AI large model takes off, rushing computing power and energy storage!

Mondo Technology Updated on 2024-03-07

The AI (artificial intelligence) frenzy has swept the world, and "violent" computing power has ushered in the big test of energy storage!

Recently, Tesla CEO Elon Musk attended the Bosch Connected World 2024 conference remotely, and when talking about AI, he said, "The computing power of artificial intelligence seems to increase 10 times every 6 months, and the shortage of chips more than a year ago was the shortage."Next year you're going to see a power shortage that won't be able to keep all the chips running.

Coincidentally, Jensen Huang, CEO of Nvidia, a leading AI chip company, also said when attending the 2024 Stanford Economic Policy Institute SummitArtificial general intelligence (AGI) will arrive as soon as five years, and computing power will continue to improve, which will increase by a factor of 1 million in the next decade.

Sam Altman, CEO of OpenAI, an American artificial intelligence startup, also recently pointed out that artificial intelligence needs a breakthrough in energy in the future, because AI will consume far more electricity than people expect. The technology of AI in the future depends on energy, and more photovoltaics and energy storage will be required.

In fact, since OpenAI released ChatGPT in November 2022, there has been a boom in AI large models around the world. In 2024, OpenAI will release the Wensheng ** model SORA, which can generate multi-angle camera movements for up to one minute with just prompts, making the AI large model detonate the technology circle again.

At present, chatbots supported by ChatGPT and a number of AI large models have realized conversations with people, and the production of Wensheng diagram models and Wensheng ** models have been applied to games, film and television, painting and other industries.

The global AI "arms race" is in full swing. With the explosive growth of computing power demand, the "100-model war" may drive a significant increase in the construction and operation speed of related power and energy storage infrastructure required by the intelligent computing center and its supporting facilities.

Electricity and energy storage will become new industry barriers

In terms of domestic news, on December 5, 2023, "People**" published an article"Let computing power be as ready to use as water and electricity as soon as possible".He also pointed out that computing power is the productivity in the era of digital economy, and it mainly provides services to the society through computing infrastructure.

On December 25 of the same year, the National Development and Reform Commission and other five departments issued it"Implementation Opinions on the In-depth Implementation of the "Eastern Data and Western Computing" Project to Accelerate the Construction of a National Integrated Computing Network;The recent meeting of the State-owned Assets Supervision and Administration Commission also required relevant central enterprises to strengthen investment in artificial intelligenceAccelerate the construction of a number of intelligent computing centers……It is expected to further promote the wave of investment and construction of domestic intelligent computing infrastructure.

It is worth noting that AI models have sprung up like mushrooms after a rain, and they have been upgraded and iterated at an astonishing speed. It turns out that in addition to algorithms, the improvement of large model performance is the result of "brute" computing powerThe threshold for the core competition of AI in the future lies in the level and cost of power and energy storage required for massive computing power.

Huang also said that if he only thought about adding computers, he would burn 14 planets of energy in the future.

The industry agreesIntelligent computing centers, supercomputing centers, and data centers are the main bodies that consume high energy。According to data from China's Ministry of Ecology and Environment, the total electricity consumption of China's data centers reached 216.6 billion kWh in 2021, accounting for 2% of the total electricity consumption of the whole society6%, carbon emissions account for 114% or so. By 2025, the proportion of electricity consumption in China's data centers will increase to 5% of the total electricity consumption of the whole society.

At present, electricity is the most important energy supply for the intelligent computing centerThe computing power industry ushered in the energy testIn the future, resources that can support the continuous improvement of the performance of the intelligent computing center ——Electricity and energy storage will build new industry barriers.

Massive computing power to meet the big test of energy storage

Industry insiders believe that the intelligent computing center needs uninterrupted, stable and massive power throughout the year**Wind and solar load storage (wind power, photovoltaic, load, energy storage).It may become a new power supply model to match it in the future.

