The end of AI computing power is photovoltaics and energy storage!

Xinjiang Supercomputing Center to be equipped with 1.6GW photovoltaic power? The end of computing power is indeed green electricity!

On November 16, the Xinjiang Production and Construction Corps held an investment promotion conference in Dongguan, Guangdong. At the conference, Tumushuke City of the Third Division of the Xinjiang Corps and Shanxi Bajian Group Co., Ltd. and Jingcai Future Intelligent Technology Co., Ltd. held a signing ceremony for the "Xinjiang Tumushuke 3000P (Flops) Intelligent Supercomputing Center and Supporting 1.6 Million KW Photovoltaic Project".

According to reports, the total investment of the project is 5.4 billion yuan, of which Jingcai Future plans to invest 800 million yuan to build a 3000P (Flops) Intelligent Supercomputing Center, and Shanxi Bajian Group plans to invest 4.6 billion yuan to build a supporting 1.6GW photovoltaic project.

The proposed photovoltaic project not only has an amazing capacity, but also an investment amount several times that of the supercomputing center itself, which naturally triggered widespread heated discussions: Can a supercomputing center use so much electricity?

Obviously, many people still don’t quite understand how crazy the demand for electricity is for computing power!

"Electricity-eating beast" - the end of computing power is electricity?

Computing power, in simple terms, is the ability to calculate, which is the ability of the data center server to process data and output results. Take cooking a dish as an analogy, data is the ingredients, algorithm is the recipe, and computing power is equivalent to the chef's cooking skills.

In recent years, AI has been booming, from the globally popular ChatGPT to the "disruptive" video generation model Sora, the computing power boom has been one after another, and competition between countries and regions has been surging.

Take Zhejiang as an example. As of September 2023, Zhejiang has built 219,000 5G base stations and 156 data centers, with a total computing power of 9EFLOPS (900 billion floating-point operations per second).

What is the concept of "900 billion times of computing power per second"? Assuming that 1.4 billion people in China perform a "1+1=2" calculation every second, then 900 billion times, it will take 1.4 billion people to calculate for 200 years without sleep.

So why do all regions have to "fight to the death" over computing power? Because computing power is a key production factor for digital transformation and the development of the artificial intelligence industry. Relevant statistics show that for every 1 point increase in the computing power index, the digital economy and GDP will grow by 3.3‰ and 1.8‰ respectively.

In other words, computing power is one of the core factors for which city will run faster and further in the future.

However, if you want to compete in computing power, you must solve the problem of electricity.

After all, computing centers are all "power-eating beasts."

Statistics show that the annual power consumption growth rate and power consumption share of data centers in my country that are only used for data storage basically maintain linear growth. It is estimated that by 2030, the total energy consumption of China's data centers will exceed 400 billion kilowatt-hours.

In March of this year, the high energy consumption of AI has caused widespread heated discussions at home and abroad. #ChatGPT consumes more than 500,000 kWh of electricity per day, which is more than 17,000 times that of an average American household# has become a hot topic.

According to The New Yorker, OpenAI's popular chatbot ChatGPT may consume more than 500,000 kWh of electricity per day to respond to about 200 million requests from users. In comparison, the average American household uses about 29 kWh of electricity per day, which means that ChatGPT's daily electricity consumption is more than 17,000 times that of a household. If generative AI is further adopted, the power consumption may be even higher.

According to estimates by the Chinese Academy of Sciences, assuming that ChatGPT has 200 million consultations per day, it is estimated that at least 0.0584 EFlops of computing power will be required per day, which will consume 792,000 kWh of electricity. Based on the average monthly electricity consumption of about 83 kWh per capita in Chinese residents, ChatGPT's monthly electricity consumption is about the same as the monthly electricity consumption of 286,000 Chinese people.

On February 29 this year, Tesla CEO Musk predicted at the Bosch Connected World 2024 Conference: The problem hindering the development of AI will change from "silicon shortage" to "power shortage" within two years.

"The growth in computing power is currently facing a bottleneck. Obviously, transformers will be in short supply next, by electricity. By 2025, we will not have enough electricity to run all the chips," Musk pointed out.

