Recently, Dongfang Supercomputing signed an agreement with SenseTime, the leader of the "AI Four Little Dragons". Based on the national integrated computing power platform and cross-regional clusters built by Dongfang Supercomputing, it will provide artificial intelligence computing power infrastructure and platform services for SenseTime. The project service period is three years, and the project amount is approximately 89 million yuan.

As the "first AI stock," SenseTime has still incurred losses three years after its listing.

In 2023, the company's overall operating revenue was 3.406 billion yuan, a year-on-year decrease of 10.6%; the annual loss was 6.494 billion yuan, an increase of 6.6% year-on-year; EBITDA (earnings before interest, taxes, depreciation, and amortization) was a loss of 5.45 billion yuan, a year-on-year narrowing of 2.2%.

Among them, generative AI revenue reached 1.2 billion yuan, accounting for 35% of the group's total revenue, achieving a growth of nearly 200%.

With a high proportion of generative AI, the corresponding proportion of traditional AI revenue has also decreased.

In 2023, SenseTime's traditional AI revenue was 1.838 billion yuan, a year-on-year decrease of 41%. Among them, the proportion of smart city revenue decreased from 30% to less than 10%.

In fact, as early as 2018, SenseTime began to invest more resources into the SenseCore large-scale device every year, while dedicating itself to the research and development of AI large models, and built the Lingang AIDC (Artificial Intelligence Computing Center) in Shanghai. Two years later, SenseTime officially launched a new AI infrastructure - the SenseCore large-scale device.

At that time, the battle of hundreds of models had not yet begun. After the advent of ChatGPT, the demand for intelligent computing power from technology companies including SenseTime has subsequently increased dramatically.

In 2023, the large language model "SenseChat," the "SenseNova" large model system, and the Chinese medical and health language large model "Dayi" were successively launched.

It can be seen that SenseTime Technology has taken on big models.

Earlier this year, SenseTime Technology jointly promoted the construction of the Qianhai Shenzhen-Hong Kong AI Computing Center with the Qianhai Administration and the Hong Kong Science and Technology Parks Corporation. The Qianhai Science and Technology Innovation Group and SenseTime Technology jointly invested in the construction, with the first phase of computing power reaching 500 Petaflops, aiming to build the largest and most advanced computing center in the Greater Bay Area.

Previously, SenseTime Technology transformed its three new business segments into generative AI, traditional AI, and smart cars.

Can it emerge from the trough under the leadership of AIGC this time?

It's no surprise that SenseTime Technology chose Dongfang Supercomputing.

As a wholly-owned subsidiary of Dongfang Materials, Dongfang Supercomputing's core technical team is led by Chen Guoliang, an academician of the Chinese Academy of Sciences, a leading expert in parallel algorithms and high-performance computing, and the chief scientist of Dongfang Supercomputing. It successfully built the first national integrated computing power platform, the "Dongfang Supercomputing·Galaxy" national integrated computing power platform, in its early days.

Through the upgrade and aggregation of cross-regional computing power facilities nationwide, based on globally leading computing power scheduling technology, this platform solves problems such as uneven allocation of cross-regional resources and cross-architecture "computing power chimneys." It quickly and flexibly matches users with nearby idle computing power of the same service level, helping users save on computing power expenses, lower the usage threshold, promote remote, heterogeneous, energy-saving, and balanced circulation of computing power, and rapidly promote the application process of artificial intelligence in various industries.

Public information shows that Dongfang Supercomputing has successfully created a "national integrated supercomputing cluster" in collaboration with the first batch of 13 data centers nationwide, and will provide "one-stop" overall supply services for computing power resources to cross-regional users in the future.

On the application side, Dongfang Supercomputing has reached in-depth cooperation with leading enterprises such as NVIDIA, Intel, Huawei, Moor Threads, Inspur, and SenseTime, as well as top research institutes such as Pengcheng Laboratory, the Peking University Institute of Artificial Intelligence, and the University of Science and Technology of China.

From B-end to E-end and then to C-end, Dongfang Supercomputing uses technological capabilities and ecological aggregation power to promote the "first mile" of the computing power industry to achieve overtaking.

Is this strong collaboration between the two parties the antidote for SenseTime Technology? Or is SenseTime Technology Dongfang Supercomputing's new round of layout?