Visual China.
Text |Silicon-based laboratory, author |Bai Jiajia.Alibaba Cloud fired the first shot in the cloud computing war in 2024.
At the press conference on February 29, Alibaba Cloud announced that the official website price of cloud products will be lowered by more than 20% on average, and the maximum price will be reduced by 55%.
In order to highlight the great strength, Liu Weiguang, senior vice president of Alibaba Cloud Intelligence Group and president of the public cloud business department, used two most and one first to summarize: the largest range of products involved in history - more than 100 products and more than 500 product specifications; The widest range of beneficiaries – millions of new and existing users; The unfulfilled portion of the customer's existing order is given to the customer for the first time.
* As soon as the battle began, it was full of gunsmoke.
On the evening of the same day, JD Cloud shouted to specific cloud providers such as Alibaba Cloud, Tencent Cloud, and Huawei Cloud on WeChat: Drop casually, compare to the end! We continue to compare prices across the network! Breakdown low price! Another 10% lower!
On March 4, Tencent Cloud also entered the ** battle. However, it does not sell cloud computing resources like its friends, but discounts AI products such as AI painting, face fusion, and audio file recognition.
It is almost certain that the battle for cloud computing will be fiercer and more ongoing this year than in 2023. However, on the back side of the competition, a tacit understanding can also be seen. After all, the war is just a means for these cloud vendors to stimulate demand and open the ceiling of the industry.
Today, how to develop the cloud computing industry is no longer a simple business issue, and how to improve the overall computing efficiency of China has been the focus of all sectors of society, and cloud vendors will obviously move forward in the complex game for a long time.
This article focuses on three main issues:
1. How is this round of ** battle different from the past?
2. How has the combination of AI and cloud changed the industry?
3. In the context of the continuous widening gap between China and the United States, what consensus have cloud vendors gathered? How will it evolve next?
* War is a normal phenomenon in the development of the cloud computing industry - with the iteration of technology and the release of scale effect, the cost of cloud computing will usher in a reduction every once in a while. The first battle is the means for cloud vendors to give back technology and scale dividends to the market.
Price reductions will inevitably lead to a decline in Alibaba Cloud's revenue and profits in the future, but this is not a short-term market competitive behavior, but a long-term strategic choice. Liu Weiguang said.
Counting the ** battles in the past few years, Alibaba Cloud is always the one who fired the first shot. Last year, Alibaba Cloud announced on April 26 that it would "cut the price for the largest time in history", with core products being lowered by 15% across the board to break 50%, and storage products being reduced by up to 50%. Immediately afterwards, JD Cloud, Tencent Cloud, Huawei Cloud, and e Cloud gradually released the news of price reduction.
From the perspective of routes, the price reduction models of major cloud vendors last year can be divided into two types. One is that Alibaba Cloud and JD Cloud have moved the play of To C to the battlefield of To B, attracting new users to settle in through means such as "0 yuan trial" and "1 yuan to enjoy 2 years of server", and then carry out continuous sales and services to earn value-added fees.
The other is to use first-class vendors and service channel providers to infiltrate the company like HUAWEI CLOUD. Last year, Huawei re-planned its channels to give its sales teams more preferential treatment and room for operation for government and enterprise customers and large KA customers.
This year, in addition to these two models, Tencent Cloud proposed for the first time to drive cloud resource usage by discounting AI products.
Behind the differentiation of fancy price reductions is the difference in the scale and business strategy of cloud vendors. As the cloud vendor with the highest market share in China, Alibaba Cloud has a significant scale effect, and can further share costs by expanding the user base.
It is worth mentioning that Alibaba Cloud's price reduction is not a simple price for volume. In response to Alibaba Cloud's price reduction strategy in 2023, Self Quadrant has pointed out that the core products of the price reduction include the old server after the iteration, which is limited by the service life, and the hardware begins to depreciate from the moment it leaves the factory. At the same time, some high-frequency, low-margin discount products also drove the sales of low-frequency and high-margin products.
Expansion of scale, high frequency with low frequency, low gross profit with high gross profit, similar logic is also reflected in Alibaba Cloud's ** policy this year.
On the one hand, this price reduction covers new and existing users, and even covers the unfulfilled part of the previous order forward in terms of effective time, apparently to further increase the upper limit while protecting the scale of customers. On the other hand, in terms of categories, ECS, OSS, and RDS are all core products that are most frequently used by cloud users.
