Light Origins Closes Pre-A Round of Several Hundred Million Yuan to Build Physical AI Foundation Models, Founded by ChatGPT Core Contributor Roger Jiang

SHENZHEN, August 14, 2026 — Light Origins, an AI company building physical AI foundation models, today announced the closing of its Pre-A funding round of several hundred million yuan. The round was led by CAS Investment, with participation from China Merchants Venture Capital, Xiang He Capital, and other institutional investors. This funding will accelerate the company’s large-scale physical AI model training, multimodal data infrastructure development, and full-stack software and hardware R&D, while supporting its continued expansion of a world-class team.

Founded in late 2024, Light Origins is dedicated to extending foundation model intelligence from the digital domain into the real world, advancing toward physical AGI by pioneering a breakthrough in the three scaling paradigms of physical AI.

Light Origins has proposed and is now putting into practice the three scaling paradigms, consisting of Scalable Pre-Training, Scalable Alignment, and Scalable Deployment. Scalable Pre-Training focuses on establishing foundational capabilities through large-scale data training, with Scalable Alignment further enhancing model capabilities through reinforcement learning and other approaches. Scalable Deployment then brings foundation models into real-world scenarios and generates high-quality data, forming a data flywheel that drives continuous improvement of foundation models.

Every major leap in artificial intelligence has come from a paradigm breakthrough, which requires more than isolated innovations in algorithms or model architectures.

— Roger Jiang, Founder and CEO of Light Origins

Roger Jiang previously worked at OpenAI, where he took part in the company's full-stack algorithm research from 2019. His work spanned the complete technical evolution from InstructGPT and ChatGPT to GPT-4, with core contributions covering pre-training, alignment, optimization and infrastructure. He is credited eight times in the GPT-4 Technical Report, tied with OpenAI co-founder Greg Brockman as the most-credited contributor.

Light Origins now has more than 100 employees, over 90% of whom work in R&D. Its algorithm team is made up of PhDs from QS Top 20 universities and engineers from world-leading technology companies. Its hardware team brings solid experience from companies including Huawei and DJI, covering the full lifecycle of consumer robotics from product design through R&D and manufacturing to commercialization, making Light Origins one of the few companies possessing hardware-software integration capability.

Zhou Xiaofeng, Managing Director at CAS Investment, said:

Physical AI is still in its early stages, where foundational technological capabilities will ultimately shape the long-term landscape. What sets the Light Origins team apart is its rare combination of expertise in large-scale model training and data processing, engineering experience in deploying models into real-world systems, and strong capabilities in robotic hardware platforms and whole-body motion control. This multidisciplinary strength is a key quality we value in founding teams. We believe Light Origins can continue to build long-term technical advantages along its roadmap and grow into a globally competitive physical AI foundation model company.

Ultimately, physical AI will need to be deployed in real-world environments and proven through rigorous validation of reliability, cost efficiency, and scalability. By building from foundational models while advancing its software-hardware stack and deployment capabilities, Light Origins has established a clearer connection between technological innovation and industrial applications. We look forward to leveraging our industrial application scenarios and ecosystem capabilities to work with the team in accelerating the commercialization of physical AI technology.

— China Merchants Venture Capital

Physical AI is not a standalone challenge, but rather the extension of language-based intelligence into the physical world. Those who truly understand the underlying logic of how intelligence emerges in foundation models stand a better chance of bringing this paradigm to robotic systems. We believe the next wave of breakthroughs may well come from founders like Roger Jiang, who equipped with solid background in foundation models. We look forward to partnering with Light Origins as it continues to push the boundaries of physical AI along this long-term technical path.

— Xiang He Capital

Looking ahead, Light Origins will continue to advance in large-scale physical AI model training, multimodal data infrastructure development and robotic systems validation. The company plans to publish a series of technical reports and model demonstrations, showcasing its progress on physical AI foundation models and cross-embodiment capabilities.