AI is on fire, cloud profits have been overturned

Original by Shenrancaijing

Author | Wang Min

Editor | Jin Yufan

As the financial reporting season approaches, the financial reports of BAT three companies come one after another, and cloud business has become a highlight.

In the past, cloud vendors have been experiencing significant losses, and whether they can continue to make profits has been an important indicator for the industry to measure the strength of cloud vendors in the past two years. In this quarter, Baidu AI Cloud announced quarterly profit for the first time, becoming the second big factory cloud after Alibaba Cloud. Alibaba Cloud has achieved profitability for two consecutive years and aims to achieve a spin off listing within one year.

According to an IDC report, in the second half of 2022, the top 5 companies in the Chinese public cloud market were Alibaba Cloud, Huawei Cloud, China Telecom Tianyi Cloud, Tencent Cloud, and Amazon AWS, with market shares of 31.9%, 12.1%, 10.3%, 9.9%, and 8.6%, respectively. By the first half of this year, the revenue scale of Huawei Cloud and Operator Cloud had not both been reported in the financial statements. However, some industry insiders have expressed to Shenran that the competition in the cloud computing market is more intense, and the market share of Huawei Cloud and Tianyi Cloud is still increasing.

However, the pattern of the top three major cloud companies is relatively stable. According to the financial report, during the first quarter, Alibaba Cloud's scale remained stable at the top, with revenue of 18.58 billion yuan, accounting for about 9% of the total revenue, but declining by 2.1% year-on-year; Tencent executives pointed out that Tencent Cloud's revenue does not exceed 10%, which means that Tencent Cloud's revenue in the first quarter will not exceed 15 billion yuan; The revenue of Baidu AI Cloud is 4.2 billion yuan, accounting for 13% of the total revenue.

In addition, the "Hundred Model War" in China has become increasingly fierce, and the outside world is more concerned about how the wave of AI big models will transform the cloud computing market.

Among the top three cloud vendors in the world, Microsoft and Google have both shifted towards cloud service providers that emphasize AI capabilities more. This is especially true of leading cloud service providers in China. Baidu and Alibaba have taken the lead in rolling up and releasing large models. Tencent and ByteDance have also participated in the infrastructure level of model training. Manufacturers not only strive to be the "shovelers" of the big model era, but also the "gold diggers" of the big model era.

From the recent actions of cloud vendors, it can be seen that they have entered a price war. In late April, Alibaba Cloud announced the largest price cut in history and took the lead in launching a new round of price war. In mid May, the night Baidu AI Cloud announced profits, Tencent Cloud and Mobile Cloud announced price cuts successively and participated. Undoubtedly, manufacturers are vying for market share and preparing for the competition in the big model era.

In the past 2022, the global cloud computing market has been experiencing a slowdown in growth. In the domestic cloud computing market, due to the peak of internet dividends, large cloud companies are facing growth bottlenecks, and operator cloud has become the "dark horse" of growth. Entering 2023, the overall slowdown in market growth is difficult to change in the short term, but what remains unchanged is that manufacturers are seizing various opportunities to seek growth.

Robin Lee, the founder of Baidu, has regarded the big model as the existence of changing the game rules of the cloud computing market in the future. After Zhang Yong, Chairman and CEO of Alibaba's Board of Directors, served as the CEO of Alibaba Cloud's intelligent business group, he has taken the big price cut, joint partners to promote the implementation of the big model, and product integration as "three board axes". The operator Cloud will not let go of the new growth curve of cloud business.

Compared to the "conservative" trend in 2022, cloud vendors are "running" and this market is bound to face a fierce battle.

Cloud vendor volume model: both as a gold digger and as a sales shovel

In 2023, cloud vendors are one of the most direct beneficiaries of the AI big model wave.

Globally, Microsoft's first quarter financial report revealed that the Azure OpenAI service has 2500 customers, a 10 fold increase compared to the previous quarter.

In China, according to Baidu's financial report, after the release of "ERNIE Bot" in mid March, the AI model has increased the sales leads of Baidu AI Cloud by four times. As for the specific increase in revenue, it may be reflected in the financial reports of the following quarters.

The logic behind cloud vendors being able to reap the dividends of AI big models is simple,On the one hand, cloud vendors have become "gold diggers" in the era of big models by providing big model services

Robin Lee pointed out when ERNIE Bot was released that in the era of big model, the mainstream business model of new cloud computing companies has changed from IaaS (infrastructure as a service) to MaaS. Nail CEO Ye Jun has also stated that SaaS (Software as a Service), PaaS (Platform as a Service), and IaaS, all three layers, may become prerequisites for data training. In the future, these three layers will be bypassed and MaaS will be directly called by the AI layer.

On the other hand, cloud vendors are also the "selling shovels" of the big model era

Zhang Yong stated at the Alibaba Cloud Summit that big models are a comprehensive competition of "AI+cloud computing". The research and development of large models with over a trillion parameters includes complex systematic engineering in various fields such as underlying computing power, networks, big data, machine learning, etc., and requires support from a large-scale AI infrastructure.

