In this era of information explosion, data centers are like the arteries of the digital world, silently supporting every search, every message delivery, and even every response of smart devices in our daily lives. With the rapid development of intelligent computing technologies such as artificial intelligence and big data analysis, data centers are experiencing a revolution.
Before you know it, intelligent computing centers have blossomed all over the country.
Intelligent computing centers have blossomed all over the country
At present, the country has set off a wave of intelligent computing center construction, and various local governments, in cooperation with enterprises, have built dozens of intelligent computing centers. Next, let's sort out the more typical intelligent computing centers.
Wuhan Artificial Intelligence Computing Center
Led by the Wuhan Municipal Party Committee and Municipal Government, the Wuhan East Lake High-tech Zone Government undertook the construction, and Hubei Science and Technology Investment Group Co., Ltd. and Huawei jointly delivered the computing center, and was officially put into operation in May 2021. The specific scale of investment is not specified, but the total construction area is about 2,300 square meters.
The peak AI computing power of the computing center can reach 200 PFLOPS, and the peak performance of HPC supercomputing can reach 4 PFLOPS. The center fully uses Huawei's Atlas900 AI cluster, which is based on the Ascend 910 and Ascend 310 chips, and supports a variety of AI model training and inference. It provides computing power support for autonomous driving, city brain, smart healthcare, etc. It is planned to launch 400PAI computing power to serve more customers.
Hefei Advanced Computing Center
The computing center is operated by Hefei Big Data Asset Operation Co., Ltd. and officially launched in June 2021. The scale of the investment has not been disclosed, but as a major technology infrastructure, the scale of investment is expected to be considerable.
In terms of computing power, the peak value of double-precision computing is 12PFlops, the peak value of integer computing is 256Pops, and the total data storage capacity is 15PB. The main feature is the use of immersion liquid cooling technology, which improves the energy efficiency ratio by more than 30%. Serve scientific research and innovation and daily life, and promote the development of science and technology and the innovation of emerging industries. It is also exploring the application of hyper-fusion computing in multiple fields to accelerate the development of related application industries.
Nanjing Intelligent Computing Center
Initiated by Nanjing Kirin Science and Technology Innovation Park, jointly built by Inspur and Cambrian, it was officially put into operation in July 2021 with a total investment of 1.3 billion yuan.
In terms of computing power scale, after the completion of the second phase, the computing power can reach 18 billion times per second (1800 pops under octal system). Cambrian Siyuan 270 and Siyuan 290 intelligent chips are used to provide powerful AI computing services. Support cutting-edge scientific and technological fields such as protein folding and astronomical research. The "one center, one base, and N platform" model is adopted to support application innovation in the fields of science and technology finance, intelligent manufacturing, etc.
China Telecom Beijing-Tianjin-Hebei Big Data Intelligent Computing Center
China Telecom invested and constructed, and it was put into operation in August 2021. As an important computing infrastructure of China Telecom in the Beijing-Tianjin-Hebei region, it aims to support the region's big data and intelligent computing needs and promote the development of the digital economy.
Zhejiang (Yangtze River Delta) a new generation of full-featured intelligent supercomputing center
Initiated by Wuzhen High-tech Industrial Park, Tongxiang City, Zhejiang Province, it was officially put into operation in May 2022. The total investment is 2.2 billion yuan, and the total computing power reaches 180P.
As the first supercomputing center in Zhejiang, it fills the gap in the field of national supercomputing centers in Zhejiang. The center, also known as the "Light of Wuzhen" supercomputing center, is located on the north bank of the Beijing-Hangzhou Grand Canal in Wuzhen, with a construction area of about 50,000 square meters. It marks Zhejiang's formal status in the national supercomputing center sequence, which is of great significance for promoting the development of science and technology and promoting the innovation of emerging industries.
