NVIDIA and Doosan Group Significantly Expand Partnership in Physical AI and AI Factories

NVIDIA and South Korea's Doosan Group have significantly expanded their partnership across physical AI, robotics, and AI factory infrastructure. The agreement covers Doosan Robotics, Doosan Bobcat, Doosan Enerbility, and the Electronic Materials Business Division of Doosan Corporation.

In **robotics**, Doosan Robotics will integrate NVIDIA Isaac Sim, Isaac Lab, Cosmos world foundation models, the Newton physics engine, and Jetson Thor to build an AI-driven platform called the "Agentic Robot OS." The initiative is expected to enhance the perception, reasoning, and action capabilities of industrial robots while improving the adaptability and deployment scale of collaborative robots. Plans also include developing reference use cases for high-value tasks such as depalletizing and polishing, as well as new form factors such as dual-arm and humanoid platforms.

Doosan Bobcat is also exploring the integration of NVIDIA physical AI technologies into its construction, agriculture, and material handling equipment, aiming to build specialized world models for autonomous construction machinery.

In **power infrastructure**, Doosan Enerbility will explore power solutions for NVIDIA AI factories through its large-scale power infrastructure portfolio, including gas turbines, steam turbines, small modular reactors (SMRs), and hydrogen fuel cells. It will also evaluate low-carbon power sources to meet the rapidly growing electricity demand of AI data centers.

In **electronic materials**, the Electronic Materials Business Division of Doosan Corporation will support AI data center infrastructure through copper-clad laminates (CCL). CCL is a material essential to the printed circuit boards found in networking equipment, AI accelerators, and server motherboards, contributing to the NVIDIA MGX modular architecture ecosystem.

The partnership shows how the AI factory ecosystem is expanding across multiple layers, from chip design to power supply and robotics.

Hitachi Secures Access to Anthropic's Claude Mythos

ITmedia reports that Hitachi has secured access to Claude Mythos, Anthropic's advanced AI model. NEC and Fujitsu have already announced collaborations with Anthropic, making Hitachi the third major Japanese company to strike a large-scale deal.

The succession of Japan's leading IT and heavy industry companies — NEC, Fujitsu, and Hitachi — securing access to Anthropic products reflects Anthropic's growing presence in Japanese companies' AI infrastructure strategies. Which use cases each company will apply Mythos to awaits future announcements, but competition in industrial AI solutions seems likely to accelerate.

SpaceX IPO Documents Mention Orbital AI Computing

A line reading "Deploy orbital AI compute at scale" has been found in IPO-related documents SpaceX filed with the U.S. Securities and Exchange Commission (SEC). It is drawing attention on Hacker News.

Ground-based AI data centers face power consumption and cooling challenges, so tapping compute resources in space is theoretically attractive. However, no specific technical details or deployment timeline have been revealed at this point, and the mention appears to be part of a long-term vision. Even so, the suggestion that SpaceX could repurpose its Starlink satellite constellation infrastructure for AI computing is worth noting.

AI-Native OS "VibeOS" Emerges

A project named "VibeOS," billed as the first AI-native operating system, has been released. It earned 8 points with 6 comments on Hacker News.

Details are limited to the official site (vibeos.sh), but the concept appears to place an LLM at the core of the OS, with system operations carried out through conversation with the user. As an attempt to implement a natural language interface at the OS level as an alternative to conventional GUIs and CLIs, the concept itself is intriguing. Whether it is ready for practical use, however, is unknown at this point.

Meanwhile, the topic "Fully Hallucinated Operating System" has also been discussed in Reddit's LocalLLaMA community, underscoring growing interest in fusing operating systems with LLMs.

Should AI Handle Workflow Approvals? 70% Say No

According to a survey by Eightred, more than 70% of respondents said that approval and decision-making processes in workflows should not be entrusted to AI. ITmedia AI+ reports.

The result indicates that even as business process automation advances, there remains a strong perception that final judgment and accountability should rest with humans. While AI adoption in business continues to grow, social consensus has yet to form around the fundamental question of "who is the decision-maker."

Other News

Claude Code v2.1.168 released**: The latest version of Claude Code is a maintenance release that includes bug fixes and reliability improvements. No new features have been added

Claude Code × tmux**: To support workflows that hand long-running tasks to AI coding agents, a Qiita article explains techniques for background execution and agent management using tmux

Ask HN: Will LLM companies absorb all the value?**: Hacker News is hosting an active debate on whether the spread of LLMs could cause young people to lose career ladders, reduce creative activity, and concentrate economic value

---