Scaling Up Domestic AI Chip Technology: SenseTime’s Galaxy Project

SenseTime, a leading AI company, has recently announced the launch of the Galaxy Project in collaboration with nearly 20 partners to enhance the AI chip infrastructure in China. The project aims to scale domestic AI chip capabilities in the country.

During a keynote presentation titled ‘Intelligent Transformation and Symbiosis,’ Yang Fan, the co-founder and president of SenseTime’s Large Device Business Group, outlined the company’s strategy to create a closed loop system that connects chip-level technology, ecosystem partnerships, and commercial deployment for locally-produced AI computing power.

In addition to the Galaxy Project, SenseTime has entered into a space computing agreement with Guoxing Aerospace, a satellite manufacturer, and established a research partnership with five institutions, including the Shanghai Artificial Intelligence Laboratory, to focus on scientific computing applications.

Yang emphasized the significance of three key trends driving this initiative: increasing demand for tokens in enterprise deployments, the growing adoption of industrial AI applications, and the commercialization of domestic chips in China, enabling the rapid establishment of intelligent computing centers based on Chinese silicon.

However, the validity of SenseTime’s claims relies heavily on undisclosed numbers that have not been independently verified, particularly regarding token throughput figures and cost-effectiveness claims associated with their infrastructure.

SenseTime projects a substantial increase in token processing capacity on its large-scale device platform, with a forecasted rise to 10 trillion tokens per day by the fourth quarter of 2026. These projections, along with the company’s claims of improved cost-effectiveness, should be viewed with caution until verified data becomes available.

The company also highlights its efforts to address the historical challenges faced by domestic AI chips, such as software stack fragmentation, through the development of a full-stack adaptation layer that enables seamless migration of workloads across different chip architectures without extensive rewrites.

Furthermore, SenseTime introduced a new metric called Tokens Per Watt to measure AI data center efficiency and unveiled a Computing-Power Collaboration Agent to optimize resource scheduling, electricity price prediction, and energy storage across their data system.

The Galaxy Project’s ecosystem includes various domestic chip vendors and infrastructure firms, reflecting a comprehensive approach to enhancing China’s innovation capabilities in AI chip production and application scenarios.

Looking ahead, SenseTime is exploring advancements in optical computing, quantum computing applications, and space computing partnerships to extend the reach of Chinese AI services into new environments and support the country’s AI exports globally.

On the physical infrastructure front, SenseTime has established state-of-the-art data centers in Shanghai, Yancheng, Hong Kong, and plans to launch China’s first overseas domestic computing cluster in Saudi Arabia. These facilities are designed to support a wide range of industries and research initiatives, emphasizing the company’s commitment to advancing AI technologies.

In conclusion, SenseTime’s ambitious plans and partnerships signal a significant step forward in China’s AI chip infrastructure development. However, the company’s claims should be carefully monitored and validated as they progress towards their future goals.

If you want to learn more about AI and big data from industry leaders, consider attending the AI & Big Data Expo events taking place in Amsterdam, California, and London as part of the TechEx series. These events provide valuable insights into the latest trends and innovations in AI technology.

AI News is brought to you by TechForge Media, offering a range of enterprise technology events and webinars. Explore upcoming events and stay informed about the latest developments in the tech industry.