In the market pattern dominated by giants such as Alibaba Cloud, Tencent Cloud, and Huawei Cloud, Kingsoft Cloud's position is a little awkward.
As an important player in China's cloud computing industry, Kingsoft Cloud has always been highly anticipated by the industry, and is also regarded as the player with the most potential to open up a third growth path, and its in-depth services in the gaming, video and financial industries have shaped its own unique market positioning.
However, recently, Kingsoft Cloud Holdings Co., Ltd., which was pinned on by Lei Jun, announced its 2023 results, and after the decline in annual revenue in 2022, the situation in 2023 has not improved, and the loss has exceeded 2 billion yuan again.
At the same time, major cloud vendors have also begun to reduce prices for core cloud products, with some of Tencent Cloud's product lines falling by up to 40%, which makes Kingsoft Cloud fall into the problem of whether to follow the price reduction, and in the capital market, Kingsoft Cloud's share price has fallen by more than 95% from its highest point, leaving many investors with no money.
Where is the back road of Jinshanyun?
01. The unspeakable hidden nature of entrepreneurial cloud vendors
In the past two years, Kingsoft Cloud's financial performance has attracted widespread attention, especially in a cloud computing market that is considered to be growing rapidly.
According to public data, Kingsoft Cloud's total turnover in 2023 will be 7.047 billion yuan, compared with 8.180 billion yuan in 2022, a decrease of 13.85%. This downward trend not only continues the negative growth that began in 2022 (when it decreased by 9.72% compared to 2021), but further exacerbates market concerns about its competitiveness.
To say that the highlight moment of Kingsoft Cloud comes from 2020.
In 2020, Kingsoft Cloud was successfully listed on the U.S. stock market at a price of $17 per share, and on the day of listing, Kingsoft Cloud's share price rose by more than 20%, and in 2021, Kingsoft Cloud's share price once rose to more than $74 per share, with a market value of more than $17.6 billion.
As the capital market's enthusiasm for cloud computing began to decline, Kingsoft Cloud's revenue also began to decline, and the stock price collapsed all the way, and many investors, including Lei Jun, suffered heavy losses.
As a third-party entrepreneurial cloud vendor, Kingsoft Cloud's echelon has been suppressed layer by layer: first the cloud vendors incubated by technology giants represented by Alibaba Cloud, Tencent Cloud, Baidu Cloud, and Huawei Cloud, and then the second-echelon cloud vendors incubated by telecom enterprises represented by Telecom Tianyi Cloud, China Unicom Cloud, and Mobile Cloud.
In the third echelon, traditional software and hardware accessory manufacturers such as Yonyou Cloud, Kingdee Cloud, and Inspur Cloud have entered the cloud service market at the right time by virtue of their past hardware strength, but on the whole, they are not particularly successful, and the sense of presence in the related market is relatively weak.
Kingsoft Cloud's position is in the echelon of entrepreneurial third-party independent cloud vendors that are not embarrassed or embarrassed, and Kingsoft Cloud coexists with Qingyun, Ucloud, etc., such cloud vendors have strong technological innovation, but suffer from the suppression of strong competitors, and are generally facing a more urgent survival crisis.
In April last year, Alibaba Cloud took the lead in launching the largest price reduction in the market, the price of core products was lowered by 15%-50% across the board, and the price of storage products was reduced by up to 50%.
At present, cloud computing is on the rise in the market, and continuous revenue growth is regarded as the cornerstone of enterprise development.
However, Kingsoft Cloud's declining revenue for two consecutive years has undoubtedly exposed the challenges it has encountered in market competition, and Kingsoft Cloud is powerless in the face of fierce competition from Alibaba Cloud, Tencent Cloud, Huawei Cloud and other two-echelon giants.
In fact, although China's cloud computing market is growing, but the overall is facing a certain development dilemma, after experiencing the rush to the cloud, most of the enterprises that have completed the road to the cloud have completed the goal, and the rest are thinking about whether the revenue in the later stage can cover the cost and other factors.
As a result, the giants have fallen into sluggish growth, and the life of the third-party entrepreneurial cloud vendors with the weakest survival roots is even more difficult, and the market cold current is the first to be transmitted to them.
In contrast, other large manufacturers can rely on their own business to support cloud computing, but entrepreneurial cloud vendors can only watch the investment of tens of billions of data centers at every turn, and they are discouraged.
What should the future of "Kingsoft Clouds" be?
02. Kingsoft Cloud, which is "de-milletized", enters AI
For this kind of cloud vendors, Kingsoft Cloud urgently needs to take more effective countermeasures.
