DeepSeek "makes a move" in a small city of 2 million people—Liang Wenfeng "takes a liking" to Ulanqab?

```

Taking the high-speed rail from Beijing, it takes less than two hours to reach Ulanqab in western Inner Mongolia.

This is a city with 2 million people in Inner Mongolia, known for its grasslands, volcanoes, milk skin, and mutton. It has also become a popular tourist destination because it sits at Jining Ancient City, a hub of the ancient Silk Road.

Note: Ulanqab scenery, source: internet

But Liang Wenfeng, a dual leader in China’s quantitative investment and large AI models, has recently begun planning in this border city.

The large model company under his command, DeepSeek, recently put up a "talent recruitment list", seeking "craftsmen" who are proficient in electromechanical, HVAC, and delivery to work locally.

What kind of “intention” lies behind this?

The hiring move is different from the norm

According to recent job postings by DeepSeek on recruitment platforms, the company is looking to hire some senior data center operations engineers and senior data center delivery managers.

The work location is—Ulanqab, Inner Mongolia.

This recruitment differs from their previous roles, especially in terms of location.

Previously, DeepSeek's recruitment focused mainly on Beijing and Hangzhou, with R&D positions as the main priority.

But this time, the recruitment is directly at the grasslands, and two positions related to data centers have been posted at once, making the entire move unusual.

Main focus on data center construction and operations

From the job settings, both positions revolve around data center operations.

In the industry, the well-known "data center" refers to IDCs (Internet Data Centers), which have high-speed internet access, high-performance local area networks, secure and reliable machine room environments, professional management, and comprehensive application service platforms.

In simple terms, it’s the "engine room" of the internet big data era.

The positions Liang Wenfeng is hiring for are senior data center operations engineers and senior data center delivery managers.

The former is usually responsible for the smooth operation and system support of the data center, while the latter is more inclined towards the construction and delivery of the data center.

Put together, it seems to "hint" that DeepSeek has roles related to data center construction, delivery, and operations in Ulanqab.

Main focus on data center construction and operations

According to the "Senior Data Center Operations Engineer" job description:

Responsible for technical support and operational management of the data center, including server hardware, network equipment operation assessment, monitoring, optimization, diagnosis, hardware performance optimization, and tool development, etc.

It emphasizes the duty of "optimizing operational processes and energy efficiency management", indicating proximity to the core operations of the data center, not just basic execution tasks.

The other position, "Senior Data Center Delivery Manager," is more focused on the construction and delivery phase.

The job description shows that this role is responsible for "managing the entire process of data center projects from initiation, construction, and delivery to operation", as well as "building IDC-related automated operation and maintenance tools and platforms," and "improving resource utilization efficiency," among other duties.

This indeed seems to involve the construction of a data center.

Key positions at the “core node” of a large model company

Some industry insiders believe that these two positions are not just a temporary move, but more like a systematic configuration around data facilities.

As is widely known, for a large model company, a dedicated data center is not a traditional "machine room". It's more like a "computing power factory" built specifically for model training and operation, equipped with large amounts of computing equipment continuously running to train models, handle requests, and output results.

Liang Wenfeng is famous for investing heavily in hardware and computing power.

Even before founding DeepSeek, his other company, High-Flyer Quant, was known for favoring large-scale computing rooms.

High-Flyer was one of the earliest quantitative private equity firms in the industry to promote their massive computing power and gymnasium-sized server rooms.

Therefore, it’s not surprising that he is preparing to build a data center for DeepSeek in Ulanqab.

Zhishitang notices that: in this large model company’s job description, "project management, electromechanics, and HVAC" are listed as a plus.

Such requirements seem to correspond to critical cooling functions in a data center. "Electromechanics" and "HVAC" directly refer to electricity distribution and cooling system management. These are essential for sustained and stable data center operations.

Why Ulanqab?

If DeepSeek really sets up the data center in Ulanqab, it’s not at all surprising.

Though Ulanqab is just a city with around 2 million people in Inner Mongolia, it's not new in the compute power industry.

Ulanqab’s main feature is—very cheap electricity.

Do you know Zhangjiakou, which is home to many IDCs? Thanks to abundant wind power, they enjoy low electricity prices, attracting many large domestic internet companies to set up their "machine rooms" there.

It’s said Ulanqab’s electricity is much cheaper than even Zhangjiakou's.

Ulanqab is covered by the unique Mengxi Power Grid, which relies on Inner Mongolia's abundant coal resources, has lots of pithead power plants, resulting in extremely low thermal power grid prices. At the same time, the Mengxi region’s new energy (wind and solar) installation scale is huge and cost-effective, with policy support, making Ulanqab’s big data center electricity prices highly competitive.

Supercomputing centers for large models use much more electricity than conventional internet data centers, so seeking places with convenient electricity and transportation may have strategic significance for DeepSeek.

Core city of computing power

In fact, Ulanqab has long been one of the core cities in Inner Mongolia’s “Eastern Data, Western Computing” project.

Information shows that as early as 2013, Ulanqab signed agreements with domestic companies to build cloud data centers. After 2016, Alibaba, Apple, Kuaishou, UCloud and others established data centers there. By 2024, 36 data center projects landed locally, 33 of which are intelligent computing centers, with total investment of 140.6 billion yuan and contracted rack scale exceeding 1.1 million racks.

Within the computing power network of big model firms, there’s a saying: "Gui (Guiyang) in the South, Wu (Ulanqab) in the North."

Specifically, southern computing power is represented by Guiyang, which has become an industrial highland in areas such as data aggregation, scheduling, data trading, and algorithm R&D;

While northern cities are represented by Ulanqab, leveraging abundant green electricity, a climate with almost year-round low temperatures, plenty of land, and an increasingly complete backbone network, it covers large-scale data center deployment and operation.

Public data shows Ulanqab’s average annual temperature is only 4.3°C, with a natural cooling period of close to 10 months, saving 20–30% in energy consumption; green electricity installed capacity is 19.402 million kilowatts, accounting for about 65.9%, providing stable, clean energy for computing; electricity costs are around 50% lower than the Beijing-Tianjin-Hebei region, further reducing long-term operation costs.

Great “imagination space” for the future

Some industry insiders believe that by building its own data center, DeepSeek shows Liang Wenfeng’s boldness, and that this may mean an even more ambitious strategic blueprint ahead.

It is widely rumored that DeepSeek’s next focus is AGI (Artificial General Intelligence), which places higher demands on compute power than ever before.

The model has to process massive data, conduct complex training and reasoning, demanding far more computing power than regular AI applications. Every training iteration needs stable and efficient hardware support; if there’s any fluctuation in electricity, cooling, or network, model training can be interrupted or grossly inefficient.

The data center roles in the hiring information may relate to further planning for the company’s computing needs. The much discussed cloud intelligence center, capable of supporting AGI training and inference—the "computing engine"—may gain a new geographic meaning because of this.

With Ulanqab’s beautiful scenery, unique resources, and convenient transportation, perhaps DeepSeek’s decision to settle in Ulanqab is not just for an office, but the launch of a grand technological blueprint…

Risk Disclosure and DisclaimerThe market involves risks, and investment needs caution. This article does not constitute personal investment advice, nor does it consider any specific user’s unique investment objectives, financial status, or needs. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific situation. Any investments based on this are at your own risk. ```