In May of this year, Alibaba Group released its latest financial report. In the 2022 fiscal year, Alibaba Cloud’s EBITA profit was 1.146 billion yuan, achieving its first profit in 13 years, and it became the only cloud service provider in China to achieve breakeven.
This seems to have released a signal: the development of cloud computing has entered the next stage, and the new battle horn has been sounded.
Who will be the next cloud vendor to win the “profit” baton?
Is it Huawei Cloud of “China’s Four Clouds”? Tencent Cloud? Or Baidu Smart Cloud?
It is undeniable that for a long time, in the arena of cloud computing, Baidu Smart Cloud has been playing the role of a chaser.
But the rivers and lakes are rivers and lakes. When the wind blows, the top players are accumulating strength, trying to define a new era.
Baidu is no exception.
From the late start and not being optimistic at the beginning, to the catching up in recent years, it has become “China’s fourth cloud”. How did Baidu Smart Cloud turn around with AI?
Perhaps, the answer can be glimpsed from a series of actions of Baidu Smart Cloud.
On December 27, at the “2022 Baidu Cloud Intelligence Summit·Smart Computing Summit” co-hosted by Baidu AI Cloud and China Institute of Electronic Technology Standardization, Baidu AI Cloud officially released a blockbuster product-Baidu AI Base.
Shen Dou, executive vice president of Baidu Group and president of Baidu Smart Cloud Business Group, said that the AI base is to help customers lower the threshold of AI industry applications and improve efficiency. Enterprises only need to focus on scene business.
Shen Dou, Executive Vice President of Baidu Group and President of Baidu Smart Cloud Business Group
01 There is no better time, now is the window period of Baidu Smart Cloud
It is irrefutable that now is the best time for Baidu Smart Cloud to catch up. Why now?
First of all, my country’s cloud computing policy environment is becoming more and more perfect, and policy support is becoming more firm.
At present, since 2006, Amazon sold its elastic computing capacity as a cloud service for the first time, and it has been more than ten years since the new business model of cloud computing was born.
Over the past ten years, the scale of the global cloud computing market has continued to grow, and my country’s cloud computing market has grown from the initial one billion to the current scale of hundreds of billions. Governments around the world have launched “cloud first” strategies, and my country’s cloud computing policy environment is also improving day by day.
In 2012, the state issued the “Twelfth Five-Year Plan for China’s Cloud Technology Development”, and the cloud computing software industry began to become a national key development task.
During the “13th Five-Year Plan” period, the state issued the “13th Five-Year Plan” National Science and Technology Innovation Plan, proposing to vigorously develop cloud computing technology and applications, support cloud computing to become a new generation of ICT infrastructure, and lay a solid foundation for development.
Last year, in the “14th Five-Year Plan”, the construction of digital China was raised to a new height, cloud computing became a key development industry, and entered a development stage of deep integration with the industry.
Secondly, in the context of the current global digital economy, cloud computing has become an inevitable choice for enterprises to transform digitally, and the pace of enterprises “going to the cloud” is accelerating.
If companies used to worry about “whether to go to the cloud”, then the question they are thinking now has become “how to better go to the cloud”?
In other words, “the industry needs it.”
Hou Zhenyu, vice president of Baidu Group, told Leifeng.com that with the popularization of cloud computing in the past ten years, the overall economy is moving in an increasingly digital direction. Whether it is the intelligent manufacturing industry or the Internet field, or the emerging life sciences, pharmaceuticals, etc., a large number of new intelligent technologies have been introduced into the industry, and the entire industry and economy are accelerating digital and intelligent upgrades.
Hou Zhenyu, Vice President of Baidu Group
For those engaged in the computing industry, the most gratifying point in recent years is that the entire large computing load has migrated from CPU homogeneous computing in the classic sense to heterogeneous computing, and to an architecture that focuses on AI and fits AI.
“This is the biggest trend that can be seen in the past ten years, and the most exciting big change brought about by the improvement of cloud-native business architecture.”
More importantly, Baidu has a deep accumulation at the technical level. The release of the AI base is not a whim, but a natural result of technology accumulation.
During nearly a hundred exchanges with Baidu people in the past few years, Leifeng.com has repeatedly heard a saying that Baidu has been a cloud-based company from the first day of its establishment.
In 2006, Baidu released a blog product – Baidu hi at its World Congress, which can provide users with image storage and display functions. Among them, the image storage function is made into a universal storage module. This move is regarded as the origin of Baidu Cloud, and also the prototype of the earliest C-end cloud storage system on the Internet in China.
