share_log

商汤-W(00020.HK)2023年中报点评:战略调整业务结构 布局生成式AI广阔市场

Shangtang-W (00020.HK) 2023 Interim Report Review: Strategic adjustment of business structure layout, broad market for generative AI

中信證券 ·  Aug 31, 2023 19:46

2023H1, affected by the external environment and its own strategic adjustments, the company's revenue structure was adjusted, and gross profit was under pressure. However, thanks to the obvious optimization of R&D expenses, the company's net loss after the adjustment narrowed year-on-year. The company's smart car business achieved major technological breakthroughs in the first half of the year. The application of relevant cutting-edge technology was imminent, and related businesses developed rapidly. At the same time, the company has invested heavily in the development of generative AI. It has a solid accumulation of computing power, model, and data levels at the bottom of the AI model, and has launched rich application products, laying a solid foundation for the accelerated growth of future performance. The company was given 16xPS in 2023, corresponding to a market capitalization of HK$73.9 billion and a target price of HK$2.2, maintaining an “gain” rating.

Matters: The company released its 2023 mid-year report. In the first half of the year, the company achieved operating income of 1,433 billion yuan, a year-on-year increase of 1.3%, a gross profit margin of 45.3%, a year-on-year decrease of 20.7 pcts, and an adjusted net loss of 2,393 billion yuan, a year-on-year narrowing of 6.7%.

The revenue structure has been adjusted, and the increase in sales costs has put pressure on gross profit. On the revenue side, the company achieved revenue of 1,433 billion yuan (YoY +1.3%) in the first half of the year. Among them, driven by new business opportunities brought by generative AI, smart business revenue was 853 million yuan, up 50.2% year on year, smart living revenue was 312 million yuan, up 6.7% year on year; smart city revenue was 184 million yuan, down 57.7% year on year, mainly due to the company's strategic adjustments focusing on leading customers; smart car revenue was 84 million yuan, down 30.5% year on year. The company's overall revenue remained stable in the first half of the year, but structurally, the share of smart commerce increased markedly, reaching 59.5% (YoY +19.3 pcts).

On the gross margin side, due to increased depreciation and amortization due to hardware costs and AIDC investment, the company's sales cost increased markedly in the first half of the year, reaching 784 million yuan (YoY +63.0%). Cost growth has put pressure on gross margin, with the company's gross profit margin of 45.4% (YoY -20.7%) in the first half of the year.

R&D expenses were clearly optimized, and net loss gradually narrowed after the adjustment. 23H1's sales/management/R&D expenses were 418/797/1,783 million yuan respectively, and the corresponding cost rates were 29.16%/55.58%/124.43% (YOY+0.75/+1.27/ -19.37pcts), respectively. The overall stability and R&D expenses were significantly reduced, mainly due to the reduction in the company's personnel expenses and the reduction of the company's AIDC investment and operation service expenses. Under cost-side optimization and pressure on the gross margin side, the company achieved an adjusted net loss of 2,393 billion yuan in the first half of the year, a decrease of 6.7% over the previous year.

Lead the advancement of big model technology and lay out a broad market for generative AI. In the first half of the year, the company followed the wave of generative artificial intelligence brought about by big models and vigorously expanded its business layout. 1) On the computing power side, the company continued to upgrade SenseCore AI, an intelligent computing device. The number of GPUs launched was increased by roughly 20% to 30,000, and the computing power scale reached 6 exaFlops. In the first half of the year, there were more than 1000 billion parameter model training. Abundant computing power strongly supported the development and iteration of the company's large-scale model technology. 2) On the algorithm side, the company invested 10,000 GPUs in self-developed large-scale model development in the first half of the year, launched the “Japan-Japan New” big model, and continued to rapidly iterate to lead the development of the domestic general model. Following the launch of InternLM against GPT-3.5 in June, the latest INTERNLM-130b model completed training in August. According to estimates given in the company's report, the model's capabilities completely surpassed GPT3.5-Turbo, and ranked second in the world in terms of performance in some data sets and tasks. 3) On the application side, based on the new Japanese and Japanese models and the base capabilities of large devices, in April 2023, the company released six generative AI products: discussion, second painting, like a shadow, Qiongyu, grid, and bright eyes. With the rapid iteration of the model, front-end products are also evolving at the same time. Taking ChatGPT-like product discussions (senseChat) as an example, the company launched 1.0 and 2.0 versions in April and July, respectively, and is expected to upgrade to version 3.0 within September. Other products will maintain a similar iteration rate. With full support from technology, the company's generative AI-related revenue increased 670.4% year-on-year in the first half of the year, accounting for 20.3% of total revenue. Among them, by providing new AI products, the number of smart lifestyle customers reached 270, an increase of 47.5% year on year; by meeting the needs of the B-side industry model, smart business revenue increased 50.2% year on year, and the customer unit price increased by 59.9%.

Major breakthroughs in core technology have accelerated the development of smart cars. In 2023, the company achieved a major breakthrough in intelligent driving technology, and the UniAd algorithm won the CVPR Best Paper at the world's top academic conference, becoming the first best paper on the subject of autonomous driving in CVPR history. Based on the Transformer framework, the technology integrates autonomous driving detection, tracking, mapping, prediction, and planning under a unified framework to optimize the overall effect. The company expects the relevant algorithm 23Q4 to complete mass production preparations and continue to lead the development of the industry. As the company's cutting-edge technology continues to transform, related business revenue is expected to show a rapid growth trend. In the first half of the year, the number of the company's smart car customers reached 34, an increase of 70% over the previous year; the company's mass production and delivery of 390,000 units, the gross bicycle profit increased 29% year over year, and the business scale increased 573% year on year; the number of fixed targets was added to more than 5 million units, with a cumulative total of more than 160 fixed-point models.

However, due to changes in the cooperation model, R&D revenue declined, and overall sector revenue declined by 30.5%. In the future, with the rapid development of front-end products, the company's smart driving business revenue is expected to recover at an accelerated pace, becoming another important support for future development.

Risk factors: macroeconomic recovery falls short of expectations; company project delivery falls short of expectations; development of artificial intelligence falls short of expectations; progress of the company's technology iteration falls short of expectations.

Profit forecasting, valuation and rating: As a leading domestic company in the field of generative AI, the company has a solid accumulation of computing power, models, and data at the bottom of the AI model, and has launched a variety of application products, laying a solid foundation for the accelerated growth of future performance. In the AI 2.0 era represented by big models, AI big model technology application industries and scenarios will tend to be diversified and intelligent, and the acceleration of widespread application and commercialization will accelerate the recovery of the company's future performance. However, considering the macroeconomic recovery situation and the company's short-term operating situation, and considering the strategic changes in the company's main business, the company's 2023-2025 revenue forecast was adjusted to 42.93/53.99/6.777 billion yuan (the original forecast was 5146/69.30/9.298 billion yuan); Yuncong Technology, Haitian Ruisheng, and Geling Deep Hitomi were selected as comparable companies, corresponding to the 2023 16XPS according to the latest closing price and WIND's consistent expectations, considering that the company has extensive development as a leading generative AI company Prospects, however, due to business adjustments in the smart city and smart car scenarios, short-term results are uncertain. The company was judiciously given 16 xPS in 2023, corresponding to a market value of HK$73.9 billion and a target price of HK$2.2 billion, maintaining an “gain” rating.

The translation is provided by third-party software.


The above content is for informational or educational purposes only and does not constitute any investment advice related to Futu. Although we strive to ensure the truthfulness, accuracy, and originality of all such content, we cannot guarantee it.
    Write a comment