US-China AI Race Shifts Toward Energy Infrastructure

Tashkent, Uzbekistan (UzDaily.uz) — The next arena of technological competition between the United States and China could be energy infrastructure rather than Silicon Valley or Beijing’s Zhongguancun, as the rapid development of artificial intelligence drives demand for computing capacity and, in turn, stable and relatively inexpensive electricity.

One of the most prominent examples is Ulanqab in Inner Mongolia. The city was previously known mainly for agricultural products, particularly potatoes and processed potato products. It is now seeking to become one of China’s largest AI computing centres.

According to China News Service, as of the end of June 2026, agreements had been signed in Ulanqab for 89 data centre projects with total investment exceeding 500 billion yuan.

Companies associated with the development of the region’s computing infrastructure include DeepSeek, Huawei, Alibaba, Apple, Kuaishou, Baidu and ByteDance.

One of the largest planned projects is expected to be a DeepSeek data centre with a capacity of about 1 GW. The generative AI technology company, after raising significant financing, allocated part of its resources to building large-scale computing infrastructure in Inner Mongolia.

The scale of planned investment has also attracted the attention of the international financial community. Analytical materials from Goldman Sachs describe Ulanqab as one of the fastest-growing AI computing clusters in the Asia-Pacific region.

As of June 2026, the combined operational and planned capacity of Ulanqab’s data centres was estimated at about 12.5 GW. Its potential for further expansion therefore substantially exceeds its existing capacity.

The main factor behind the region’s appeal is energy.

As large language models and other AI systems develop, demand is growing not only for computing chips but also for the electricity required to operate them. Modern data centres are highly energy-intensive facilities, requiring power for servers as well as cooling systems and other infrastructure.

Experts are increasingly describing electricity availability as a limiting factor for the scaling of AI.

Inner Mongolia has several advantages in this regard. First, the region has significant renewable energy potential. A large share of China’s technically available wind and solar generation potential is concentrated there. The total installed capacity of new energy sources in the autonomous region has already reached 174 million kW.

Second, electricity costs are lower than in China’s largest economic centres. While industrial electricity tariffs in Beijing, Shanghai, Guangzhou and Shenzhen are around 0.7 yuan per kWh, prices for end users in some parts of Inner Mongolia can be nearly half that level, at about 0.35 yuan per kWh.

Third, the region’s climate can reduce cooling costs. Ulanqab is located on a high-altitude plateau where the average annual temperature is about 4.3°C. During a significant part of the year, data centres can use outdoor cold air to cool equipment, reducing the energy consumption of their cooling systems.

Data transmission speed is another advantage.

Ulanqab is connected to Beijing by two direct-access fibre-optic lines, with data transmission latency of about 4.2 milliseconds. This allows the region to serve as a computing base for the country’s larger economic centres.

China is also developing a model for directly supplying data centres with green electricity. In Inner Mongolia, a demonstration project has already been implemented in which renewable electricity from solar power plants is supplied to a computing park through dedicated lines under a “point-to-point” model.

In other words, computing infrastructure is gradually moving to locations where inexpensive and clean energy is available.

Ulanqab’s development is part of China’s broader “East Data, West Computing” strategy. Its basic logic is that the country’s developed eastern regions generate large volumes of data and account for most demand for digital services, while western and northern territories have greater land and energy resources.

Computing workloads are therefore gradually being distributed among regions according to their infrastructure advantages.

Gansu province provides one example. The total computing capacity of the Qingyang data centre cluster has reached 215,000 P, with about 99% allocated to intelligent computing.

A large direct-supply project for green electricity to data centres has also been implemented in the Ningxia Hui autonomous region.

As a result, western regions of China already account for about 32.6% of the country’s total intelligent computing capacity.

The situation is different in the United States.

The US AI industry has leading chip developers, major technology companies and substantial investment. However, the rapid expansion of data centres is increasingly constrained by the capabilities of the power grid.

According to Morgan Stanley estimates cited in the source, US data centres’ electricity demand could reach 68 GW in 2026–2028, while existing infrastructure can provide only about 30 GW of additional capacity.

This means that potentially more than half of the demand could remain without the required power supply.

The problem has several dimensions.

One is public opposition to the construction of new data centres. Their development increases pressure on regional power grids and often requires the construction of new power plants, transmission lines and substations.

Another is the condition of the US power grid itself.

The US electricity system historically developed as several major interconnected systems — eastern, western and Texas. Interregional electricity transmission is limited, while construction of new high-voltage lines can take many years.

At the same time, US electricity demand grew relatively slowly for an extended period. As a result, parts of the energy infrastructure were not prepared for the sharp increase in demand associated with the development of artificial intelligence.

The global AI race is therefore gradually moving beyond competition between processor manufacturers and AI model developers.

Chips remain a critical resource, but without electricity they cannot become a functioning computing system.

China is relying on the geographical distribution of computing capacity and the relocation of energy-intensive workloads to regions with substantial renewable energy potential. Inner Mongolia occupies a particularly important position in this strategy because of its combination of relatively inexpensive electricity, wind and solar resources, a cold climate and proximity to major centres of demand.

The United States, by contrast, faces the simultaneous challenges of rising electricity consumption, upgrading power grids, building new generation capacity and addressing public opposition to large-scale infrastructure development.

The next stage of US-China technological competition may therefore centre not only on which country has better AI chips, but also on which has enough electricity to operate those chips at the required scale.

In Inner Mongolia’s steppe regions, the deserts of Ningxia and the energy sites of Gansu, China is developing a new foundation for the digital economy based on solar and wind resources, available land and the ability to generate electricity at relatively low cost.

Ulanqab could become one symbol of this emerging era, in which computing power increasingly begins with energy capacity.