Just a two-hour train journey west of Beijing lies a landscape transformed. The rolling grasslands and ancient cinder cones of Inner Mongolia, long synonymous with sheep farming and coal mining, are now at the epicenter of China’s artificial intelligence revolution. This vast, arid region has rapidly emerged as the nation’s premier destination for building cutting-edge AI data centers, a phenomenon driven by a confluence of strategic advantages and burgeoning demand from Chinese tech giants.
Ulanqab, a city within Inner Mongolia with a population of approximately 1.5 million, has witnessed an unprecedented surge in data center construction. Since 2016, nearly 100 such facilities have either commenced operations or are currently under development. The scale of these commitments is staggering: Chinese companies have pledged to build projects with a combined estimated capacity of 12.5 gigawatts. What’s particularly noteworthy is the accelerated pace of these announcements, with over 70 percent of the total commitments made in the last year alone, positioning Ulanqab as one of Asia’s fastest-growing compute clusters, according to a recent research note from Goldman Sachs. For perspective, the ambitious $500 billion "Stargate Project" by OpenAI, a leading figure in global AI development, aims for a total capacity of 10 gigawatts upon completion. This comparison underscores the immense scale of investment and expansion occurring in Inner Mongolia.
The Allure of Ulanqab: A Strategic Nexus
Several key factors contribute to Ulanqab’s magnetic pull for data center developers. Its strategic location on the Inner Mongolian Plateau, characterized by high elevation and prolonged, frigid winters, significantly reduces the energy expenditure required for cooling these massive computing facilities. This natural advantage translates directly into operational cost savings, a critical consideration in the capital-intensive data center industry.
Furthermore, Ulanqab’s relative proximity to Beijing, China’s densely populated capital, allows for minimal latency in data transmission to major population centers. This is crucial for applications requiring rapid data exchange, such as real-time AI inference. However, the most compelling draw remains the cost of electricity. Inner Mongolia boasts some of the cheapest power rates in China, a benefit stemming from both the robust growth of its wind and solar energy sectors and the continued, albeit transitioning, abundance of coal resources. This cost advantage, coupled with the region’s cooling capabilities and connectivity, creates an economically attractive environment for large-scale data infrastructure development.
A Paradigm Shift: Chinese AI Companies Building Their Own Infrastructure
A significant development accompanying this data center boom is the discernible shift in investment strategy among Chinese AI companies. Historically, these firms have relied on renting compute power from established cloud providers. However, a growing number of leading Chinese AI companies are now making substantial direct investments in their own physical infrastructure.
Reports indicate that DeepSeek, a prominent AI research firm, is developing a massive AI data center in Ulanqab. This initiative is mirrored by major tech conglomerates such as ByteDance, Alibaba, and Xiaohongshu, all of whom are reportedly investing in their own data center facilities within the region. This trend marks a pivotal moment for China’s AI sector. Despite developing sophisticated AI models with impressive capabilities, Chinese AI companies have historically invested less in physical infrastructure compared to their American counterparts. The current wave of data center construction in Ulanqab signifies their intent to bridge this gap and establish a more self-sufficient and robust technological foundation. This strategic move is likely aimed at gaining greater control over their computing resources, optimizing performance, and potentially reducing long-term operational costs.
The Water Conundrum: A Growing Environmental Challenge
Despite the compelling economic and strategic advantages, the rapid expansion of data centers in Ulanqab presents a significant environmental challenge, particularly concerning water resources. Ulanqab’s arid climate, receiving only about 14 inches of rainfall annually, places considerable strain on its water supply. The local government is already grappling with meeting the existing water demands of its residents, a situation that will be further exacerbated by the numerous planned data center projects.
Evidence of this strain is already apparent. Last month, the local water company in Ulanqab was compelled to implement rolling shutdowns of several waterworks for seven hours each night to manage peak demand. While data from the local government suggests that data centers require additional water for cooling only during approximately two months of the year, the cumulative demand from such a high concentration of facilities, especially as new ones come online, poses a substantial and ongoing environmental concern for the region. This situation highlights the complex trade-offs between rapid technological advancement and sustainable resource management in water-scarce areas.
