Tech firms push global open-source AI cooperation at CES 2026

Tech companies from China, the US, and Europe highlighted open-source AI collaboration at CES 2026 in Las Vegas, which ended on January 10.

Industry leaders noted that open-source models are driving wider AI adoption and international cooperation.

Chinese firms, including Alibaba and DeepSeek, released AI models under open licenses, allowing developers worldwide to modify and build on them.

Robotics companies like Unitree and Agibot demonstrated applications of these open models in both home and industrial settings.

US-based Nvidia and Germany’s Siemens announced an expanded partnership to develop an industrial AI operating system that supports virtual factory simulations.

Boston Dynamics, with Google DeepMind and Nvidia, unveiled a new humanoid robot developed through open machine learning research.

In healthcare, Abridge and other firms are using open-source tools to automate tasks and advance drug discovery.

New partnerships, such as Dolby’s collaborations with Hisense and TCL, focused on integrating software and hardware for improved display technology.

🔗 Source: Xinhua

🧠 Food for thought

Implications, context, and why it matters.

Most “open” AI releases lack commercial clarity

DeepSeek-V3 uses MIT License for code and a modified Open Responsible AI License (OpenRAIL) for the model that blocks illegal or hazardous use 1. Commercial use has no fees, and developers must retain copyright notices and state modifications in any distributed derivative 1.
DeepSeek-V3 weights total 685 billion across a 671 billion main model plus a 14 billion Multi-Token Prediction (MTP) module, and sit on Hugging Face (a popular repository for AI models) 2. The license allows commercial deployment plus fine-tuning, quantization, or platform integration 2 1. Dataset access is unclear, though pretraining used 14.8 trillion tokens 2.
Alibaba’s Mobile Neural Network (MNN) uses Apache 2.0, which is more permissive than modified OpenRAIL 3. That gap matters for compliance in regulated or public work.

Third-party integrators can build tooling for the Nvidia-Siemens platform

Nvidia with Siemens built an Industrial AI Operating System that joins NVIDIA Omniverse (tools for industrial digital twins), CUDA-X (GPU-accelerated computing libraries) and PhysicsNeMo (physics simulation models) with the Xcelerator portfolio (an industrial software suite) 4 5. Integrators and ISVs can build connectors or middleware for legacy plants.
Erlangen factory in 2026 acts as a blueprint site 6 7. Foxconn, HD Hyundai, KION Group, plus PepsiCo will need partners 7. MLOps (machine learning operations) teams can add monitoring and governance for the AI Brain (the central decision engine) that reviews digital twins (virtual replicas of real-world systems) 4. Entry needs alignment with NVIDIA CUDA-X and Siemens operations software 6.