From grasslands to GPUs: Why China is building its AI powerhouse in Mongolia – explained in 5 charts
China’s artificial-intelligence boom is no longer confined to laboratories, software companies or technology districts in its biggest cities.By the end of 2025, China had 602 million generative-AI users, up 141.7 per cent from a year earlier, according to the China Internet Network Information Center (CNNIC). The country’s core AI industry was valued at more than 1.2 trillion yuan in 2025, while the number of AI companies surpassed 6,200, according to China’s Ministry of Industry and Information Technology.That surge in AI adoption is driving an equally rapid expansion in computing infrastructure. China’s intelligent computing capacity reached 1.59 million PFLOPS by the end of 2025, according to the National Data Administration. A petaflop (P) measures computing performance, with one P representing one quadrillion floating-point operations per second.But AI needs more than algorithms. It needs enormous amounts of computing power, and the electricity, cooling, land and network infrastructure required to keep that computing power running.One of the unexpected places emerging as a supplier of this infrastructure is Ulanqab, a city in Inner Mongolia better known for its grasslands, livestock and potatoes than for artificial intelligence.The transformation is striking. In 2010, Ulanqab’s economy was still firmly rooted in agriculture and animal husbandry. By 2013, the city had begun laying the foundations of a data-centre industry, with Huawei’s data-centre project marking an early milestone in its computing build-out.Over the following years, Ulanqab developed its cloud-computing base before making a sharper shift from general-purpose CPU computing to GPU-based intelligent computing in late 2022 and 2023, according to Ulanqab authorities.By August 2026, the city’s operating computing capacity had reached 172,000P, with intelligent computing accounting for more than 95 per cent, according to figures published by the Saihan District People’s Government, citing Xinhua.Computing is now becoming an increasingly important part of Ulanqab’s economic and infrastructure landscape.
Livestock and farming to AI hub
Five charts show how a grassland economy became one of China’s emerging computing hubs, and why scaling that transformation will require much more than simply adding servers.
From farms and livestock to computing
Ulanqab remains very much an agricultural and livestock economy. In 2025, its agriculture, forestry, animal husbandry and fisheries sector generated ¥34.75 billion, more than double the ¥14.1 billion recorded in 2011.But alongside that established economy, a new one has been taking shape.Ulanqab’s data-centre industry began with Huawei’s arrival in 2013 and has since grown into a major computing cluster. By 2025, the city’s computing industry was generating ¥6.8 billion in annual revenue, according to Ulanqab authorities.The point is not that computing has replaced farms, livestock or potatoes. Instead, a city long associated with grasslands and animal husbandry is developing a second economic identity around something far less visible: servers, data centres and AI computing.That transformation received another major push from China’s “East Data, West Computing” strategy, which seeks to move computing infrastructure towards resource-rich western regions while connecting it to demand in the country’s densely populated east.Inner Mongolia was designated one of China’s eight national computing hub nodes, with Ulanqab’s Jining Big Data Industrial Park included in the starting area of the national hub.The result is an unusual economic mix: agricultural and livestock production continues at scale, while a growing computing industry is being built around the region’s wind, land, cooling conditions and access to renewable power.
The computing base
The next question is whether Ulanqab’s data-centre push is simply a collection of individual projects or whether it has evolved into a genuinely expanding computing cluster.The numbers point to the latter.Ulanqab’s operating computing capacity rose from 24,000P in 2024 to more than 71,000P by June 2025 and 125,000P by the end of 2025. By August 2026, it had reached 172,000P, with intelligent computing accounting for more than 95 per cent of the total, according to the latest city figures.Physical infrastructure has expanded alongside computing capacity. Ulanqab now has more than 100 data-centre projects, more than 5 million contracted standard racks, about 650,000 racks built and 430,000 in operation, according to the same source.The scale increasingly reflects demand for AI workloads rather than conventional data storage and cloud services.That shift is also visible in where Ulanqab’s computing power is being used. The city says about 75,000P of its capacity is currently serving demand originating from Beijing, linking the Inner Mongolian computing cluster to one of China’s largest technology markets.Ulanqab is therefore becoming part of a wider computing network in which AI workloads can be processed in a region with abundant energy, cooler temperatures and high-speed links to major technology centres.
Why put AI servers on grasslands?