Alibaba Cloud's ** battle seems to be menacing, but the spearhead is not aimed at peers, as Liu Weiguang said:"The price reduction move is not aimed at Tencent and Huawei, but to attract 72% of incremental users. ”
72%** data from the China Academy of Information and Communications Technology. As of 2022, 60% of the public cloud in the United States is in China, while only 28% in China, and the remaining 72% are enterprise-owned servers. The purpose of Alibaba Cloud is to pull this part into its own business territory. Liu Weiguang put forward the concept of "time cost" at the press conference, emphasizing that cloud computing is ready to use, and also pointing out the disadvantages of self-owned servers or self-built clouds that take a long time.
The non-Internet industry's unwillingness to accept cloud computing is the biggest headache for Alibaba Cloud at the moment. In the past decade, the penetration rate of cloud computing in the Internet industry has increased from 60% to 74%, which is similar to the overall cloud migration rate in Europe and the United States, while the penetration rate of the non-Internet industry has only increased from 24% to 29% during this period, far behind the rest of the market.
Objectively speaking, the low penetration rate cannot be blamed entirely on these companies not keeping up with the times.
In the government and enterprise market, for example, even if information security is not considered, SOEs are more inclined to build their own cloud servers because they have a long enough life to spread the cost below the purchase of public cloud services, and these enterprises often have IT departments in place and do not increase personnel expenditures too much.
It can be foreseen that if the purpose is to open the market, 2024 will not be the last time Alibaba Cloud fights the best battle, but it remains to be seen whether it can achieve the desired effect.
Judging from the development of the global cloud computing industry in recent years, it is not the core element that impresses customers, but the product.
In a way, Alibaba Cloud is following the path that AWS once did. AWS has become the world's No. 1 cloud provider in terms of market share, relying on price reductions. As of May 2020, Amazon AWS has reduced prices a total of 82 times. At its peak, AWS had cut prices 12 times a year for 3 consecutive years.
But in recent years, the low-price strategy has lost its charm. In terms of revenue scale, profit margin, growth rate and other aspects, Microsoft has surpassed AWS. The market believes that it is mainly due to Microsoft's bucket structure, which has product layouts in IaaS, SaaS, and PaaS to provide customers with full-link services.
Similar changes are brewing in the Chinese market. Previously, some industry insiders revealed that learning from Microsoft has almost become politically correct within the company, and the top management is thinking about how to form a bucket structure all day long.
On the whole, behind this year's first-class war, there are not only the objective laws of industry development at work, but also the intention of Alibaba Cloud to force peers to stimulate demand and open the ceiling.
The industry generally agrees that AI+ cloud is a new growth point for cloud vendors, but compared with overseas giants, the financial reports of domestic cloud vendors have not yet clearly reflected the contribution of AI to enterprise revenue.
Last year, Amazon, Microsoft, and Google all achieved double-digit year-on-year revenue growth, and Google Cloud's year-on-year growth rate reached 259%。Previously, the Silicon-based Research Laboratory revealed in "What signals are revealed in the performance of the three major cloud vendors in the United States?" pointed out that AI is the core driving force for the high growth of the three companies.
However, the revenue growth of domestic cloud vendors is still hovering in the single digits.
A few days ago, he and Ali respectively released their 2023 financial reports. Among them, Alibaba Cloud's revenue in the fourth quarter was 280$6.6 billion, up 3%, and adjusted EBITA profit up 86% to 236.4 billion yuan, a new high in the fiscal year.
Although the revenue of the intelligent cloud was not listed separately in the financial report, Robin Li revealed in the follow-up meeting that the total revenue of the intelligent cloud in the fourth quarter was 8.4 billion yuan, of which the large model brought about 6$5.6 billion in incremental revenue.
Looking through these two financial reports, AI has been the focus of attention, although the current revenue is not enough to directly support the performance, but it has faintly shown its potential.
For example, this set of data is given: by the end of 2023, the PaddlePaddle platform has gathered more than 10 million developers and served 2350,000 business units. So far, the developers have created 860,000 models on PaddlePaddle.
These developers will undoubtedly bring considerable increments to the performance of cloud providers in the future.
