BIG DATA’S DRIVE WEST

The National Development and Reform Commission (NDRC), which announced the plan, likened the importance of computing power for today’s digital economy to water conservancy facilities for agriculture and energy for industry. China aims to increase its digital economy to 10 percent of its GDP by 2025 from 7.8 percent in 2020. As part of China’s 14th Five-Year Plan (2021- 2025), this east-west computing network project will increase annual investment by 400 billion yuan (US$62.9b) along its supply chain, according to an NDRC forecast cited by China Central Television. In addition, with many big data centers planned for the country’s less-developed western regions where renewable energy is abundant, experts expect the project will help China achieve its 2030 carbon peak and 2060 carbon neutral goals. Data processing consumes a considerable amount of power.　

Big data centers join 5G, AI and the industrial internet as seven new infrastructure fields targeted by China to drive growth and upgrade the country’s economy. Investment enthusiasm in these sectors has risen over the past two years. Many local governments have put new infrastructure high on their agendas since 2020, announcing investments in the trillions of yuan. But the scale is still far from enough to replace the role of conventional infrastructure projects that have driven China’s growth. The computing project, however, compares to other cross-regional megaprojects such as the South-to-North Water Diversion Project and the west-east power transmission program, both of which took decades. The project has a similar name: Eastern Data and Western Computing. 　

But there are already concerns over excessive construction of data centers driven by local governments and capital, as market demand has yet to match supply. Besides, unlike traditional infrastructure, enterprises are the likely principal investors, which are more sensitive to market demand and input-output ratios. This means local governments need to adjust their role to better serve the new situation. 

According to a recent report by the International Data Corporation, China’s Inspur Information and Tsinghua University’s Institute for Global Industry, one percentage point increase in the computing power index (which considers computing power and efficiency) would translate to 0.33 percent growth in the digital economy and a 0.18-percent increase in GDP. 

　

The scale of China’s computing power has ballooned in recent years. In 2020, it reached 31 percent of the global computing scale with an increase of 55 percent year-on-year, according to a report published by the China Academy of Information and Communications Technology in September 2021, about 16 percentage points higher than the global growth rate. But there is an obvious gap. Big cities like Beijing, Shanghai and Guangzhou in Guangdong Province top the list, while less developed regions in the west, including Inner Mongolia Autonomous Region, and Guizhou and Gansu provinces, fall far behind. 　

“Most of our data centers are in the east where resources, including land and energy, are increasingly strained. The west is rich in these resources, particularly renewable energy, so there is great potential to develop data centers and share computing demand in the east,” the NDRC said on its website. 　

The NDRC points out that due to time lags over distance, some internet-reliant sectors such as the industrial internet, financial securities, telemedical services and video calls may remain in more developed clusters, including the Beijing-Tianjin-Hebei region, the Yangtze River Delta, the Guangdong-Hong Kong-Macao Greater Bay Area and the Chengdu-Chongqing region. Businesses that have less demand for internet transmission, such as data processing and offline analysis may be among the first to go west. 　

Huang Lihua, a professor at Shanghai’s Fudan University who took part in surveys on building data hubs in the west in 2021, told NewsChina that the project mainly involves data-intensive companies like telecom operators and internet companies. 　

For enterprises, cost and proximity to users are major considerations behind the cross-regional project. “When building data centers, companies need to keep customers’ demands in mind and consider regional factors including support facilities, land availability and power supply,” said a senior manager with Tencent Cloud, Tencent’s cloud computing arm, under the condition of anonymity. According to him, energy consumption makes up 40-50 percent of the entire cost of a data center, from construction to decommission. “China’s west is rich in renewable resources, such as hydropower in southwestern regions and wind and solar in the northwest. Building data centers in these areas could reduce energy costs while increasing consumption of green power locally,” he said. 　

Yang Fan, founder and vice president of SenseTime, an AI software company, said that low electricity prices in western areas are alluring for data centers. “Infrastructure like data centers are cost-driven businesses. Besides, using green power could help save enterprises’ carbon emissions quotas.” 　

Yang Fan said while demand for AI is high, it remains cost restrictive. “Reducing costs is a core motivation behind SenseTime’s investment in its large AI computing center [in Shanghai],” Yang said. “After the infrastructure is finished, we could provide services as soon as there is user demand. It’s the shortest path to generating profit,” Yang said. 　

Yang said SenseTime chose Shanghai to build its AI computing center partly because of the strong purchasing power in the Yangtze River Delta, where companies are more willing to invest in AI. “The future distribution of AI computing centers will surround industry clusters that have greater demand for AI-oriented industrial upgrading,” he added. 

“Local governments mainly took charge of preparing the business environment, like helping companies get land and electricity. But private companies are the major investors,” said a person engaged in the data center industry. 　

A 2018 report released by the Data Center Industry Alliance of China shows that 50 percent of data centers in the west were vacant. In some regions, only 10 percent of data centers were online. The report pointed out that local governments and capital drove many projects instead of market demand. 　

The same happened to data centers in first-tier cities. Research by Kezhi Consulting in 2021 on the internet data center market around Shanghai shows that an average 16.4 percent of computer rooms in the city were vacant in 2020. In some data centers, the rate exceeded 40 percent, showing a severe mismatch of supply and demand. 　

The Tencent manager told NewsChina that in terms of profitability, data centers are not like internet businesses, which have potential for quick market growth and viability. “But they can also generate steady rental income from good customers, which may bring higher return rates than traditional businesses. This is why so much capital entered the field in recent years,” he said. 　