Ulanqab’s appeal comes down to an unusual combination of resources that AI computing consumes heavily: electricity, cooling and connectivity.The city sits in a region rich in wind and solar power. By the end of 2025, Ulanqab’s new-energy installed capacity had exceeded 20.3GW, while green electricity accounted for 67 per cent of the city’s overall electricity use. For newly built data centres, the share was more than 86 per cent.The climate provides another advantage.Ulanqab has an annual average temperature of about 4.3°C, allowing its data centres to use natural cooling for nearly 10 months of the year. Data-centre power usage effectiveness, or PUE, can fall as low as 1.1. In simple terms, that means relatively little additional electricity is required for cooling and other infrastructure beyond what is consumed by the computing equipment itself.Then there is connectivity.Ulanqab is about 240km from Beijing, and two dedicated 144-core optical routes connect the city directly to the capital. The latest reported minimum one-way latency is 2.1 milliseconds, allowing computing workloads to be processed in Ulanqab while maintaining a fast connection with Beijing.Put together, these advantages explain the economics of locating a large computing cluster here: abundant renewable power, a naturally cool climate, available land and high-speed connectivity to a major technology market.
Ulanqab’s national footprint
The scale of Ulanqab’s transformation becomes clearer when it is placed against China’s wider AI-compute build-out.The city’s 172,000P of operating computing capacity is roughly 7 per cent of China’s 2.45 million PFLOPS of intelligent computing capacity, based on the latest available figures.For a city of Ulanqab’s size and economic profile, that represents a significant footprint.But Ulanqab is not trying to compete with Beijing, Shanghai or Shenzhen as a centre for AI companies and research. Its role is different: it is becoming part of the physical infrastructure that allows those technology centres to scale.Much of the logic behind this model comes from China’s “East Data, West Computing” strategy. Workloads that do not require the lowest possible latency can be processed in regions with greater energy availability, cooler climates and more space for large computing facilities.Ulanqab is particularly well placed for that role because of its proximity to Beijing and its high-speed connectivity to the capital.This gives the city an unusual position in China’s AI economy. The algorithms, companies and users may be concentrated elsewhere, but an increasingly meaningful share of the computing capacity supporting them can sit hundreds of kilometres away, on land historically associated with farming and livestock.Ulanqab’s contribution, therefore, is less about becoming an AI headquarters and more about becoming an AI infrastructure base, turning its natural resources and geography into computing capacity usable across China.
Rising power bill
The same numbers that demonstrate Ulanqab’s success also reveal its emerging constraint.The city’s computing industry consumed 3.267 billion kWh of electricity in the first half of 2026, up 89.52 per cent from a year earlier. That implies first-half 2025 consumption of roughly 1.72 billion kWh.In other words, electricity demand almost doubled in a year.That is the catch in building an AI hub around cheap power. Once thousands of high-density GPUs arrive, cheap electricity alone is not enough. The grid must deliver power reliably, transmission and substations must keep pace, and renewable generation has to be backed by storage or other balancing capacity.Ulanqab is already responding. Its latest projects increasingly combine renewable generation, storage and dedicated power infrastructure rather than treating electricity as something that can simply be purchased from the grid.But the scale of the planned build-out remains enormous.At the end of 2025, Ulanqab had 125,000P of computing capacity. By June 2026, that had risen to 172,000P, while the city’s target for the end of 2026 is more than 200,000P.There is another limitation: talent and local value creation.Ulanqab has built training programmes and computing-related courses, but local officials themselves identify skilled digital workers as a continuing constraint. The challenge for the city is to build an ecosystem around its data centres rather than remain merely a place where servers sit.
Takeaway: What India could learn
Ulanqab’s transformation points to a broader shift in the economics of AI. As computing demand grows, the location of servers could matter almost as much as the location of the companies building the AI.The city’s advantage does not come from any single factor. It comes from the combination of renewable power, a naturally cool climate, available land, grid infrastructure and fast connectivity to a major technology market.That combination has allowed Ulanqab to build computing infrastructure far from China’s traditional technology centres while remaining closely connected to them.The lesson for India is not to recreate Ulanqab in one particular location. It is to think of AI infrastructure as an energy-and-computing ecosystem rather than simply a collection of data centres.India’s data-centre capacity is expanding rapidly, but much of the investment remains concentrated around established technology and business centres. As AI workloads become more power-intensive, states with abundant renewable energy, reliable grids, cooler climates and available land could potentially emerge as the next generation of computing hubs.Ulanqab offers a glimpse of what that future could look like: a city once better known for livestock and potatoes becoming an important part of the infrastructure powering a national AI economy.
Source link
