SenseTime: Accelerating Large-Scale Infrastructure and Model Deployments, Positioning Generative AI as Core Engine

2024-03-29

Recently, SenseTime successfully held a performance exchange meeting. Xu Li, the chairman and CEO of the company, revealed at the meeting that in 2023, the company has fully implemented the strategic layout of large-scale devices and models. In terms of computing power, the total operating power of large-scale devices has exceeded 12,000P (Petaflops), doubling compared to the beginning of the year. At the same time, the number of GPUs launched by the company has also increased to 45,000.



In the field of large-scale models, SenseTime's daily new large-scale model performance has been significantly improved, and it has the ability to train large-scale models with tens of thousands of cards and participants. In terms of basic models, multimodal processing, programming and tool invocation, million-word lossless context processing, and terminal small models, SenseTime has reached the leading level in China. The new model and service business model have also attracted a large number of new customers for the company, making SenseTime one of the first eight models in China to pass the Interim Measures for the Administration of Generated Artificial Intelligence Services, and can provide services widely to the public and enterprises.

Xu Li also pointed out that under the new business classification, the company focuses on the development of generative AI business while maintaining the advantages of traditional AI business and intelligent automotive business. The data shows that the revenue of generative AI business reached 1.18 billion yuan last year, a year-on-year increase of 200%, and the proportion of total revenue increased to 35%. The revenue of traditional AI business was 1.84 billion yuan, a year-on-year decrease of 41%, and the proportion of total revenue also decreased from 82% in the previous year to 54%. Among them, the proportion of revenue from smart city business has significantly decreased, reflecting the company's significantly reduced dependence on smart cities.

In addition, the Jueying automotive business has also shown a steady growth trend. Last year, the revenue of this business reached 380 million yuan, a year-on-year increase of 31%, accounting for 11% of total revenue. Generative AI not only brings technological innovation and innovation to SenseTime, but also replaces smart city business as the core business of the company. It is expected that this business will continue to maintain high-speed growth in 2024.

The rapid growth of SenseTime's generative AI business is due to the widespread demand for large-scale model training and inference in various industries, marking the beginning of a new cycle of investment in Chinese hard technology. By deeply integrating generative AI capabilities into various business levels, SenseTime is continuously winning new customers and comprehensively improving efficiency and productivity.

Xu Bing, Executive Director and Board Secretary, said when looking forward to 2024, the company will move forward around three strategic goals. First, the company will continue to maintain its leading position in technology, further leverage the synergies of large-scale devices and models, expand computing power, and enhance the comprehensive service capabilities of infrastructure and large-scale models. At the same time, the company will continue to invest in the iterative upgrade of the daily new model system to ensure that the model capabilities remain at the forefront of the industry.

Second, the company will be committed to promoting business growth by accelerating the commercialization process of generative AI business to provide customers with cost-effective solutions. Finally, the company will strive to achieve the goal of core business profitability by optimizing operational efficiency, incubating and spinning off non-core businesses, and concentrating resources on generative AI business to improve cash flow and reduce losses. With its strong capabilities in large-scale devices, models, and cloud collaboration, SenseTime is confident about the future. Despite the challenges of the transformation period, the company will still occupy a favorable position and achieve long-term leapfrog development in the fierce competition of the AI 2.0 era.