OpenAI scraps standalone robotics team as ASUS puts humanoid to work
OpenAI has dissolved its independent robotics research group, folding those researchers into its foundation model and applied AI divisions — a reorg Fortune reported on Nov 13 and the company confirmed. Separately, ASUS showed off Gallop, its first humanoid, heading for factory trials, while Samsung is pushing into AI data centers.
- Original source: note.com.
- The source published it on Nov 15, 2026 at 11:07 PM.
- No second outlet has reported it yet. Treat it as a single-source claim.
OpenAI has broken up its standalone robotics research team and moved the researchers into other internal groups. Fortune reported the restructuring on Nov 13, and OpenAI confirmed it when the magazine asked. The robotics researchers land in the foundation model research department run by Chief Research Officer Mark Chen, while the commercialization and hardware side joins the Applied department under President Greg Brockman.
The company frames this as a strategy shift, not a retreat. OpenAI says it has validated that general-purpose models such as GPT-6 Astra can handle robot control, which reduces the case for training dedicated robotics models that read camera images and generate joint movements on their own. Work on a humanoid continues: according to Fortune, CEO Sam Altman has pointed to a first public demonstration in 2027, though specs, mass production plans and a commercial release date are still unset.
It's the second time OpenAI has shut down an independent robotics organization. The first came in 2020, when real-world data for robot learning was hard to come by and the company chose to pour resources into large language models instead. The team was rebuilt in 2024 and worked with humanoid maker Figure, a collaboration that ended in February 2025.
ASUS announced on Nov 13 where its first humanoid, Gallop, stands. The robot is learning to pack 15 small motherboard accessories, including parts sealed in thin transparent bags that deform and slip when grabbed — a harder job than lifting rigid components. ASUS wants each process done in under 20 seconds at a 99.5% success rate, but that's a target, not a result: the company hasn't disclosed its current measured rate. Gallop learns by watching skilled workers, plus simulation and data collected through VR remote operation.
The machine runs on a single NVIDIA Jetson AGX Thor computing platform, with ASUS building the body and hands in-house and separating work planning, progress monitoring, movement and machine control in software so the AI models can be swapped out later. At this stage Gallop is two arms on a fixed base, not a bipedal robot. Customer trials are planned for 2027 and a market launch for late 2028.
Elsewhere in the same wave: US-based World Labs released its spatial generation AI, Atlas, for general use alongside a developer API. South Korea's Samsung is looking to expand into building and operating AI data centers, leaning on the group's existing businesses in semiconductors, cooling equipment, batteries and construction. The competitive ground is shifting from single models toward real-world deployment and the infrastructure underneath it.
TopicsOpenAI · ASUS · Gallop · Samsung · World Labs · Sam Altman · Mark Chen · Greg Brockman
Source: note.com
Written by AI from verified reporting.
More stories
- Tech & AI blockchain.news 0 viewsGrid Robotics paper pitches 'agentic robotics' powered by GPT-6 Astra
- Tech & AI x.com 0 viewsNvidia director warns board of losing "irreplaceable talent" if Huang isn't reinstated as CEO
- Tech & AI blockchain.news 0 viewsNVIDIA's Vera Rubin aces MLPerf debut with 10x agentic efficiency gain
- Tech & AI blockchain.news 0 viewsOpenAI Adds Computer-Use Agents and 20+ Plugins to ChatGPT for Healthcare