What kind of dividends will AI bring to cloud vendors? According to the silicon-based laboratory, there are two aspects:
First, the large model is naturally compatible with the cloud, and the demand for high-quality computing power has brought more orders to the technology flow players. Training a large model requires a large number of GPUs to run efficiently and stably, which means that the technical threshold, money, and time costs are very high, and cloud services that can be used as you buy and use will be a better choice than self-built clouds.
At the same time, in the past, the mainstream computing power provided by the computing center in the market was the general computing power with CPU computing as the core, which was not suitable for the GPU-centered intelligent computing power required for large model training, and the difference between this also brought a new round of computing infrastructure. Internet cloud vendors benefit from technical advantages and can dig more dividends in it.
According to the bidding information for government and enterprise digitalization in the fourth quarter of 2023 compiled by Digital Intelligence Frontline, among the total 266 large orders of more than one million yuan, Alibaba Cloud has become the one that has won the most bids among non-operator clouds, and has won more than 70 projects of more than one million yuan in a single quarter.
Second, after the emergence of the concept of MaaS (Model as a Service), the original IaaS, SaaS, and PaaS service paradigms in the cloud computing industry have changed, and they have been transformed into underlying capabilities or integration platforms, accessing enterprise business systems and data, and directly providing services as the central brain.
The viability of this service model was first verified by Microsoft. As the cloud vendor behind OpenAI, all GPT4-based MaaS services are completed through Azure (Microsoft's cloud service business), which not only increases its market share, but also polishes its own intelligent capabilities with greater intensity.
Following the cooperation between OpenAI and Microsoft, Chinese cloud vendors have derived two routes of AI + cloud.
The first is the AI for Cloud route, which is based on the innovation capabilities of cloud vendors, that is, the large model is a new carrier, which changes the channel for cloud vendors to provide services to enterprises, and also solves the problem between SaaS lightweight and PaaS customization.
Tencent Cloud's cloud computing resources are not cloud computing resources in this battle, but direct AI products, which is the embodiment of this route.
The second is the cloud for AI route adopted by cloud vendors with deep accumulation and large scale, that is, large models have created new computing power gaps, the importance of cloud computing has risen sharply, and the scale advantage established in the past has exploded.
Alibaba Cloud, which has the highest market share, has naturally become the "leader" of the second route. At last year's Alibaba Yunqi conference, Chairman of Alibaba's board of directors, Joe Tsai, pointed out that 80% of China's technology companies and half of the large models run on Alibaba Cloud.
It is worth mentioning that in addition to route differentiation, almost all Internet cloud vendors have chosen to close their formations and shift from scale to profit. For example, Alibaba Cloud has reduced projects with low profit margins, Robin Li has set profit targets for business segments such as content, cloud computing, big health, personal cloud, and Xiaodu, and Tencent Cloud has retreated from total integration with thin margins and large orders to integrated with thick profits and small orders.
A similar situation of opening up sources and reducing expenditure also occurs overseas. So far this year, nearly 150 companies in the U.S. tech industry have announced layoffs of nearly 40,000 jobs, according to the financial community. This phenomenon is similar to the cycle in the tech industry, where every time a new key technology emerges, companies need to cut out the tentative business and focus on real certainty.
Standing at the moment, looking back on the overall situation of China's cloud computing industry in the past year or so, Internet cloud manufacturers have a clearer direction under the catalysis of AI and have basically completed a new round of formation adjustment.
However, due to the lack of a dominant model like GPT-4, new models such as MaaS have not yet driven cloud vendors to achieve high growth and drive the attractiveness of traditional enterprises to the cloud.
Many customers still see the cloud as a server resource, but cloud computing is more than just a substitute for informatization, it brings much more value than that. ”
At Alibaba Cloud's 2024 strategy conference, Liu Weiguang emphasized several times that the market has insufficient understanding of cloud computing.
In this regard, he gave many sets of data similar to cloud computing that reduced the average duration of offline tasks by 42%, the duration of long-tail tasks by 70%, and improved the real-time query performance by 30%.
Behind these data, of course, there is Alibaba Cloud's intention to sell goods, but it also hides the underlying anxiety of China's cloud computing industry.
Anxiety first comes from the gap between the cloud computing industry in China and the United States.
From the perspective of revenue gap, Microsoft Cloud in 2022 is 76 times (at the current exchange rate, the same below), narrowing to 67 times. However, in terms of market share, according to the latest global cloud computing market ranking released by the Synergy Research Institute, Alibaba Cloud (4%) has fallen out of the top three in the world, and the gap with the third-place Google Cloud (11%) has opened a gap of nearly three times.