Wang Qing, deputy chief engineer with the Industry and Planning Research Institute of China Academy of Information and Communications Technology, complained about the rush to invest in building data centers, noting that many are built without strategic coordination, which can easily cause redundancy and waste. 　

The NDRC requires a minimum occupancy average of 65 percent for server racks in the 10 data center clusters. “Now the project has started, companies will be more careful when they build data centers to avoid overly scattered distribution. This will help generate economies of scale and reduce costs, which are important considerations behind the project,” the Tencent manager said. 　

Despite the post-2020 rush to build data centers that are likely behind the oversupply, he said the overall industry remains healthy. “There used to be a supply shortage, after all,” he said. 　

In analyzing the supply-demand mismatch, Yang Fan distinguished data centers, which collect, store, process and distribute data, from AI computing centers, as the two operate differently. 

　

“Investment [in data centers] is mainly in land, construction and data center racks and climate control equipment. AI computing centers need to purchase servers and AI software and hardware beyond that, and directly provide services to industries... The two models vary a lot in unit economic value added,” Yang said. 　

He observed local governments now cared more about economic value added per unit of energy consumption brought about by data centers than before. “Previously, construction of industrial park-style data centers was going on everywhere with little thought into attracting high-quality enterprises. That’s how the mismatch occurred,” said Yang, adding that building computing centers requires greater integration capability along industrial chains to serve market demand. 

Some local governments already changed their approach, particularly under increasing pressure to reduce energy consumption. Wang Hui, an investment official in a second-tier eastern coastal city, told newspaper the Economic Observer they have no plans to court big data centers this year, as such projects are difficult to get approved now. 　

Besides energy consumption concerns, the government examined the necessity of data centers for the city when related fields such as the industrial internet and autonomous vehicles are yet to take off, Wang said. 　

By 2025, direct investment in seven major fields of new infrastructure will total 10 trillion yuan (US$1.6t), predicted the CCID, a research institution under the Ministry of Industry and Information Technology of China. 　

Unlike traditional infrastructure, new infrastructure hinges on market demand, so enterprises are the primary source of investment. This will require a role change for the government, said Yang Xiaoyi, a senior researcher with BRI Data, a data service provider engaged in infrastructure investment. 　

“What the government needs to do is stop direct investment and improve the institutional environment, stimulate market demand and clear obstacles hindering relevant industries,” Yang said. 　

Yang said they hope governments can help figure out their industrial needs to stimulate demand for AI to boost local economic development. 

Demand will determine not only the sustainability of projects but also the investment scale of new infrastructure for industrial upgrading and economic development, Yang Xiaoyi said. 　

Given this, experts expect more government involvement in attracting social capital for new infrastructure. But as new infrastructure is technology and innovation-driven, it is more complex and faces greater uncertainties, said Wu Liangcheng, vice president of the China Development Institute, an independent think tank. Compared to traditional infrastructure, it involves a much longer industrial chain and a wider scope of application, so it takes enormous investment, Wu said. 　

Investment in data or AI computing centers, for example, remains a high threshold for many companies. SenseTime’s AI computing center in Shanghai, which went online in January, cost 5.6 billion yuan (US$880.3m). A technician at SenseTime said it is their largest investment in fixed assets. The project started in July 2020, before the company’s Hong Kong IPO in late 2021, when its disposable income was only 19 billion yuan (US$3.0b), the technician said. 　

Xiao Gang, former president of China Securities Regulatory Commission, said at a financial forum held in Beijing in May 2021 that the lack of non-government investment is restricting the development of new infrastructure. Xiao pointed out that besides government, China’s three major telecom operators and some State-owned enterprises are major investors for new infrastructure. 　

Wu Liangcheng noted: “It’s difficult for non-government capital to get involved in new infrastructure.” Wu, who took part in the Shenzhen CPPCC survey on new infrastructure, said the lack of consensus between the government and market surprised him most. 　

“The recoup period for new infrastructure projects is long. A clear benefit-sharing mechanism should be designed to attract private enterprises,” Wu said. 　

Wen Laicheng, executive director of Zhongcai-CSCI Pengyuan Local Finance Investment and Funding Research Institute, said that public-private-partnership (PPP) projects (which encourage private capital to take part in the construction of public infrastructure and have a 30-year cap) mainly adhere to Ministry of Finance and NDRC rules, without State-level regulation. Stateowned enterprises make up the bulk of PPP projects, with private capital only accounting for 30 percent. 　

“Without legal guarantees, it’s challenging for private capital to enter projects that require large and long-term investment,” Wen said. 　

The Tencent manager said while he had not observed discrimination against private companies taking part in new infrastructure projects, building data centers requires an operating cycle of at least 10 years. Therefore, stable local business environments are probably “the most important thing” to companies, he said, adding he also expects clear regulations for public capital regarding what projects it can enter. 　

Wu admitted it will be hard to separate government from market forces with new infrastructure as it involves public goods. “The government should play a leading role, but that does not mean it has to be the chief investor in each project. New infrastructure projects should be classified by significance, such as whether a project has nationwide benefits, and be matched with a system of suppliers [by public capital or government funds],” Wu said. 　

“Also, there should be a negative list for market access, depending on the project. For example, market players should not enter projects involving public security and privacy. For projects off the list, a diverse supply mechanism could be built based on benefit and risk sharing,” Wu added.