news

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

sensetime can continue to make rapid progress in the generative ai business, which is inseparable from the years of layout of the underlying computing power. in 2018, sensetime built its own computing power. now it has achieved the second place in the market share of large models and the top three in the market share of aidc, all of which are achieved step by step. however, the market is more concerned about when sensetime will be profitable, and this is the biggest problem.

text｜han xiang

sensetime, "asia's no. 1 ai stock", released its 2024 interim report showing continued losses.

data shows that sensetime achieved revenue of 1.74 billion yuan in the first half of the year, a year-on-year increase of 21.4%; the net loss attributable to the parent was 2.477 billion yuan, a year-on-year narrowing of 21.2%.

after years of development, sensetime’s profitability problem has not been effectively solved.however, there are still bright spots in this interim report. the revenue of the generative ai sector and jueying intelligent automobile business has increased significantly. among them, the revenue of generative ai was 1.05 billion yuan, a year-on-year increase of 255.7%, accounting for 60.4% of the total revenue, becoming the core business of jueying intelligent automobile. the business revenue of jueying intelligent automobile increased by 100.4% to 168 million yuan.

the most obvious decline in performance was in the traditional ai business, with revenue halved to 520 million yuan, and the proportion of revenue dropped from 73.5% to 29.9%, giving up the core business position. overseas business is also growing steadily, but currently accounts for only 18.5% of revenue.

regarding the decline of traditional ai business, "digital intelligence research society" believes thatthe bulk of traditional ai business is smart city business, and most of its customers are government and enterprises.due to changes in the macroeconomic situation, it has become relatively difficult to collect payments from these customers. to a certain extent, the provision for impairment of accounts receivable from automobile manufacturing and trade has increased. the net impairment loss on financial assets and contract assets in the first half of the year increased from 397 million yuan in the same period last year to 588 million yuan.

to a certain extent, this has also forced sensetime to transform to generative ai. among them, the leap-forward improvement of the "sensenova" big model capabilities and the rapid decline in inference costs have driven a substantial increase in sensetime's generative ai business revenue. currently, more than 3,000 leading companies in the industry are using sensetime's big models and intelligent computing services, and the overall call volume of the "sensenova" big model has increased by 400%.

in fact, as early as in its strategic goals for 2024, sensetime mentioned the need to drive profitability of its core businesses and concentrate resources on generative ai businesses to improve cash flow and reduce losses.

at present, only one of the sentences has been achieved, and profitability of the core business is still a long way off.

lost 52.8 billion yuan in 6 and a half years

for a long time, sensetime, which has been crowned with many halos such as asia's leader, one of the four ai dragons, and the first ai stock, has been running on the road of losses.

"numerical research society" found that from 2018 to 2023, sensetime's net losses attributable to shareholders of the parent company were 3.433 billion yuan, 4.968 billion yuan, 12.158 billion yuan, 17.177 billion yuan, 6.093 billion yuan and 6.495 billion yuan, respectively, with a cumulative loss of 50.324 billion yuan in six years.

combined with the loss of 2.477 billion yuan in the first half of 2024, sensetime has lost 52.8 billion yuan in six and a half years.

in the secondary market, sensetime closed at hk$1.12 per share on september 4, with a total market value of only hk$39.359 billion. compared with the high point of hk$9.7 per share and a market value of more than hk$320 billion, only a fraction of that is left now.

for current generative ai companies, losses are a very common phenomenon. on the one hand, there is high r&d investment, and on the other hand, commercialization progress is slow. "digital intelligence research society" found that from 2018 to 2023, sensetime invested 849 million yuan, 1.916 billion yuan, 2.454 billion yuan, 3.614 billion yuan, 4.014 billion yuan and 3.466 billion yuan respectively, plus 1.892 billion yuan invested in the first half of 2024. in 6 and a half years, sensetime has invested a total of 18.205 billion yuan in r&d expenses.

it is worth noting that while sensetime continues to increase its investment in research and development, its employee count has been reduced.at the end of 2022, sensetime had a total of 5,098 employees, which dropped to 4,531 by the end of 2023, a reduction of more than 500 employees in one year. this also basically confirms the rumors of layoffs.

betting on generative ai

sensetime did not initially focus on generative ai business.

before entering the capital market in 2021, sensetime was china's largest computer vision software company, and its business at the time was relatively fragmented, covering four major sectors: smart business, smart cities, smart cars, and smart life.

but when it went public in 2021, the concept of the metaverse was just right. at that time, sensetime mentioned the "metaverse" many times in its annual report, such as "we have also established a digital space business group to fully integrate our advantages in ai perception, ar and mr technology, computing power and customer ecology, and are committed to building an industry-leading metaverse enabling platform."

at this time, sensetime was immersed in enjoying the bright future of the metaverse.

in 2022, the craze for the metaverse gradually faded. at the end of the year, openai on the east coast of the pacific fired the first shot in artificial intelligence, big models, and generative ai. in 2023, sensetime did not mention the metaverse at all, but directly embraced big models and generative ai. at the same time, it integrated the original four business segments into three major segments: generative ai, traditional ai, and smart cars. among them, the generative ai business mainly includes big models and intelligent computing services.

by the end of 2023, the revenue from generative ai increased by 199.9% to 1.184 billion yuan, and the proportion of the group's revenue increased from 10.4% to 34.8%. by the mid-year report of 2024, the generative ai business ushered in a big explosion, and the proportion of revenue increased to 60.4%, becoming the core of the business. this shows that the training and reasoning of generative ai have shown a strong growth trend.

amid the boom in generative ai business, sensetime has become the second largest market share in china's large model market with a market share of 16%, and its aidc (automatic identification and data collection) service ranks among the top three with a market share of 15.4%.

in the view of "digital intelligence research society", sensetime's rapid growth in generative ai business is inseparable from its years of layout in underlying computing power. in 2018, sensetime built the sensecore ai device based on its own computing power, and later built the aidc intelligent computing center in lingang, shanghai. in recent years, it has expanded new computing nodes in guangzhou, shenzhen and other places.

after building its own computing power, sensetime has also updated its large-scale devices based on sensecore in terms of platform and algorithm, thus achieving the goal of providing solutions for thousands of industries. sensetime has also become the first manufacturer in china to release a large model with more than 100 billion parameters.

at present, it is still in the early stage of the outbreak of generative ai, and it is not easy for sensetime to achieve such results. on the one hand, it shows that the layout over the years has paid off. but at the same time, the continuous losses also trouble the market's confidence in sensetime.