From Hinterlands to High-Tech Hubs: The "Eastern Data, Western Compute" Initiative
Inner Mongolia’s role as a data center hub predates the current AI fervor. The region has been recognized for its potential for some time, notably with Huawei establishing its first data center in Ulanqab in 2016, followed by Apple three years later. In 2021, the area was officially designated as a principal hub within a nationwide government initiative known as "Eastern Data, Western Compute." This ambitious national strategy aims to establish major data centers in China’s western and central regions, seeking to balance the country’s digital infrastructure and leverage the unique geographical and resource advantages of these less developed areas.
Initially, these remote data centers faced a significant hurdle: high latency rates when transmitting data to the predominantly eastern-based user base. This led to their primary utilization being relegated to backup storage and less time-sensitive applications. However, the advent of artificial intelligence has fundamentally reshaped this dynamic.
AI’s Game-Changing Impact on Latency
The rise of AI, particularly in 2022, brought a critical realization: these geographically dispersed data centers were ideally suited for AI model training. As Andrew Stokols, a professor at Singapore Management University specializing in China’s compute infrastructure, explains, "With the rise of AI in 2022, there was the realization that actually, those remote data centers could be well-utilized for model training." AI model training is a computationally intensive process that can span months and does not necessitate real-time user interaction, thus mitigating the impact of latency. This discovery breathed new life into the "Eastern Data, Western Compute" initiative, transforming previously underutilized facilities into crucial components of China’s AI ecosystem.
Moreover, the perception of Ulanqab’s remoteness is relative. Compared to other major data center hubs in Western China, Inner Mongolia is considerably closer to Beijing and other key metropolitan areas. The development of two dedicated fiber optic cables, laid in 2017 and 2019, has further enhanced connectivity. These cables have reduced average latency speeds to under five milliseconds, a benchmark fast enough to support real-time data exchanges essential for AI inference, thereby closing the performance gap with eastern data centers.
Commercial Demand Fuels Ulanqab’s Growth
Analysis suggests that Ulanqab’s data center expansion is largely driven by commercial demand, rather than solely by government-led investment, a distinction from some other Chinese data center hubs. As Chinese AI startups, including DeepSeek, Moonshot AI, and Zhipu AI, have successfully attracted a growing user base domestically, they have identified a clear business imperative for establishing inference data centers within reasonable proximity to their customers. This proximity ensures efficient and responsive AI services, a critical factor in retaining and growing their market share. The commercial viability of these locations, supported by technological advancements in connectivity, is a powerful engine for Ulanqab’s rapid development.
Harnessing Renewable Energy: A Dual Objective
The Chinese government views the development of data centers in Inner Mongolia as a strategic opportunity to address another critical national objective: absorbing the country’s surplus renewable energy capacity. Damien Ma, Director of Carnegie China, highlights a strong positive correlation between regions with significant unused renewable energy and those that have become centers for data center construction.
This symbiotic relationship presents a "win-win" scenario for the government. It allows China to accelerate its data center construction, potentially catching up to the pace set by the United States, while simultaneously stimulating demand for its burgeoning renewable energy sector. In line with this strategy, Envision, a major Chinese manufacturer of wind turbines, announced this month its intention to build a 2-gigawatt AI data center in Ulanqab, directly connected to the company’s own clean power supply. This move exemplifies the government’s vision of integrating renewable energy infrastructure with the growing demands of the digital economy.
The Persistent Reliance on Coal: A Transitional Reality
However, experts caution against an overly simplistic portrayal of AI data centers and renewable energy in China as a perfect synergy. Stokols’ research indicates that approximately 37 percent of Ulanqab’s electricity supply still originates from coal. The inherent operational requirement for data centers to function continuously, 24/7, has historically led operators to favor fossil fuels due to their perceived reliability.
"Inner Mongolia has long been the West Virginia of China. It’s a coal country," observes Ma, referencing the region’s deep-rooted ties to the coal industry. While the region is actively pursuing a transition to wind and solar power, the pace and ultimate extent of this shift remain uncertain. Consequently, the burgeoning data centers in China’s remote highlands will, for the foreseeable future, continue to rely on coal to some degree. Nevertheless, there is optimism that this reliance will diminish. Ma suggests that within the next three years, the region’s power mix could potentially be dominated by renewables, signaling a significant step towards a more sustainable energy future for its expanding digital infrastructure. The challenge lies in balancing the immediate demands of the AI boom with the long-term environmental goals of the nation.