In 2009, Baidu began to use GPU for acceleration;
In 2012, Baidu developed AI accelerators based on self-developed FPGAs. At the same time, it began to purchase and build GPU computing clusters on a large scale, and developed the world’s first parallel deep learning platform Paddle supporting GPU&CPU, which later evolved into a deep learning framework – Paddle;
In 2018, Baidu released China’s first cloud full-featured AI chip – “Kunlun Chip”, which has been deployed on a large scale with more than tens of thousands of chips;
In 2019, Baidu began to develop the pre-training model. In March of the same year, it released China’s first officially opened pre-training model ERNIE1.0, and in December 2021, it upgraded ERNIE3.0 to the world’s first 100-billion-enhanced large-scale model. ——Baidu Wenxin Large Model.
In the past, these product lines were “kings of their own” and developed independently. And when they are brought together for integrated integration, the Baidu AI base is born.
The first domestic full-stack self-developed AI infrastructure “Baidu AI Base” released by Baidu Smart Cloud
Just combining several product lines with one name will not achieve the real integration and optimization effect. Only vertical integration, the flying paddle deep learning framework can play the characteristics of the Kunlun chip, can achieve better results, and this is Baidu. The “soul” of the AI base.
Li Yanhong once publicly stated that Baidu has invested more than 100 billion yuan in research and development in the past 10 years. With 100 billion yuan in exchange for the initiative on the road of “chasing the cloud”, this deal is not a loss.
02 Baidu has saved for so many years, and what it saves is an initiative
In the cloud computing market, Baidu Smart Cloud is not the most advanced, it is undeniable.
But when it comes to the cloud vendors that were the first to deploy AI and use the AI card with ease in the cloud computing market, Baidu must have a name.
Data shows that in the past Q3 of 2022, Baidu Smart Cloud’s revenue was 4.5 billion yuan, a year-on-year increase of 24%, and the growth rate was higher than the industry average.
Canalys pointed out in the 2022 Q2 report that Baidu Smart Cloud grew by 25% year-on-year in Q2, more than double the overall market growth rate of 11%.
Not only that, in June 2022, IDC released the “2021H2 China AI Cloud Service Market Research Report”, and Baidu Smart Cloud once again won the first place in the market share of AI public cloud services. And this is Baidu Smart Cloud’s sixth consecutive domination of the number one throne.
The rapid growth and rising status are mainly due to Baidu’s self-confidence in technology and a large amount of R&D investment.
If you look carefully at Baidu’s 2022 Q2 financial report, you will find a detail worth pondering: supporting the company’s core business revenue is 6.3 billion yuan in research and development expenses, which exceeds the company’s core revenue by 25%.
And this is the seventh consecutive quarter that Baidu’s core R&D expenses accounted for more than 20%.
This is why Baidu was able to create the first full-stack self-developed AI infrastructure in China – the AI base. “Baidu’s investment in the AI field is indeed too long.” Others may have infrastructure similar to Baidu’s AI base, but they are not full-stack, full-element, and self-developed like Baidu.
According to Hou Zhenyu’s original words, there is currently no domestic company like Baidu that has a full-stack self-developed key technology in the core field of intelligent computing, and is very leading in every layer of chips, frameworks, models, and application elements. Baidu’s technical foundation, including the bottom-level hardware AI chip to the autonomous and controllable flying paddle deep learning framework, especially the recently popular large model, Baidu is currently very leading in these aspects.
It is precisely based on these large technological investments in the past, especially the forward-looking investment and technology accumulation of AI frameworks and AI chips, that Baidu Smart Cloud can continue to improve both in terms of revenue and industry status.
In fact, the current players in the field of cloud computing also have their own advantages.
The difference is that they either focus on AI but do not have the underlying chip and system capabilities, or they are from chip background but do not have the upper-level software capabilities, especially the ability to face needs directly. Thanks to its own search genes, Baidu has a unique opportunity to forge scenes.
It is reported that at present, tens of thousands of Baidu “Kunlun chips” have been launched, which are widely used in Baidu search and self-driving cars, iQiyi, Xiaodu and other businesses. For any other start-up AI chip company, there does not seem to be such a large-scale scene to polish.
Based on its own hard power, Baidu Smart Cloud is expanding its tentacles into more industry scenarios, such as finance, hydropower, industrial manufacturing, autonomous driving, and biocomputing. The application of the AI base in the State Grid is a typical case.
As we all know, in the energy field, inspection of high-voltage cables is an essential work, but many high-voltage cables are often located in deep mountains and dense forests, it is difficult for inspection personnel to enter, and inspection personnel need to climb up the cables to do various tasks. All kinds of inspections are very dangerous.
Therefore, Baidu helped State Grid Fujian build an AI middle platform with its AI middle platform capabilities. On this basis, it collaborated to build a large electric power model and built a full-chain intelligent processing capability for AI power key business data, realizing requirements, samples, models, The whole-process business integration of application and iteration supports State Grid Fujian Electric Power to build AI high-quality applications in a more efficient and standardized manner. At present, this method has helped them improve the recognition accuracy by 30% and the recognition efficiency by 5 times.
