SenseTime Galaxy Project Plans Major Expansion of Chinese AI Chip Production

SenseTime has officially launched the Galaxy Project, partnering with nearly 20 industry leaders to dramatically scale domestic AI chip infrastructure across China. This ambitious initiative marks a significant milestone in the country's push toward technological self-reliance in artificial intelligence computing.

During a keynote presentation titled 'Intelligent Transformation and Symbiosis,' Yang Fan—SenseTime's co-founder and president of its Large Device Business Group—outlined the company's comprehensive strategy. He described what SenseTime characterizes as a closed-loop ecosystem that seamlessly integrates chip-level technology, strategic ecosystem partnerships, and commercial deployment of domestically-produced AI computing power.

🤝 Strategic Partnerships Expand Ecosystem

Alongside the Galaxy Project announcement, SenseTime signed a space computing agreement with satellite manufacturer Guoxing Aerospace. Additionally, the company established a research partnership with five prestigious institutions—including the Shanghai Artificial Intelligence Laboratory—specifically targeting scientific computing applications.

Yang positioned this strategic timing around three critical converging trends:

• Token demand surging across enterprise deployments
• Industrial AI adoption accelerating to match consumer-facing applications
• Domestic chip commercialization reaching maturity for rapid intelligent computing center deployment

However, the viability of this opportunity window depends substantially on performance metrics that have not yet received independent verification.

📊 Token Throughput Performance Claims

SenseTime reports that its large-scale device platform currently processes an average of 2.42 trillion tokens daily. The company projects this figure will experience a remarkable 25-fold increase to 10 trillion tokens per day by the fourth quarter of 2026. Enterprise buyers should note this represents a forecast rather than verified results, requiring validation through actual quarterly performance data.

Compared with domestic homogeneous inference configurations, SenseTime claims a 2.5x increase in token output at equivalent cost. The company asserts this advancement pushes optimized hybrid inference clusters beyond what the industry previously considered the minimum profitability threshold for domestic computing power.

Important note: These figures lack third-party benchmarking. The performance gap between vendor-optimized test clusters and customer production environments—with real-world data pipeline variations and firmware update delays—typically results in adjusted performance metrics.

🔧 Cross-Platform Adaptability Solutions

Domestic AI chips have historically faced challenges with fragmented software stacks, where models trained for one architecture frequently require substantial rework for alternative platforms. SenseTime addresses this with a full-stack adaptation layer spanning models, frameworks, operators, toolchains, and hardware, enabling customers to migrate workloads across domestic chip vendors with minimal code modifications.

The company highlights two practical implementation examples:

These performance metrics demonstrate promising potential in controlled testing environments, though their real-world effectiveness requires validation against actual customer pipelines operating mixed hardware generations.

⚡ Energy Efficiency Innovation

SenseTime introduced a new efficiency metric called Tokens Per Watt, positioned as an industry benchmark for measuring AI data center efficiency. The company also unveiled a Computing-Power Collaboration Agent that manages resource scheduling, electricity price prediction, and energy storage optimization across an eight-level data system with five decision chains.

By integrating compute resources, electricity pricing dynamics, and automated scheduling, SenseTime reports:

• 80% increase in token output per unit of electricity cost
• Average power prices 10% below comparable regional data centers
• 96% accuracy in computing load prediction

These claims warrant monitoring over multiple quarters rather than immediate acceptance. Electricity price arbitrage and load forecasting accuracy typically demonstrate different performance characteristics during full seasonal cycles with genuine demand volatility.

🌐 Comprehensive Partner Ecosystem

The Galaxy Project's ecosystem encompasses an impressive roster of domestic technology leaders:

SenseTime's roadmap includes:

• Construction of one "token factory"
• Five computing clusters at "10,000-calorie" scale
• Joint development across ten technology directions
• Support for 200 AI startups

"Domestic production is not simply about replacing individual chips, but rather a collaborative effort across the entire chain of China's innovation capabilities, from chips and components to infrastructure and application scenarios," — Yang Fan, SenseTime Co-founder

🚀 Future-Focused Technology Initiatives

Beyond immediate infrastructure development, SenseTime outlined ambitious long-term research initiatives:

The space computing initiative follows an ambitious timeline: first satellite launch in 2026, scaling toward thousands of computing satellites and computing capacity in the tens of thousands of petabytes by 2030.

Yang emphasized that the value extends beyond raw computational capability, positioning space-based computing as a strategic solution to extend Chinese AI services into challenging network environments such as maritime operations and disaster response scenarios, while supporting China's international AI service exports.

Note: The 2030 target timeline extends five years into the future, and satellite computing deployments at this scale lack industry precedent for timeline validation.

🏢 Global Physical Infrastructure Expansion

SenseTime's infrastructure footprint spans multiple strategic locations:

🔬 Scientific Research Collaboration

On the research front, SenseTime established partnerships with leading institutions including:

• Shanghai AI Laboratory
• Beijing Zhongguancun Academy
• Shenzhen Hetao Academy
• Shanghai Algorithm Innovation Research Institute
• Shanghai Jiao Tong University's AI School

This collaborative platform integrates compute resources, tooling, and model capabilities for advanced research in life sciences, materials science, and manufacturing. Yang characterized AI for Science as "a key lever for paradigm innovation in basic research," aligning the initiative with China's broader "Artificial Intelligence+" policy framework.