First of all, in terms of cost, in 2023, Kingsoft Cloud has made significant progress in cost control, and its operating costs have decreased by 20.05% year-on-year to 6.197 billion yuan, which has had a significant impact on the improvement of gross profit margin, and gross profit has increased by 97.67% year-on-year to 850 million yuan.
The effect of this cost control directly mitigated the negative impact of the revenue decline and laid the foundation for the company's future earnings improvement. In particular, the decline in cost of sales and administrative expenses exceeded the decline in revenue, indicating that the company's adjustment of business strategy began to achieve initial results.
In addition, Kingsoft Cloud is also actively seeking transformation.
In the past two years, ChatGPT has become popular, and major cloud service providers at home and abroad have accelerated their speed in AI layout, such as Baidu Cloud under the strategy of "cloud intelligence integration", which has added a lot of color to its products, and Alibaba Cloud has also begun to access the Tongyi Qianwen AI model, and the vigorous war has kicked off.
Of course, Kingsoft Cloud is no exception, and it will release 3 generative AI large model images in 2023.
The three large models are ChatGLM-6B, Stable Diffusion and Alpaca-Lora, all of which are provided free of charge, and customers only need to pay for the GPU cloud server of their choice, and can complete the fine-tuning of the generative AI large model or the construction of the inference environment in a few simple steps, enabling the productivity needs of various fields.
Among them, ChatGLM-6B has the most outstanding performance and is most likely to be widely used in major industries.
It is reported that ChatGLM-6B is an entry-level bilingual open source model with 6.2 billion parameters released by Tsinghua University Open Source.
In addition, ChatGLM-6B also trains 1T tokens on 1:1 ratio of Chinese and English materials, that is, compared with foreign models, ChatGLM-6B has better Chinese dialogue capabilities, and is more suitable for domestic enterprises and developers to explore landing application directions.
However, it is very difficult for Kingsoft Cloud to train a model of the same scale in the future, and it is important to know that a large model product of Kingsoft Cloud's size has not entered the scope of the "2024 China Large Model Capability Assessment" by Frost & Sullivan, a global growth consulting company.
The research and development of large models requires a large amount of funds as the basis for research and development, and in the early stage, it even takes 2-300 million yuan to train a new large model.
Nowadays, with the increase in the price of computing power, the cost of training a large model is even as high as 500 million yuan.
Compared with the 2.255 billion yuan in cash and cash equivalents of Kingsoft Cloud by the end of 2023, choosing to continue to increase the research and development of AI large models will obviously further expand the loss.
Judging from the equity composition and enterprise user composition of Kingsoft Cloud, Xiaomi has invested in Kingsoft Cloud many times, accounting for 12.3% of Kingsoft Cloud's shares, and is the main user of Kingsoft Cloud. From 2019 to 2021, related party transactions between Xiaomi and Kingsoft Cloud accounted for 14.4%, 10.0%, and 8.3% of Kingsoft Cloud's revenue, respectively. Lei Jun, founder, chairman and CEO of Xiaomi, is also the chairman of the board of directors of Kingsoft Cloud.
"At the group level, what bets on the next ten years is Kingsoft Cloud. Lei Jun said this in an interview in June 2013.
Since the beginning of this year, Kingsoft Cloud has decided to bet on the TO B market, Kingsoft Cloud's "de-milletization" has begun to become more and more obvious.
In the early days of Kingsoft Cloud's development, Xiaomi Group contributed more than 80% of Kingsoft Cloud's revenue, and now, this contribution is declining year by year - which indicates that Kingsoft Cloud is accelerating its "de-Xiaomi" and embracing a broader enterprise-level market.
In the words of Cui Baoqiu, vice president of Xiaomi Group: Kingsoft Cloud and our Xiaomi are brothers, Xiaomi's Internet business has fully benefited from big data today, and will continue to take the road of data-driven and AI-enabled in the future.
Kingsoft Cloud plays an important role in the Xiaomi system, but what Xiaomi really empowers Kingsoft Cloud, so far the most noteworthy is only smart home.
In this context, for Kingsoft Cloud, perhaps the C-end will be more cost-effective than continuing to develop AI large models.
Reference:
The stock price fell by more than 95%, and Kingsoft Cloud bet that AI will be difficult to break through - the forefront of entrepreneurship
Kingsoft Cloud's Q4 revenue was 1.722 billion yuan, and technological innovation drove the growth of both revenue scale and profitability - Shikr shared
The shrinking "Kingsoft clouds", the collective dilemma of independent third-party cloud providers - odd and even