In order to seize this growth point, in the past few months, large cloud companies have been accelerating their layout with their own ideas.

Among them, Baidu and Alibaba are the most urgent, striving to become both gold miners and those who want to sell and shovel people. After the release of ERNIE Bot, Baidu said that it would open API interfaces directly to enterprise customers in the future and provide them with large model services directly. It is reported that ERNIE Bot has completed 4 technical iterations since its release in January, which has increased reasoning efficiency by 10 times compared with the original version, doubled the utilization rate of model computing power, and achieved significant optimization in model performance and cost.

In May of this year, Baidu also brought the Wenxin Qianfan Big Model Platform, where enterprises can develop their own exclusive big models based on any open source or closed source big models.

Similarly, Alibaba directly released the Tongyi Qianwen Big Model in April, announcing that all businesses will be connected to the big model in the future and will provide model services externally. Before releasing the big model, Zhang Yong proposed to provide computational support for the industry's big model.

Among cloud vendors,Tencent and Volcano Engine have not yet directly released the latest AI big model, but have chosen to become "shovel sellers" and provide services from the infrastructure level

Zhang Xiaorong, the director of the Deep Science and Technology Research Institute, told Shenran that AI computing relies on GPU chip computing power, and even Dachang Cloud is facing a situation where production capacity is in short supply. This is because Dachang Cloud has already laid out its layout in the AI computing field, and operator cloud is relatively weak.

The operator cloud will obviously not give up the opportunity of AI big models. From the financial reports of the three major operators in the first quarter, it can be seen that cloud business is becoming a new driving force for growth, and the growth rate of related business sectors in the first quarter is around 20%.

In the wave of AI big models, the role of computing power as infrastructure has become increasingly prominent, and the three major operators have previously expressed their willingness to increase investment in computing power. China Unicom's network investment in 2023 is expected to grow by more than 20% year-on-year, no less than 14.6 billion yuan. China Telecom's digital investment has increased to 38 billion yuan, including 19.5 billion yuan in computing power investment. Zhang Yi, CEO of iMedia Consulting, believes that "the AI big model is expected to open up a second spring for the massive computing power demand in the cloud computing market

Ma Tongjuan, a partner of Locke Capital, pointed out that the AI big model is intensifying competition in the cloud computing market, and also promoting cloud vendors to strengthen innovation and improve service quality. In the future, with the continuous development and application of AI big models, it may bring more increment to the cloud computing market, with more efficient resource utilization, more intelligent service management, and more accurate data analysis, all of which are typical cases.

The price war has started

As the competition for big models intensifies, cloud vendors have also launched a new round of price wars.

Alibaba Cloud was the first to start the war. On April 26, it announced that the price of core products would be reduced by 15% to 50% across the board, involving multiple core cloud products such as cloud storage, cloud network, database, and cloud security, with a maximum reduction of 50% for storage products. This price adjustment is known as the "largest price reduction in Alibaba Cloud's history".

20 days later, Tencent Cloud and Mobile Cloud joined the price war simultaneously. Tencent Cloud can achieve a maximum price reduction of 40%, with product lines including cloud networks, cloud servers, databases, cloud security, etc. The maximum price reduction for some mobile cloud products can reach 60%, covering more than ten products in six major sectors including computing, storage, databases, secure backup, and big data.

Three of the top domestic cloud manufacturers have joined this price war.

Industry insiders have analyzed that Alibaba Cloud may have been the first to launch a price war due to its intention to split and list. After the release of the first quarter financial report, Alibaba CEO Zhang Yong released a full staff letter stating that he plans to completely spin off and list Alibaba Cloud Intelligent Group from Alibaba in the next 12 months, and Alibaba Cloud will also introduce external strategic investors.

In the past few years, although Alibaba Cloud has maintained a leading position in China, its market share has been declining, from 36.7% in the second half of 2021 to 31.9% in the second half of 2022.

The declining market share and the pressure of self financing after the spin off may be the reasons for Alibaba Cloud's shift from defensive to offensive.

In addition, part of the reason is that technology is iterating, and manufacturers are increasing their own research to improve performance and reduce costs. Taking Alibaba Cloud as an example, its self-developed chip has significantly improved energy efficiency. The performance of the Yitian 710 exceeds the industry benchmark by 20%, and the energy efficiency ratio has increased by more than 50%. Currently, it has been widely applied.

Regarding the price reduction, Tencent executives stated at the financial report that from an industry perspective, the price reduction is due to a decrease in investment costs. However, Tencent's price adjustment is within a few point range and the impact is not significant.

Ma Tongjuan believes that Alibaba Cloud's big strategy is to sacrifice gross profit for a larger market share in order to gain more advantages in fierce competition, in order to further consolidate its leading position. Tencent Cloud's strategy in recent years has been to focus on self-developed business and improve business performance. If a low-priced strategy is used to cope, it will have an impact on its business performance, but the overall strategy is currently unclear.

