the list of unknowns is long. OpenAI has not disclosed:

• What the 2027 demo will actually show
• What the robot will look like
• When people outside the company will get to try it
• A price, a subscription model or a release date
• Any privacy framework designed for robots

That last point deserves attention. A home robot would carry cameras and microphones into private spaces. It would see faces, hear conversations and map rooms. OpenAI already faces scrutiny over how its software behaves. The company recently disclosed several misalignment incidents involving its AI systems. Putting similar systems in a machine that can move around a house raises the stakes considerably.

Who is affected

For most people, nothing changes soon. A 2027 demo means a consumer robot is likely years away, and Altman himself said robots on the streets are "a few years longer" beyond that.

The people who should pay attention now are:

• Workers in construction, energy and infrastructure. OpenAI's stated short term goal is robots that help skilled workers build infrastructure. That suggests early robots would assist on job sites rather than replace whole crews, but it is worth watching.
• Robotics engineers. With 27 open roles, OpenAI is competing with Tesla, startups and other labs for a limited pool of hardware and controls talent.
• Investors and partners. OpenAI has not named any hardware partners. Whoever supplies its actuators, sensors or manufacturing could gain a lot.
• Regulators. Physical AI raises safety questions that software alone does not. A chatbot that makes a mistake gives a wrong answer. A robot that makes a mistake can hurt someone.

What to watch next

The next year will show how serious this effort is. A few signals to look for:

1. Hiring growth. If the robotics job count keeps rising, OpenAI is scaling the team ahead of the demo.
2. Hardware partners. A named manufacturing or component partner would suggest plans beyond a lab prototype.
3. Model updates. Watch for OpenAI models that handle video, spatial reasoning and real world planning. Those are the building blocks of a robot brain, and they could show up first in products like ChatGPT, which recently moved to GPT-6 with Intelligent UI.
4. Safety commitments. Any robot specific safety or privacy policy would be a sign OpenAI is preparing for public use.

Altman has a long record of setting ambitious targets. A cool demo in 2027 is a modest promise by his standards, which may be a sign of how difficult this really is. For more on where AI is heading, visit our AI page.

Frequently asked questions

Is OpenAI really building a humanoid robot?

Yes. Sam Altman said on the Sources podcast that OpenAI will build a humanoid, and the company lists 27 robotics jobs on its careers page, according to Engadget.

When will the OpenAI humanoid robot be available?

There is no release date. Altman told Fortune he expects "a cool demo in 2027," and said robots walking the streets are a few years beyond that.

How much will OpenAI's robot cost?

OpenAI has not announced a price, a subscription model or any consumer plans. Its stated short term focus is robots that help skilled workers build infrastructure.

Who are OpenAI's competitors in humanoid robots?

Tesla has worked on its Optimus humanoid for years, and XPENG recently showed its IRON humanoid walking off its production line on its own.