Moore Threads Makes a Splashy Debut, Achieves Deep Integration with SenseTime's Algorithm to Jointly Lay the Groundwork for Domestic AI Computing Power

12/08 2025 526

Introduction: The splashy debut of Moore Threads marks a significant milestone in the collaborative advancements within China's AI industry chain. Behind the scenes, SenseTime is paving a vital path for domestic chips, transitioning them from merely being 'usable' to genuinely 'user-friendly' through its innovative 'Computing Power Mall' model.

On December 5th, domestic GPU company Moore Threads (688795.SH) made a successful market entry, with its stock price skyrocketing over 400% on its debut day and its market valuation surpassing 270 billion yuan. This moment undoubtedly stands out as a highlight for China's AI chip sector.

Beyond the excitement in the capital market, it's essential to take a calm look at the underlying industrial logic: the true bottleneck for domestic GPUs is no longer just a competition of raw computing power but rather an incomplete ecosystem. NVIDIA's core strength lies in the CUDA software ecosystem barrier it has constructed over the years.

How can this barrier be broken? The longstanding collaboration between Moore Threads and SenseTime (00020.HK) offers a blueprint: only through deep ecological synergy can a 'software-hardware integration' breakthrough in localization be achieved.

I. Efficient Synergy Between SenseTime's Algorithm and Moore Threads' Computing Power

In response to investor queries, SenseTime officially stated, "Our company has maintained a business partnership with Moore Threads for several years, and our algorithms have been tailored to fit their relevant products. The synergy between the two sides will bolster the market adaptability of our offerings."

It's reported that the two entities have forged efficient collaboration in areas such as large model training and inference technology optimization, software ecosystem support, and core technology research. On one hand, Moore Threads excels with its full-featured GPU, which strikes a balance between AI computing and graphics rendering. The rich algorithm demands and increasingly robust computing power supply from Moore Threads create a solid foundation for cooperation.

On the other hand, SenseTime's computing power requirements are multifaceted. Large model training necessitates extreme floating-point operation capabilities, while applications like AIGC video generation demand powerful rendering capabilities. Moore Threads boasts differentiated strengths in graphics rendering, meeting SenseTime's composite computing power needs for large model operations across various scenarios, and in turn, boosting the efficiency of SenseTime's algorithm deployment.

Consequently, the two sides have established a closed-loop cooperation mechanism of 'demand-supply-optimization,' with their market competitiveness enhancing in tandem.

II. Building the Optimal Testbed for Domestic Computing Power

For Moore Threads, SenseTime's leading 'Day Day New' large model system, SenseCore AI infrastructure, and its broad array of application scenarios constitute an exemplary large-scale scenario verification platform. This provides Moore Threads with the most crucial product validation and endorsement, aiding in the expansion of its industry customer base.

Currently, Moore Threads' MTT S series GPUs have been fully integrated with SenseTime's SenseCore, enabling support for SenseTime's 'Day Day New' multimodal large model system. This marks the inaugural instance of domestic GPUs undergoing rigorous industrial-grade testing in large model training and inference tasks at the scale of hundreds of billions of parameters.

Under the pressure of real-world business demands, SenseTime not only assists Moore Threads in verifying performance and refining products, completing the transition from 'usable' to 'user-friendly,' but also leverages the channel resources of chip companies to reach a wider pool of potential customers. This fosters a virtuous cycle of 'ecosystem expansion - customer growth - performance enhancement,' further solidifying SenseTime's valuation premium as an 'ecosystem leader.'

III. Ecosystem Co-construction: 'SenseTime Computing Power Mall' Fuels China's AI Computing Power Self-Sufficiency

Against the backdrop of constrained global high-end AI computing power availability, establishing a diversified and stable domestic computing power supply chain and fostering ecological synergy is paramount.

The 'SenseCore Computing Power Mall' collaborates with numerous chip companies to forge a domestic AI computing power ecosystem alliance, constructing a comprehensive computing power solution. This lowers the barrier for enterprises to access high-performance computing power, rapidly expands customer reach, increases market penetration, and facilitates the swift integration of AI technology into various industry scenarios.

Taking Moore Threads as an illustration, it became a core member of the 'SenseCore Computing Power Mall' in July this year. Enterprises in fields such as digital twins and embodied intelligence downstream of the platform can utilize the adapted and verified computing power of Moore Threads chips through SenseTime's platform. SenseTime acts as a bridge connecting the upstream and downstream of the industry chain, enabling both computing power providers and users to accelerate market expansion through SenseTime's ecological influence.

The industry widely acknowledges that SenseTime's deep collaboration with multiple domestic AI chip companies to jointly tackle 'bottleneck' technologies and achieve computing power self-sufficiency has propelled the self-sufficiency and industrial development of China's AI computing power. This fosters a cooperative ecosystem characterized by 'technological complementarity, resource sharing, and scenario mutual nourishment,' alleviating market concerns about supply chain shortages.

Solemnly declare: the copyright of this article belongs to the original author. The reprinted article is only for the purpose of spreading more information. If the author's information is marked incorrectly, please contact us immediately to modify or delete it. Thank you.