Zhang Xiaorong told Shenran that the cloud computing industry has suffered long-term losses, with only a few large companies making profits. When the price war starts, the share of large cloud factories may increase, but small cloud factories cannot afford it. As a result, they may have to close or transform, and the market concentration ratio will further increase.

This is not the first time that the cloud computing market has engaged in a price war. In the past, cloud manufacturers often had 0 yuan or 1 yuan bids to win industry benchmark cases, and they often traded losses for scale.

However, this price war is due to the fact that manufacturers are seizing market share, but it is also considered that a new round of arms war is being waged by manufacturers in the era of big model. Expanding the user scale and base, and increasing market penetration are also adding ammunition for new market competition.

In the era of big models, reassessing cloud computing

In 2023, the cloud computing market is still full of uncertainty, but it is certain that the domestic market will bring a new round of competition. Looking rationally, whether globally or domestically,The slowdown in the growth rate of the cloud computing market is a reality that manufacturers need to face.

According to a report by Canalys, in the global market, global cloud infrastructure service expenditures increased by 19% in the first quarter of 2023, reaching $66.4 billion. Although cloud computing remains one of the fastest-growing parts of the IT market, facing ongoing macroeconomic uncertainty, customer investment continues to slow down, with cloud growth falling below 20% for the first time.

From the financial reports, all three major cloud giants are facing the dilemma of declining growth rates. In the first quarter of this year, the AWS growth rate of Amazon's cloud business, which ranked first in the industry, slowed to 16%, falling below 20% for the first time, marking the lowest growth rate since Amazon began splitting the department's performance in 2015.

During this quarter, Microsoft Intelligent Cloud, the world's second largest cloud service provider, saw a year-on-year growth of 16% in revenue, but its growth rate has slowed down for seven consecutive quarters. As a rising star, Gu Geyun has a relatively high growth rate, with a year-on-year increase of 28%, maintaining a double-digit growth rate. However, compared with the previous two quarters, the same growth momentum is slowing down.

In the Chinese market, the latest report released by IDC titled "China Public Cloud Service Market (Second Half of 2022) Tracking" shows that the two major businesses in the Chinese public cloud market (IaaS+PaaS) grew by 19.0% year-on-year in the second half of 2022, showing a significant decline compared to the 42.9% growth rate in the same period in 2021.

This is inevitably due to factors such as a slowdown in macroeconomic activities, a decrease in revenue for major internet clients, and weak demand from Chinese internet clients. It is precisely for this reason that in 2022, the domestic cloud computing market experienced a trend of "big factory cloud down, operator cloud up".

With its accumulation in government and enterprise key customers, operator cloud has grown faster and faster, to the extent that in the second half of 2022, Tencent Cloud was overtaken and ranked fourth, with Alibaba Cloud, Huawei Cloud, and Tianyi Cloud becoming the top three manufacturers.

However, with the advent of the big model era, leading cloud vendors both domestically and internationally have long regarded it as a new incremental market. Among the top three giants in the global cloud market, especially Microsoft and Google, they believe that the cloud business is facing a critical moment in the AI big model and are using AI capabilities as a new lever to start a new round of market competition.

Among them, Microsoft has invested in OpenAI, launched a combination campaign, and has begun to accelerate the commercialization of technological achievements, including NewBing for search scenarios, Copilot for Office's intelligent assistant, and so on. Google also had to accelerate to keep up, so there was the question and answer robot Bard.

Similarly, domestic cloud vendors are also seizing the benefits of infrastructure while accelerating the commercialization of technological achievements.

In 2022, cloud vendors are no longer adhering to the path of "taking on all responsibilities" when accelerating the expansion of government and enterprise key customers. Instead, they are shifting from integration to being integrated (unlike "whole package", which only provides partial technology and products). In order to reduce costs and quickly make money, they are "taking a step back". From the strategy of manufacturers this year, the trend of "being integrated" is still ongoing. The MaaS business model of the AI big model, which is suitable for integrated features, also brings new opportunities for cloud vendors.

In 2023, the big cloud companies shifted from being conservative to actively attacking. Although they have seen the direction, the pressure of sustained profitability and technological iteration still hangs on their heads.

Especially the iteration of technology cannot be completed overnight. Tencent founder Ma Huateng recently stated at the shareholders' meeting that AI is a rare opportunity similar to the industrial revolution that invented electricity in hundreds of years. For the industrial revolution, taking out the electric light bulb one month earlier is not so important in the long term. The key is to solidly do a good job in the underlying algorithms, computing power, and data, and more importantly, implement the scenario. Tencent is currently doing some thinking.

The era of big models is approaching, and cloud vendors are slowly expanding their new responses.

*The captions and illustrations in the text originate from Visual China.