Xi'an Future Artificial Intelligence Computing Center
Invested by the Future Industrial City in Yanta District, Xi'an City, Shaanxi Province, undertaken by Shaanxi Construction Engineering First Construction Group Co., Ltd., and participated in the construction by Huawei, it was officially launched on September 9, 2021. The total investment is about 1.9 billion yuan, and the first phase includes 300P AI computing power and 8P HPC computing power, and the total planned computing power will reach 1000P in the future.
As the first large-scale AI computing cluster in Northwest China, relying on abundant computing resources, it is committed to promoting the rapid evolution of the artificial intelligence industry and creating a new highland for artificial intelligence innovation and development in Northwest China. Realize full-stack independent innovation in artificial intelligence and high-performance computing, and break through the key bottlenecks of software and hardware foundation.
Zhongyuan Artificial Intelligence Computing Center
Huawei cooperated with the Xuchang Municipal Government of Henan Province to build the project and officially put it into operation in October 2021. It is planned to build a computing power scale of 300P, of which the first phase has been completed with a scale of 100P.
The first fully autonomous and controllable all-flash, green and low-carbon artificial intelligence computing center in China. Huawei's Atlas900 AI training cluster has a leading edge in technology, and its powerful computing power is widely used in scientific research and business innovation such as smart manufacturing, smart agriculture, autonomous driving, and urban governance. At the same time, it is committed to building a green data center, reducing power consumption by 70% and increasing storage efficiency by 35%.
Harbin Artificial Intelligence Advanced Computing Center
Harbin Municipal Government invested in the construction and put into operation in December 2021. As an important measure to promote the development of the artificial intelligence industry in Harbin, the center is expected to reach the domestic advanced level in terms of data security and privacy protection, and provide support for the research and application of core AI technologies.
Alibaba Cloud Zhangbei Super Intelligent Computing Center
Alibaba Cloud invested and operated, and was officially launched on August 30, 2022.
The computing power scale reaches 12,000 PFLOPS (1,200 trillion floating point operations per second), making it one of the largest intelligent computing centers in the world.
The center uses advanced power consumption solutions and clean energy to achieve a high energy efficiency ratio. With the support of the Feitian intelligent computing platform, the computing efficiency has been significantly improved, the parallel computing efficiency of kilocalories has been greatly improved, and the AI large model training, autonomous driving, spatial geography and other artificial intelligence exploration applications have been supported.
In addition, Alibaba Cloud also has the Ulanqab Intelligent Computing Center. As part of Alibaba Cloud's full-stack intelligent computing solution, the construction and operation of the Ulanqab Intelligent Computing Center will further promote the development of intelligent computing services.
SenseTime Artificial Intelligence Computing Center
Invested by SenseTime, it started operation on January 24, 2022. The investment is about 5.6 billion yuan, and the number of cabinets in the first phase is 5,000. In terms of computing power, the peak training computing power reached 3740PFLOPS, ranking among the largest AI computing centers in Asia.
The main feature of the center is an open, large-scale, low-carbon advanced computing infrastructure that serves large-scale AI R&D and industrialization. It will accelerate the R&D and application process of smart chips such as domestic GPUs, and it is expected that when all servers are in place, the proportion of hardware localization will exceed 50%.
Baidu Intelligent Cloud-Kunlun Core (Yancheng) Intelligent Computing Center Project
Baidu Intelligent Cloud Investment, launched in September 2022.
The computing power scale is 200PFLOPS, and its core feature is the use of Baidu's self-developed AI chip - Kunlun chip. The launch of the Yancheng Intelligent Computing Center demonstrates Baidu's technical strength and application potential in the field of intelligent computing, and the construction and operation of the intelligent computing center will further promote Baidu's development in intelligent cloud services.
Tencent Yangtze River Delta (Shanghai) Artificial Intelligence Advanced Computing Center
As an important part of Tencent's presence in the Yangtze River Delta region, the intelligent computing center is expected to play a key role in promoting the development of the regional artificial intelligence industry, and the construction and operation of the center will further consolidate Tencent's leading position in the field of intelligent computing and artificial intelligence.
