At CES 2026, NVIDIA announced the open-sourcing of its autonomous driving software solution Alpamayo during the keynote, including the Alpamayo series AI models, simulation tools like AlpaSim, and large datasets. This solution is designed to address the long-tail issues in autonomous driving and supports autonomous capabilities from L2++ to L4 levels.

However, on the hardware side (DRIVE AGX Thor SoC), it remains proprietary and is only available to partners. The open-source components have been launched on Hugging Face and GitHub, opening the door for developers worldwide.

What does this mean? Small and medium-sized automakers and traditional brands now have new opportunities. Players previously blocked by technical barriers can now quickly enhance their autonomous driving capabilities by purchasing NVIDIA chips. In the short term, this puts pressure on leading manufacturers who have heavily invested in self-developed solutions.

Some believe Tesla FSD still leads, but essentially, NVIDIA is just selling chips—its supporting solutions are becoming more complete, and its ecosystem is maturing. The idea of a "physical AI GPT moment" sounds exciting, but there’s still a distance to go before it’s realized.

What lessons can Huawei, Li Auto, and Xpeng learn? One path is to go all-in on self-developing FSD; another is to rely entirely on platform solutions. The latter seems easier, but platform providers will eventually catch up. If you don’t maintain continuous technological leadership, your previous R&D investments could be wasted. This is an unavoidable choice.