In addition, with the help of the AI base, the retraining of the general large model and the model specific to the scene is realized, and positive feedback is realized. In other words, the State Grid is a large group. Due to the different natural environments in different provinces and cities, a general large model trained in one place cannot be properly applied in every place. However, local sample information contributes to the general general model. Therefore, the data collected locally can be poured into the large model, trained locally, and fed back to the large model to enhance the generalization ability of the large model and achieve a comprehensive enhancement of the entire cycle.
03Baidu is down-to-earth, more lethal
Perhaps no one would have imagined that Baidu AI, once criticized for being unable to be commercialized, would become Baidu’s most powerful “killer” in the era of cloud computing.
This year, the report “Cloud Native AI-Accelerating AI Engineering Implementation” released by IDC pointed out that cloud native AI can effectively accelerate the implementation of AI engineering.
And Baidu Smart Cloud, which has taken advantage of the early deployment of AI-native cloud the day after tomorrow, may become the big winner after the new round of cloud computing market reshuffle: AI and cloud computing can not only tell a good story, but also change the way of “war”.
Baidu Smart Cloud is fighting “a very new kind of war”.
In 2020, Baidu Smart Cloud proposed the concept of “AI native cloud” for the first time.
In the past two years, Baidu Smart Cloud has rapidly improved its ranking in both the public cloud market and the smart cloud market.
In 2022, Baidu Smart Cloud has undergone a strategic upgrade, and the core strategy comes down to “integration of cloud and intelligence, deepening the industry.”
In general, strategies can be roughly divided into four layers:
The first layer is to deeply cultivate the industry and focus on actual scenarios. Utilize AI and cloud computing capabilities to strive to penetrate deeply in scenarios such as core manufacturing, urban governance, water affairs, and intelligent transportation.
The second layer is general-purpose products, especially technologies such as digital human technology, intelligent customer service, and digital twins that have been accumulated across modalities. These are general-purpose AI products.
The third layer is the AI base, which is composed of AI IaaS and AI PaaS. It integrates Baidu’s self-developed AI chip “Kunlun Core”, the Fei Paddle deep learning framework, and Wenxin large model, providing a complete solution for enterprise AI development and application. .
The fourth layer is infrastructure, common Cloud.
Frankly speaking, the cloud vendors who are currently well-known in the cloud computing arena are trying to define a new era of cloud computing and continue to launch corresponding strategic deployments. Therefore, it is unavoidable for each family to repeat: everyone thinks and does the same.
But for Baidu Smart Cloud, the AI base, which consists of AI IaaS and AI PaaS to form the entire AI environment layer, is Baidu’s most differentiated approach in this war.
Everyone in the industry knows that because AI applications are very scattered and the application scenarios are ever-changing, it is very difficult to extract a general AI model. Therefore, most players cannot provide an AI middle platform for the industry.
Based on the dual accumulation in the cloud and AI fields, Baidu submitted a new answer to the industry.
In addition, at this Smart Computing Summit, Baidu also released a new Baidu Yangquan Smart Computing Center, which is the largest single smart computing center in Asia built based on the full-stack capabilities of Baidu’s AI base.
Yangquan Intelligent Computing Center, formerly known as Yangquan Data Center, is a data center built by Baidu in 2012. Since it was put into operation, it has been the largest single data center in Asia.
This year, Baidu Smart Cloud officially upgraded the Yangquan Data Center to Yangquan Smart Computing Center. It has a computing power of 4EFLOPS—for reference, the Ali Wulanchabu Intelligent Computing Center that has been built so far has a computing power of 3 EFLOPS; SenseTime Shanghai Lingang Computing Center (AIDC) has a peak computing power of 3.74EFLOPS.
As Shen Dou said, the wave of intelligentization creating value for the industry has just begun. Cloud computing vendors need to standardize the output of intelligent underlying capabilities, and turn high-threshold technologies such as chips, large models, and deep learning frameworks into hydropower It can also be used by customers on demand.
And this new release means that Baidu Yangquan Intelligent Computing Center will be fully opened to the outside world to meet the ultra-large-scale AI computing needs of various industries.
Hou Zhenyu once told Leifeng.com that the future is a golden 10 years for the transformation of the entire computer system architecture, and there are not many companies in China that can stand at the level of affecting the global computer system architecture, so Baidu’s advantages in cloud computing will become more and more late. manifested.
Today, Baidu’s advantages seem to be emerging little by little.
Perhaps, in the entire cloud computing game, Ali’s development has indeed gained a first-mover advantage, but the market space is still very large. China’s cloud computing market is still in its infancy in a strict sense, and players from all walks of life still have a lot to do. And Baidu is trying to redefine a new game rule.
This article is transferred from:https://www.leiphone.com/category/ai/qDjDbmZ2iyqWt6OV.html
This site is only for collection, and the copyright belongs to the original author.