By analyzing these intelligent computing centers, some distinguishing features can be uncovered. For example, the Wuhan Artificial Intelligence Computing Center is led by the Wuhan Municipal Party Committee and Municipal Government, and the Hefei Advanced Computing Center is operated by Hefei Big Data Asset Operation Co., Ltd., and the Intelligent Computing Center widely uses self-developed AI chips and technologies, such as the Cambrian Intelligent Chip in Nanjing Intelligent Computing Center and the Huawei Atlas900 in Zhongyuan Artificial Intelligence Computing Center AI training cluster, Baidu Kunlun Core (Yancheng) Intelligent Computing Center project uses self-developed Kunlun chips, and the intelligent computing center has a wide range of service fields, not only supporting scientific research and innovation, but also serving multiple industries such as intelligent manufacturing, smart agriculture, urban governance, etc., and promoting the development of related application industries.
Why are local governments and enterprises suddenly keen to build intelligent computing centers?
In just a few years, dozens of intelligent computing centers have sprung up all over the country. So, what led to this round of intelligent computing center construction boom?
The Intelligent Computing Data Center, imagine being the kind of master artist who can play with both a traditional paintbrush and a digital tablet. Not only can data be stored and processed in traditional ways, but it can also play with the latest technologies, such as artificial intelligence and big data. They don't stop there, they can also use the brains of artificial intelligence to find gold mines in the data while processing massive amounts of information. To put it simply, this kind of data center is like an intelligent robot that can think and solve problems on its own, and it is amazingly efficient.
Traditional data centers, on the other hand, are more like craftsmen who focus on traditional craftsmanship, with a solid foundation, focusing on the day-to-day tasks of storing data, protecting it from loss, and managing the network. While they may not seem like a big job, in the early days of the digital world, these data centers were like solid bricks and stones that supported the entire online world.
In other words, the intelligent computing data center is like the pioneer of future technology, which is not only fast, efficient, and self-taught, while the traditional data center is more like a veteran guarding the early memories of the Internet, stable and reliable. The difference between the two is like a dialogue between the future and the past, with both stark contrasts and room to learn from each other.
In terms of hardware, energy efficiency, and operation and maintenance, intelligent computing data centers are very different
When exploring the world of intelligent computing data centers versus traditional data centers, one of the key differences we have to mention is their hardware deployment strategies. It's like comparing a supercar equipped with the latest technology to a reliable family car, both with their own unique designs and features that serve different needs.
Intelligent computing data centers, where the hardware is like the engine of a sports car, must not only be powerful, but also intelligent. This kind of data center requires not only speed, but also the ability to handle large-scale, complex data. As a result, GPU-intensive servers have become the standard. GPUs, or graphics processing units, were originally designed to process complex graphics and video content, but they are now being used to accelerate a variety of computing tasks, especially those involving artificial intelligence.
In addition to this, to further optimize the performance of intelligent computing, these data centers will also deploy dedicated AI processors. Designed specifically for AI computing tasks, these processors deliver more efficient and accurate computing power, like a racetrack-specific supercharger in a sports car, taking data center processing power to the next level.
In contrast, traditional data centers have hardware deployment strategies that are similar to those designed for family cars: reliable, standardized, and suitable for the masses. Servers and storage devices here are typically configured in a more standardized configuration to meet general-purpose computing needs. This doesn't mean that their technology is lagging behind, but that they are more focused on providing stable, reliable services than on pursuing extreme performance. Such hardware configurations, while may not be able to handle the most cutting-edge AI algorithms, are more than adequate for most everyday computing tasks.
This difference essentially reflects the different design philosophies and target users behind the two data centers. Intelligent computing data centers are boundary, the ability to apply the latest technological innovations to solve complex problems. Traditional data centers, on the other hand, focus more on providing stable and reliable computing services for the majority of users. Just like a supercar and a family car, both have their value and uses, and the key is how to choose the right solution for your needs.
In addition to the differences in hardware configuration, the intelligent computing center is very different from the traditional data center in terms of energy efficiency, heat dissipation, and O&M costs.
