It took SenseTime merely a year to achieve revenue of 1.2 billion yuan ($170 billion) from generative artificial Intelligence (GAI) despite starting from scratch, and the strategy of developing large AI models and large AI computing infrastructure has played a very important role in the achievement, said the Chinese AI company.

New quality productive forces — a buzz phrase that features high technology, high efficiency and high quality — can be interpreted as great improvements in productivity through enhanced computing capability and algorithms in the AI domain. With that in mind, SenseCore large-scale computing infrastructure is SenseTime's answer to new quality productive forces, said Sheng Shiwei, managing director of the board secretary office with SenseTime.

"In this regard, SenseTime made an extremely important technological transition and upgrade last year, which enabled us to become one of the first Chinese companies to make a profit from generative AI, creating 1.2 billion yuan in revenue," said Sheng.

The managing director said such transactions cannot be achieved without large-scale models and large scale computing infrastructure.

"As we all know, AI models' training and reasoning must be based on fine infrastructures with sufficient computing capacity," he added.

Sheng said the Shanghai-based firm has so far achieved a combined AI computing power of 12 EFLOPS (12 quadrillion floating-point operations per second), with 8.1 EFLOPS deployed in Shanghai's Lin-gang Special Area under China (Shanghai) Pilot Free Trade Zone.

After about 18 months of construction, SenseTime commenced the operation of its next-generation Artificial Intelligence Data Center (AIDC) in Shanghai in 2022. The first phase of the center was designed with a capacity of 3.74 EFLOPS, and its computing power soon exceeded 8.1 EFLOPS in two years.

"We expect capacity will continue to rise in the coming years," said Lin Hai, general manager of the intelligent computing center at SenseTime.

Lin said the open, large-scale and energy-efficient advanced computing infrastructure is so far one of Asia's largest AIDCs. And considering the second phase of the center in the planning stage, the facilities are believed to become a benchmark project in supporting Shanghai's transition into a technological and innovative center with global influence.

Such capability was not created overnight, but driven by forward-looking strategies by the company.

"A few years ago, only companies like SenseTime had the need for training large-scale models. Unable to find large-scale interconnected computing power that could be used for AI in the market, SenseTime decided to build our own large-scale computing infrastructure for high-dimensional research and development, as well as business needs," said Sheng.

Since 2018, SenseTime spent billions of yuan annually in developing large-scale computing infrastructure SenseCore, including pouring 5.6 billion yuan into the first phase of the AIDC in Shanghai.

Thanks to advanced, forward-looking planning amid thriving generative AI development worldwide, SenseTime's niche requirements turned out to be a mainstream trend for the sector.

The global generative AI market size was valued at $43.87 billion in 2023, and the figure is projected to grow to $67.18 billion in 2024. According to a report published on the website of Fortune Business Insights, the market scale is expected to snowball to $967.65 billion by 2032, exhibiting a compound annual growth rate of 39.6 percent during the forecast period between 2024 and 2032.

Having the AI computing infrastructure widely used in various industries by companies ranging from big name tech companies to startups, SenseTime will continue to expand its computing capacity in the future.

"Along with the cost reductions for generative AI, we expect it will be applied in a lot more industries in the future. Therefore, generative AI will become the foundation of productivity and technological development, and that will be phenomenal," said Sheng.

wang_ying@chinadaily.com.cn