The end of AI is photovoltaics and energy storage!

Regarding the threat of power shortage faced by computing power development, Huang Renxun, founder of Nvidia, said in a public speech at the beginning of this year, "The end of AI is photovoltaics and energy storage system. Don't just think about computing power. If you only think about computers, you need to burn 14 Earth's energy."

Coincidentally, Sam Altman, founder of OpenAI, also said, "The future of AI technology depends on energy. We need more photovoltaics and energy storage."

It can be seen that the views of the big guys are surprisingly consistent: to develop computing power, it is inseparable from the development of new energy power.

So why must it be new energy power?

In traditional concepts, data centers pursue highly reliable and stable sources of electricity, while new energy power such as wind power and solar power fluctuates due to factors such as geography and weather, which does not seem to meet the energy characteristics of data centers.

Behind this seemingly contradictory choice, there is actually an inevitable trend in the development of data centers:

1. Urgent need for low-carbon transformation

Data centers are high-energy-consuming industries. With the continuous increase in computing power demand, their energy consumption and carbon emissions continue to grow. This not only affects the sustainable development of the industry, but also runs counter to the "dual carbon" goal. Therefore, data centers must accelerate the pace of energy conservation and carbon reduction, and the use of new energy is the key to achieving this goal.

The country attaches great importance to this. In June this year, the National Development and Reform Commission made it clear at a press conference that it would speed up the introduction of a special action plan for energy conservation and carbon reduction in key industries such as data centers.

In addition, in recent years, many places have also successively issued clear regulations on data center PUE (power utilization efficiency), and data centers that do not meet energy efficiency standards will face elimination. For example, Beijing issued the "Beijing Data Center Coordinated Development Implementation Plan (2021-2023)" as early as 2021, proposing to transform a batch of high-consumption and low-efficiency data centers.

2. Stability issues have been improved

The instability of new energy was once the biggest contradiction with the energy consumption characteristics of data centers. Today, data centers have made many explorations in the use of new energy, and this problem is gradually being solved.

Specifically, there are currently two mainstream routes:

First, by fully developing new energy, accelerating the access of green electricity to the grid, and increasing the proportion of green electricity in the power grid, data centers can obtain a more stable and sufficient supply of green energy.

For example, my country's "East Data West Computing" strategy encourages data centers to move to the resource-rich western region to utilize the abundant renewable energy in the western region. As of 2023, 246 national green data centers have been built, which play an important role in increasing the proportion of green electricity use and achieving green and low-carbon operations.

The second is the green electricity direct supply model. For example, the Inner Mongolia Autonomous Region proposed to support data center power users in the cluster to enjoy the electricity policy of strategic emerging industries, encourage new energy power generation companies to sign multi-year electricity trading contracts with intelligent computing centers, adopt green electricity direct supply and grid-type energy storage, and appropriately configure security new energy indicators for large intelligent computing centers. Measures such as stable guarantee of the country's best electricity price and a green electricity ratio of no less than 80%.

In terms of specific practice, the first data center cluster green electricity direct supply demonstration project has been built and put into use in the Horqin New District of Inner Mongolia, with a green electricity substitution rate of up to 43%, which significantly reduces the energy cost of data centers.

3. Obtain additional benefits

The large-scale consumption of new energy will not only help data centers achieve green transformation, but also bring them considerable market benefits.

In recent years, with the gradual improvement of my country's green electricity and green certificate markets, more and more data centers have changed from buyers to sellers in green electricity trading and carbon trading markets, thereby saving costs and achieving profitability.

Statistics show that by the end of 2023, data centers in provinces such as Hebei, Zhejiang and Jiangsu will purchase green electricity on a large scale, with a total transaction volume of more than 1.7 billion kWh, especially data centers in Zhangbei area, which will purchase more than 1.4 billion kWh of green electricity; from January to May 2024, the green electricity and green certificate transaction volume in the five southern provinces of China (Guangdong, Guangxi, Yunnan, Guizhou, and Hainan) exceeded 13.8 billion kWh, exceeding the scale of the whole year last year. Data centers and traditional high-energy-consuming industries have become the main force of green consumption.