Ulanqab: Quietly Accumulating Wealth, Now Impossible to Ignore!

In August, Goldman Sachs published a report on China's data center industry, highlighting a small city in central Inner Mongolia, with a population of under 2 million, as "one of the largest and fastest-growing AI computing hubs in the Asia-Pacific region."

Recently, a Bloomberg journalist paid a visit to the city. At a bustling intersection in eastern Ulanqab during Friday lunchtime, workers streamed out from a data center construction site, while vendors beside food trucks busily served chicken and tofu skin. "Hundreds of people every day, without fail," she remarked.

Why is Wall Street Casting Its Gaze on Ulanqab?

From the "Potato Capital of China" to the "Token Hub": How Ulanqab's AI Computing Resurgence Captured Global Capital's Interest

Some cities thrive on mineral wealth; others on manufacturing prowess.

Ulanqab's transformation has been a subtle yet significant shift. Once synonymous with potatoes, grasslands, and a cool climate, the city now increasingly finds itself in discussions centered around data centers, intelligent computing, and tokens.

This narrative extends beyond mere city rebranding.

As AI large models enter a phase demanding intensive computation, factors such as electricity supply, land availability, network connectivity, cooling efficiency, and capital expenditure are reshaping a city's industrial standing. Ulanqab has successfully amalgamated several previously undervalued resources into a new industrial gateway.

Recently, Wall Street Journal, Caixin, and various overseas tech media outlets have turned their spotlight on Ulanqab.

On one hand, reports situate the city within the broader context of China's AI infrastructure expansion and the global computing power race. On the other hand, leading model companies like DeepSeek are reportedly making significant strides in deploying large-scale computing power in Inner Mongolia, prompting global investors to reassess China's underlying AI infrastructure.

It's important to note that reports on specific investment scales, operational timelines, and self-built data centers for certain enterprises are often based on "reports" or information still in the planning stages and should not be universally regarded as completed projects.

The true focus should not be on grand planned figures but on whether Ulanqab can successfully navigate the layers of signing agreements, construction, operations, utilization rates, and industrial revenue.

Why Ulanqab Suddenly Captured Wall Street and Overseas Media's Attention

The global AI industry is transitioning from model competition to infrastructure competition. While model parameters, training data, and application ecosystems remain vital, the capacity to supply computing power has emerged as a critical constraint determining model iteration speed and inference costs.

In the U.S. market, tech giants continue to ramp up AI capital expenditures, with data centers, electricity, and chip supply chains becoming focal points for capital markets.

A similar trend is unfolding in China. The National Data Administration revealed that by the end of 2025, the national intelligent computing power scale had reached 1.59 million PFLOPS, with 42 ten-thousand-card intelligent computing clusters built and over 13.73 million standard racks of operational computing facilities in use.

Data from the China Academy of Information and Communications Technology also indicates rapid growth in China's computing device capacity. Computing power is no longer just a backend resource for cloud providers but has evolved into infrastructure for the AI industry.

Against this backdrop, Ulanqab offers three compelling points for capital markets.

Firstly, it's not starting its AI journey from scratch. The city began deploying cloud computing and data centers around 2016 and has since established a solid foundation in parks, networks, and operations.

Secondly, it's in close proximity to the Beijing-Tianjin-Hebei demand market. Training tasks can be more heavily deployed in resource-rich regions, while inference and real-time operations necessitate greater attention to network latency.

Ulanqab, located approximately 350 kilometers from Beijing, offers more land and energy space than first-tier cities without being entirely isolated from core clients.

Thirdly, AI computing power is increasingly becoming a business of electricity and energy management. Ulanqab boasts favorable wind and solar resources, with local initiatives exploring direct green power connections, integrated wind-solar-storage systems, and computing-electricity coordination.

For data centers, electricity pricing is just one aspect. Power supply stability, green power ratio, carbon management, and long-term contracts all determine the lifecycle costs of a project.

Why This City is Ideal for Computing Power Development

Ulanqab, situated in central Inner Mongolia, serves as a crucial node connecting North China, Northwest China, and Northeast China.

Its traditional industrial base does not excel in high-end electronics manufacturing, which, paradoxically, reduces path dependency when making new industrial choices.

The city requires a growth model not entirely reliant on population size or traditional manufacturing, making the computing power industry a pivotal breakthrough.

1. Cool Climate Eases Cooling Burden

While servers represent the most visible cost for data centers, electricity and cooling are equally critical for long-term operations.

Ulanqab's extended cool periods favor natural or hybrid cooling methods, reducing the demand for mechanical cooling.

Although "coolness" is not a decisive advantage, it creates sustained operational value in the era of high-power-density machine rooms.

2. Wind and Solar Resources Enable Green Computing

Inner Mongolia is China's significant wind and solar power base. Developing data centers in Ulanqab is not merely about relocating servers to the grasslands but integrating new energy consumption, energy storage regulation, and computing power loads.

