Key Takeaways
- OpenAI now lists 27 open robotics roles, up from 11 in May 2026, with base salaries reaching $500,000, continuing its build-out since restarting the robotics program it shut down in 2021.
- The highest-paying role, at up to $500,000, goes to a machine learning engineer building distributed data infrastructure for robotics training, OpenAI’s data-scarcity problem, not humanoid hardware, is where the company is paying the most.
- Aditya Ramesh, the engineer behind DALL-E and a contributor to Sora, now heads the robotics division, with a stated mission of reaching AGI-level intelligence through general-purpose robotics in real-world settings. Five years after walking away from its first robotics effort, OpenAI is back, and it is paying $500,000 base salaries to prove it is serious. The company now lists 27 open robotics roles, up from 11 in May 2026, spanning actuator design, firmware, PCB layout and distributed data systems: silicon-to-software work, not software wrappers around someone else’s hardware.
Building Custom Hardware
The job listings point to full vertical integration of OpenAI’s robotics stack. Actuator design engineers, firmware engineers and PCB layout specialists all appear, confirming the company intends to design custom robotic systems rather than write control software for third-party hardware, the approach most AI companies settle for. Controlling everything from the motors driving robotic joints to the embedded electronics gives OpenAI direct influence over how its models interact with the physical world. Hardware constraints were the cited reason for the 2021 shutdown; owning the stack this time removes that dependency.
The Data Bottleneck
The top-paying role, base salary up to $500,000, goes to a machine learning engineer focused on distributed data systems, responsible for moving and processing large volumes of robotics training data across compute infrastructure. That prioritisation reflects where OpenAI sees the real constraint. Large language models could draw on existing internet text; physical AI cannot. Real-world interaction data has to be generated from scratch, at scale, under controlled conditions. OpenAI is also hiring to run what the listings describe as “data collection facilities,” suggesting the company plans to manufacture its own training data rather than wait for it to accumulate organically. The data sourcing problem in physical AI is structurally different from anything solved in language modelling, and OpenAI is treating it as an engineering problem rather than a research one.
AGI Through Embodied Intelligence
Aditya Ramesh, the engineer behind DALL-E and a contributor to Sora, now leads the robotics division. The team’s stated mission is general-purpose robotics as a route to AGI-level intelligence in dynamic, real-world settings, the argument being that physical embodiment surfaces learning problems that digital-only environments cannot. CEO Sam Altman has said OpenAI plans to develop humanoid robots alongside other physical form factors, according to public comments, though no product timeline has been disclosed. Whether physical interaction is necessary for AGI remains an open research question the field has not settled; OpenAI’s hiring pattern makes clear which side of that debate the company is funding.
A Tighter Talent Market
Base salaries from $177,000 to $500,000, excluding equity, put pressure across the entire robotics hiring market. Figure AI, in which OpenAI holds an investment stake, and 1X Technologies both compete for the same hardware, firmware and ML engineers. So do Tesla and Google. Smaller robotics firms running leaner budgets have less room to counter offers at this level. The skills involved, physical system design, embedded firmware, robot learning, are genuinely scarce, and OpenAI’s ceiling makes a supply gap that already existed at more typical pay bands considerably sharper.
Starting in Data Centers
The long-term pitch is a robot in every home. The near-term plan is more grounded. Altman’s public comments point to data centers and infrastructure-heavy settings as the first deployment targets, environments where tasks are repetitive, conditions are predictable and performance is measurable. That controlled context also generates cleaner training data, letting OpenAI refine models and hardware iteratively before confronting the variables of domestic or public spaces. Prove the system works where failure is recoverable before shipping it somewhere it isn’t.
Originally published at https://autonainews.com/openai-offers-robotics-engineers-base-salaries-up-to-500k/
Top comments (0)