For example, intelligent computing data centers often leverage the latest AI technology to optimize energy distribution and ensure that every watt of power is getting the most out of it. Imagine a smart home assistant that automatically adjusts the use of appliances to achieve optimal energy efficiency based on the home's activity schedule and changes in energy prices.
In addition, efficient power and cooling systems are an integral part of intelligent computing data centers, which automatically adjust to the real-time needs of the data center, thereby minimizing energy waste.
There are significant differences in the cost of construction and operation and maintenance between intelligent computing data centers and traditional data centers, and these differences come from their different design concepts and technical applications. The intelligent computing data center is like a luxury car equipped with the most advanced technology, and its construction cost is naturally higher than that of traditional data centers that are like classic family cars. Smart data centers invest in high-performance computing hardware, advanced cooling systems, and automated energy management systems, which are expensive initial investments that can reduce operational and maintenance costs in the long run by improving energy efficiency and reducing manual intervention.
Will the intelligent computing data center gradually replace the traditional data center?
Just as a race car driver must constantly update his vehicles to stay competitive, smart data centers need to constantly introduce the latest technologies to handle the ever-increasing amount of data and complex computing demands. However, this rapid technology refresh requires not only a significant capital investment, but also ongoing technical evaluation and personnel training to ensure that data centers can take full advantage of the benefits of the new technology.
As the demand for computing increases, so does the energy consumption of intelligent computing data centers. How to effectively manage this high energy consumption to meet computing needs while maintaining energy efficiency has become a difficult problem for intelligent data centers. It's like maintaining a car's top speed while keeping fuel consumption under control, which requires careful strategy and management.
The rise of intelligent computing data centers has brought new requirements for energy efficiency, high-performance computing power, and environmental sustainability, while also driving innovation and development of data center technology.
With advances in artificial intelligence and machine learning, data centers will increasingly adopt these technologies to optimize operations, including energy management, failure prediction, and automated maintenance. As technology continues to evolve and the focus on sustainability increases, the data centers of the future will be more efficient, smarter, and more environmentally friendly, supporting the development of a digital society.
Another round of data center "false fire"?
It should be pointed out that China has experienced a round of data center boom, and many data centers have been built, resulting in low cabinet utilization in many data centers. So, will this wave of intelligent computing center construction repeat the mistake of "overcapacity" in the data center?
At present, China is experiencing a round of intelligent computing center construction climax, behind which there are not only optimistic expectations for the development potential of artificial intelligence, but also the strong desire of local governments to promote the upgrading of the technology industry. However, as some of the legacy of the previous data center construction boom shows, indiscriminate mass expansion can lead to wasted resources and inefficient utilization.
The construction of intelligent computing centers should be guided by market demand, and historically, the "overcapacity" of data centers was partly due to the mismatch between overly optimistic market expectations and actual demand. The construction of the intelligent computing center also needs to conduct accurate market research to ensure that the supply of computing power matches the future technology needs and industrial upgrading needs. For example, the demand for computing power in deep learning, autonomous driving, biomedicine and other fields is growing rapidly, and the planning and construction of intelligent computing centers should be tailored to the specific needs of these fields to avoid blind expansion.
The government plays an important role in the construction of intelligent computing centers, and should avoid homogeneous competition and disorderly expansion through policy guidance. On the one hand, the government can encourage the development of intelligent computing centers in a green, efficient and professional direction by formulating standards and guiding funds. On the other hand, through regional coordination and planning, the rational layout of intelligent computing centers can be realized to avoid excessive concentration and waste of resources in a certain region.
Through close cooperation with upstream and downstream enterprises, including hardware suppliers, software developers, and ultimately application scenario enterprises, the intelligent computing center can better serve the needs of the industry and promote the rapid application and industrialization of new technologies.
In general, although the construction of intelligent computing centers faces the risk of repeating the mistake of "overcapacity" in data centers, through market-oriented precise planning, continuous innovation in technology and management, policy guidance, regional coordination, and industrial chain