Chinese artificial intelligence giant SenseTime has officially launched a large-scale infrastructure initiative called "Galaxy." As part of this project, the company has joined forces with nearly two dozen chip and telecommunications equipment manufacturers. The ultimate goal is to create five independent computing clusters, each equipped with 10,000 Chinese AI accelerators.
Five Clusters, One Hub, and Startup Support
The Galaxy project involves not only the construction of the clusters themselves. SenseTime's plans include deploying a specialized token processing hub, as well as providing direct support to 200 Chinese AI startups. Thus, the initiative covers the entire value chain: from hardware to software solutions and venture funding.
A key feature of the project is the complete abandonment of foreign semiconductor solutions. All accelerators that will form the basis of the clusters are manufactured within China. This is a direct continuation of Beijing's policy to achieve technological sovereignty amid tightening export restrictions from the United States.
Scale and Uncertainty
The total computing power of the five clusters (50,000 Chinese AI accelerators) is capable of training GPT-level models, but the efficiency of these chips relative to Western counterparts (NVIDIA H100/B200) remains an open question. SenseTime has not disclosed either the construction completion timelines or the planned investment volume. This is typical for Chinese megaprojects in the semiconductor sector, where many details are classified.
Commentary from Cryptalist analyst: The Galaxy project is not just a technological initiative but a strategic signal to the market. SenseTime is effectively challenging NVIDIA's monopoly in the AI accelerator market, but success will depend on the real performance of Chinese chips. If the clusters demonstrate competitive results in training large language models, we will witness the formation of a new power center in the global AI infrastructure. However, without clear timelines and a budget, it is still difficult to assess the realism of these ambitions.