Among them, by virtue ofHigher cycle times, longer life, and higher energy densityand other advantages,Energy storage lithium iron phosphate batteryIn the application of backup power storage in the intelligent computing center, it has higher power supply reliability and economic value of carbon reduction and consumption reduction, and its application prospects are broad.

At present, many provinces and cities in China, such as Henan, Shandong, and Shenzhen, have introduced incentive policies to encourage intelligent computing centers and supercomputing centers to allocate and store.

Battery companies, including:CATL, BYD, Guoxuan Hi-Tech, EVE Lithium Energy, Honeycomb Energy, Funeng Technology, Envision Power, Xinwangda, Lishen Battery, Far East Battery, Penghui Energyand other China's leading battery companies have already laid out the research and development of energy storage batteries, and energy storage battery products have covered wind and solar energy storage, peak shaving, industrial and commercial energy storage, etcData Centersand other scenarios.

Recently, Sunwoda Energy Storage released intelligent IDC high-voltage modular lithium battery products at the 2024 Mobile World Congress. It is reported that this product is a small and precise representative product of the third-generation lithium battery system focusing on medium and large data centers and edge data centers.

In addition, domestic and foreign companies, including Huawei Digital Power, Crestec, Kehua Data, Schneider, Eaton, etc., have also widely deployed the lithium battery uninterruptible power supply market.

The energy storage construction of intelligent computing centers is facing challenges

It is worth noting that although China's R&D and commercialization of energy storage battery products leads the world, there are still some challenges in achieving large-scale application in intelligent computing centers

First, the electricity consumption of large-scale intelligent computing centers is hugeAlthough distributed photovoltaic + energy storage is widely used, its upper limit of power generation still needs to be increased;

Second, in terms of energy storage capacity and cell rateIn the future, the power demand of the backup power supply of the intelligent computing center will increase exponentially, and considering the redundant configuration, the actual requirements for energy storage capacity will be higher. At the same time, the massive computing power also puts forward extremely high requirements for the rate performance of energy storage cells;

Third, in terms of energy storage grid connection, the grid connection process is complexHow to achieve large-scale grid connection involving the coordination and compatibility of multiple power systems, the bottleneck needs to be broken;

Fourth, the construction planning of energy storage facilitiesDue to the large footprint of large-capacity energy storage equipment, which affects the layout of the entire intelligent computing center, at the beginning of construction, the intelligent computing center needs to be coordinated with energy storage battery enterprises.

Fifth, the construction and operation and maintenance costs of energy storage facilitiesDue to the long operation cycle of the intelligent computing center for decades, higher requirements are put forward for the service life and cost of energy storage cells and systems.

At present, the construction of energy storage facilities in China's intelligent computing center is still relatively slow, and it is expected to further accelerate the pace of construction driven by the future demand for large electricity.

For example, the Action Plan for the High-Quality Development of Computing Infrastructure in Shenzhen (2024-2025) issued by the Shenzhen Municipal Bureau of Industry and Information Technology on December 5, 2023 states that:It will accelerate the implementation of green energy-saving products and technologies such as lithium batteries, and explore the establishment of a collaborative construction and linkage operation mechanism between the computing center and the power grid; explore implementation plans such as power energy storage; Explore the all-electric mode power guarantee scheme for mutual backup of power supply between computing centers; Encourage data center operators to reduce carbon emissions by self-built photovoltaics and other methods.

Looking further, if you compare the intelligent computing center withEnergy storage power stationWith the same matching, the energy storage system can not only be used as an uninterruptible power supply, use the energy storage capacity to shave the peak and fill the valley, but also generate electricity to relieve the pressure on the power grid. At the same time, with the upgrade of the power grid, the core competitiveness of future computing infrastructure will be greatly improved.

At present, the high-safety lithium iron phosphate energy storage cells of China's leading battery companies have been widely recognized in the world, and they are expected to show their strength in the field of computing power and energy under the wave of AI large models in the future.

Related Pages