TechWhoKnows
Log inSign up
Latest
AI
Startups
Venture
Tech
Security
Hardware
Biotech & Climate
Markets & Policy
Companies
TopicsNewsletterAbout
TechWhoKnows

Independent tech news from Singapore, with the sources linked.

Sections

  • AI
  • Startups
  • Venture
  • Tech
  • Security
  • Hardware
  • Biotech & Climate
  • Markets & Policy

TechWhoKnows

  • Companies
  • About
  • Newsletter
  • All topics
  • Search
  • Account

© 2026 TechWhoKnows. All rights reserved.

  • Privacy
  • Terms
Home/AI/Models & Research

OpenAI Dissolves Standalone Robotics Unit to Fold Hardware Work Into Core Models

OpenAI disbanded its dedicated robotics research group, moving engineers into core foundation-model teams as multimodal models take over physical control.

OpenAI Dissolves Standalone Robotics Unit to Fold Hardware Work Into Core Models

By Jin, Founder & Startups

15 Nov 20262 min read

OpenAI has dissolved its dedicated robotics research group, folding its researchers and engineers directly into the company's core foundation-model and application divisions.

The internal restructuring ends OpenAI's second major experiment with an isolated robotics team. Executives concluded that general-purpose multimodal foundation models have reached a level of spatial reasoning and tool execution where physical control no longer requires a separate research silo.

Shifting from bespoke robotics to unified models

Historically, robotics stacks separated perception, motion planning, and low-level actuator control into distinct software pipelines. Modern vision-language-action (VLA) architectures instead map camera feeds and natural-language instructions directly to joint trajectories inside a single neural network.

By embedding robotics specialists inside its main pre-training and post-training teams, OpenAI intends to train future flagship models natively on physical-world trajectories, simulation data, and teleoperation logs alongside text, code, and video.

How rivals are approaching physical AI

The reorganisation puts OpenAI's internal structure in contrast with competitors maintaining dedicated embodied-AI divisions. Google DeepMind recently introduced Gemini Robotics 2.5 as a specialised VLA family, while Asian hardware ventures such as Xilect, a joint robot-chip venture backed by MetaX and UBTech that raised CNY 500 million, are building custom edge silicon to challenge Nvidia's Thor robotics platform.

For robotics startups building on third-party APIs, OpenAI's consolidation means embodied capabilities will likely arrive as standard multimodal endpoints inside upcoming model releases rather than through a separate robotics SDK.

Subscribe to the TechWhoKnows newsletter for AI lab restructurings, model releases, and engineering analysis.

Sources

  1. 1The Decoder Report on OpenAI Robotics Restructuringthe-decoder.com
  2. 2OpenAI Research Updatesopenai.com

More on

  • OpenAI
  • Robotics
  • Multimodal-AI
  • Foundation-models
  • Physical-AI

Further reading

Agents & Tools

Agile Robots Raises $500M Series F at $8.5B Valuation Backed by SoftBank and Nvidia

13 Nov 2026

Models & Research

OpenAI Splits GPT-6 API Into Three Pricing Tiers and Lists ChatGPT Enterprise on AWS Marketplace

13 Nov 2026

Semiconductors

Nvidia Robotics Engineer Darrell Dai Resigns Over How Jensen Huang's Exit Was Handled

15 Nov 2026

Models & Research

Bye Bye Llama, Hello Muse Spark

9 Apr 2026

Follow TechWhoKnows for real-time coverage

Stay updated with the latest developments in technology.

XBlueskyThreadsInstagramLinkedInTikTokYouTube

The TWK Newsletter

Pick your topics and we'll email you our stories on them, with links to the sources.

In this story

Related stories

Agile Robots Raises $500M Series F at $8.5B Valuation Backed by SoftBank and Nvidia

13 Nov 2026

OpenAI Splits GPT-6 API Into Three Pricing Tiers and Lists ChatGPT Enterprise on AWS Marketplace

13 Nov 2026

Nvidia Robotics Engineer Darrell Dai Resigns Over How Jensen Huang's Exit Was Handled

15 Nov 2026