The China International Capital Corporation's zero-carbon computing power base, operational in 2025, is publicly reported as an integrated wind-solar-storage green power direct-connected computing park, paired with 300,000 kW of wind-solar capacity and 45,000 kW of energy storage, embodying the computing-electricity coordination direction.

3. National Projects Offer Infrastructure Dividends

Following the launch of the "East Data, West Computing" project, the Inner Mongolia node undertakes non-real-time computing, offline analysis, and model training tasks for regions like Beijing-Tianjin-Hebei.

The significance of national hub nodes lies not just in a prestigious project name but in elevating network, energy consumption, industrial coordination, and resource allocation to a higher level.

4. Local Policies Shift from "Investment Attraction" to "Supporting Facilities"

Ulanqab has continuously advanced data center cluster construction in recent years, focusing on supporting measures for power supply, green power trading, park infrastructure, project approvals, talent introduction, and computing power applications.

Local public information reveals that by the end of 2025, the city had signed 84 data center projects, including 81 intelligent computing center projects, with total investments exceeding 500 billion yuan and cumulative completed investments reaching approximately 39.975 billion yuan.

These figures should be viewed separately.

The 500 billion yuan represents total signed project investments, while the 39.975 billion yuan reflects cumulative completed investments. Neither directly equates to local current industrial revenue.

The city's true capabilities will ultimately be validated by operational racks, stable clients, sustained utilization rates, and sustainable cash flows.

From "Numerous Signings" to "Genuine Industrial Formation": Ulanqab's Progress

Latest public reports indicate that Ulanqab's industrial scale has advanced from early-stage data center construction to concentrated expansion of intelligent computing centers.

Public information from August 2026 shows that the city has secured over 100 data center projects, with operational computing power reaching 172,000 P and intelligent computing power accounting for over 95%, marking a 112% year-on-year increase.

In the first half of the year, fixed asset investments in the computing power industry and big data electricity consumption maintained rapid growth.

Caixin, citing Ulanqab Bureau of Statistics data, reported that local internet data services (i.e., big data electricity consumption) reached 4 billion kWh in the first half of 2026, an 85.2% year-on-year increase.

Rapid electricity growth indicates increasing server and machine room loads but raises new questions: How much of this power translates into effective computing power, orders, and local economic value added?

From an enterprise perspective, Ulanqab attracts a diverse range of companies.

One category comprises cloud services and computing power infrastructure enterprises, such as Alibaba Cloud and Huawei, which bring cloud platforms, servers, networks, and software ecosystems.

Another category includes data center operators like China Data Center Group, GDS Holdings, 21Vianet, and China International Capital Corporation, which focus on park operations, cabinet services, computing power hosting, and client delivery.

A third category encompasses internet platforms and model companies like Kuaishou and ByteDance, whose businesses inherently require massive computing resources.

For model companies like DeepSeek, having more stable and controllable computing power supplies helps reduce reliance on external cloud resources. However, their specific construction timelines, investment entities, and chip configurations remain subject to official corporate disclosures.

Envision Technology's deployment of the Xinghe Base in Ulanqab has opened up new imaginative space for the industrial chain.

Public reports of the base's official operation in August 2026 link million-card parallel computing capacity, green power supply, and large-scale AI infrastructure.

However, planned parallel capacity does not equate to simultaneous full-load operation, requiring continued observation of actual utilization rates by investors and local governments.

Figure: Enterprise Deployment Matrix

Impact on Surrounding Cities Extends Beyond Data Centers

Data centers are not labor-intensive industries but generate a long supporting chain.

First come power facilities, energy storage, substations, communication networks, and cooling equipment, followed by server operations and maintenance, machine room management, security services, network operations, and engineering construction. Further downstream lie model training, industrial simulation, financial risk control, autonomous driving, and government applications.

Ulanqab's spillover effects extend in at least three directions.

Firstly, it can form computing power and application synergies with Hohhot. Ulanqab is better suited for large-scale infrastructure, while Hohhot possesses universities, research institutions, administrative, and industrial service resources, creating a combination of resource and application centers.

Secondly, it can drive computing power demand spillover to Beijing-Tianjin-Hebei. Beijing lacks no clients or technology but faces tighter constraints on land, electricity, energy consumption quotas, and data center expansion space.

Ulanqab can undertake some training, rendering, backup, and offline computing tasks, forming an "eastern demand, western supply" computing power division of labor.

Thirdly, it can prompt the energy industry to seek new clients. Previously, wind and solar power's value lay primarily in grid connection. Now, large computing power parks have become stable major clients.

Computing power can help absorb new energy, but the premise is that true synergy is achieved between the power grid, energy storage, dispatching, and data center loads, rather than simply placing two signs together.

The inspiration for other cities is not to "all build data centers."