Secondly, the gap in computing power utilization also makes practitioners nervous.
Data from the China Academy of Information and Communications Technology shows that although the gap between China and the United States in the number of servers is not large, China has 20 million units, while the United States has 21 million. But 60% of the computing power in the United States is exported by the public cloud, while only 28% in China.
There is a consensus among technologists in the cloud computing industry that public clouds are more advanced productivity than private clouds. Previously, Caijing gave a set of data that the utilization efficiency of public cloud CPU is as high as 25%-50%. The CPU usage efficiency of private cloud deployments is usually only 1% to 2%, and generally does not exceed 5%.
The last anxiety is the dispersion of computing power.
Caijing previously compared the growth rate of public cloud IaaS and data center investment in China and internationally in the past five years, and found that the development trend of the two is diametrically opposed. According to IDC and Gartner data, the growth rate of China's data center investment has been rising since 2019, while the growth rate of the public cloud IaaS market has declined. In the international market, the growth rate of the public cloud IaaS market has always been higher than the growth rate of data center investment.
This phenomenon not only means that the new AI computing power that leading technology companies can obtain is reduced, but also may lead to their living space being squeezed by local intelligent computing centers, making it difficult for them to give full play to the scale effect. The bigger hidden danger is that if the local intelligent computing centers lack self-viability, they will need more market subsidies, which will bring financial burdens.
Caijing contacted a number of policy people told them,Local governments are not encouraged to build intelligent computing centers in person, because this can easily lead to duplicate construction and waste of resources.
The above can be said to be the common anxiety of the industry, and it can also be said to be the consensus of the industry, in the reshuffle set off by AI, based on these origins, players are looking for a new win-win path. The good news is that some results are already being seen.
The most intuitive change is that carrier clouds and Internet clouds have begun to share the dividends brought by computing infrastructure.
As mentioned above, Alibaba Cloud won more than 70 large orders of more than one million yuan in the fourth quarter of 2023. Among them, it can be seen that in some orders, Alibaba Cloud, Tencent Cloud, and China Mobile have cooperated to jointly build cloud platforms or computing centers. For example, the Zhuhai Smart Cloud Platform project is jointly built by Tencent, Alibaba, China Telecom, and China Mobile.
The other side of the change is happening with domestic chips.
Huawei's Ascend 910B, released last year, is widely regarded as the most competitive AI product in the Chinese market, with performance comparable to the NVIDIA A100.
Although the Ascend 910B is only usable at present, it is not easy to use. In terms of ecosystem, Huawei still has a big gap with NVIDIA CUDA. However, in order to avoid the instability caused by geopolitics, more and more cloud vendors are gradually adopting or compatible with Huawei's chips. To facilitate this process, Huawei has even dispatched on-site engineers to many enterprises to solve the problem of adaptation to Ascend.
In terms of policy, the plight of the cloud computing industry has been taken seriously, and the binding clauses in the relevant documents have left more room for the development of enterprises.
At the end of December 2023, the "Implementation Opinions on the In-depth Implementation of the "Eastern Data and Western Computing" Project and the Acceleration of the Construction of a National Integrated Computing Network" jointly issued by five ministries and commissions pointed out that "in principle, all kinds of large and super-large data centers shall not be built outside the national hub node, and blind and disorderly competition between regions shall be resolutely avoided." ”
For Chinese cloud vendors, 2024 is a year of uncertainty and certainty.
The uncertainty lies in the fact that while there has been some consensus and positive factors, the end is still uncertain, and all players will need to find a way out of the dust of history.
And the certainty is this, at the moment when no one knows the correct solution, cloud vendors still choose the direction and path according to their own characteristics, and will implement it on a yearly basis.
At the end of the day, action is the only answer.
References:
1. Self-Quadrant: Apocalypse 2023: The Year of Cloud Computing
2. Ebang Power: Cloud Computing 2023: Who Grows and Who Falls Behind?
3. Digital intelligence frontline: from 266 large orders, look at the nine major signals of the cloud computing market
4. Finance and economics: There are five real problems in China's computing power industry
5. Snow Leopard Finance Club: Dialogue丨Two "largest price reductions in history" in 10 months, Alibaba Cloud aims at 72% of the incremental market