OpenAI has officially restarted its robotics business after a six-year hiatus, announcing a major hiring push to build humanoid robots that can assist workers and eventually serve as personal assistants. The company's robotics division is being led by Aditya Ramesh, the same vice president of research who architected Sora, OpenAI's advanced video generation model. This leadership choice reveals how OpenAI plans to leverage its world-simulation capabilities, originally developed for video, to understand and control physical robots.

On May 31, 2026, OpenAI CEO Sam Altman posted a public recruitment notice announcing the formation of "OpenAI Robotics" and calling for full-stack hardware engineers, operations engineers, systems engineers, and machine learning engineers. Altman described the team's progress as "rapid," using the same language he employed when describing ChatGPT's development before it transformed global technology use. The decision to restart robotics comes after the company disbanded its robotics team around 2020 to focus resources on large language models, a strategic choice that ultimately led to ChatGPT's creation.

What Changed to Make OpenAI Return to Robotics?

OpenAI's robotics restart is rooted in a major internal research project called "Worldsim," which evolved into the robotics division over the past year. This project represents the deep integration of robotics hardware research with machine learning research, according to company leadership. The timing is significant: OpenAI's return to robotics follows the breakdown of its partnership with Figure AI in February 2025, just one year after the companies announced a collaboration on a dedicated multimodal AI model for humanoid robots.

Figure AI founder Brett Adcock cited diverging technical approaches as the reason for ending the partnership. Figure believed that general large models cannot meet the specific hardware requirements of robots and that an end-to-end model with vertical integration was necessary. This disagreement prompted OpenAI to "revive" its robotics team and upgrade robotics from an investment strategy to an internal strategic business. The move also serves another purpose: it provides investors with a new growth narrative before OpenAI's anticipated initial public offering, expected as early as September 2026, with potential valuations exceeding one trillion dollars.

How Does OpenAI's Robotics Strategy Differ From Competitors?

OpenAI's approach to robotics follows a philosophy of "build the brain first, then grow the body." Rather than starting with hardware design, the company plans to leverage its world model, the same technology underlying Sora, to help AI understand physical laws and simulate the physical world. This AI understanding would then transfer to physical robots. The strategy contrasts with companies like Tesla, Figure AI, and Agility Robotics, which have built robotics programs from the hardware side up.

OpenAI's competitive advantages in this space include its globally leading large AI model capabilities and its deep expertise in world simulation. If this approach succeeds, it could reshape the research and development model of the entire robotics industry by defining hardware with software and algorithms rather than the traditional reverse.

What Are OpenAI's Short-Term and Long-Term Robotics Goals?

OpenAI has articulated two distinct phases for its robotics ambitions:

• Short-term focus: Developing robots that can assist skilled workers in building future infrastructure, addressing immediate labor needs in construction, manufacturing, and warehousing.
• Long-term vision: Creating personal robots that can meet various needs for individual users, eventually making humanoid robots accessible to everyone.
• Sector applications: Healthcare and elder care, where labor shortages are acute; retail and hospitality, where early deployments are already underway; and home and personal use, which remains the longer-term target.

Figure AI has already begun deploying robots in real retail environments through partnerships with JCPenney, Aéropostale, and Brooks Brothers, demonstrating that humanoid robots are moving beyond research labs into commercial settings. OpenAI's entry into the market is expected to accelerate adoption across these sectors.

Why Does OpenAI's Robotics Move Matter Now?

OpenAI's return to robotics signals a fundamental shift in how the company views its future. The company faces significant financial pressures: it expects to incur losses of approximately 14 billion dollars in 2026, with cash consumption expected to increase further. The company is not projected to achieve positive cash flow until 2030 at the earliest, and its gross profit margin stands at only about 33 percent due to high AI model inference costs.

By expanding from pure software into hardware and from the virtual world into the physical world, OpenAI is attempting to depict a new growth curve for capital markets. The company is using the narrative of "embodied intelligence" to address investor concerns about business model sustainability and massive losses. This strategic repositioning comes as OpenAI prepares for its anticipated public offering, where the robotics story could significantly influence valuation.

The broader industry context matters too. When OpenAI enters a market, it typically accelerates competition and innovation across the entire sector. The company has deep financial resources, access to top talent, and a proven ability to generate global headlines, a combination that tends to shake up competition in ways few other organizations can match. Industry insiders note that when a company begins recruiting openly at this scale, it usually means substantial work has already progressed behind the scenes.

How Does Sora's Technology Connect to Robotics?

The appointment of Aditya Ramesh, Sora's creator, to lead OpenAI Robotics is not coincidental. Sora's core technology involves understanding and simulating the physical world, predicting how objects move, how physics works, and how scenes evolve over time. These same capabilities are essential for robots to understand their environment, predict outcomes of their actions, and operate autonomously in complex physical spaces.

Ramesh's background includes being a core developer of DALL-E, OpenAI's text-to-image model, before moving to video generation. His expertise in visual understanding and world simulation makes him uniquely positioned to bridge the gap between video generation and physical robotics. The Worldsim project that evolved into OpenAI Robotics represents the culmination of years of research into how AI can model and understand the physical world.

OpenAI's historical robotics work provides additional context. From 2016 to 2019, the company developed the OpenAI Gym reinforcement learning benchmark and the Roboschool simulation platform. In 2019, OpenAI trained a humanoid robotic hand to solve a Rubik's Cube using reinforcement learning and automatic domain randomization technology, proving that AI trained in simulation could transfer its capabilities to real robots. That research foundation, combined with modern large language models and world simulation, positions OpenAI to approach robotics from a fundamentally different angle than traditional robotics companies.

The timeline for OpenAI's robotics products remains unclear. Altman's announcement was a hiring call rather than a product launch date, but industry observers suggest that public recruitment at this scale typically indicates that foundational work has already begun. The robotics division is expected to move quickly, given OpenAI's track record of rapid development cycles and the company's need to demonstrate new growth opportunities before its public offering.