Why DeepSeek Is Moving Its AI Heart to the Mongolian Grasslands
DeepSeek’s latest hiring push reveals a strategic shift from algorithmic research to building and operating a high‑efficiency data center in Inner Mongolia’s Ulanqab, leveraging low‑temperature climate and existing cloud infrastructure to cut TCO, while gearing up for the upcoming V4 trillion‑parameter model.
DeepSeek data‑center recruitment
DeepSeek announced senior positions for a data‑center senior operations engineer and a senior delivery manager, each with a monthly salary of 30,000 CNY. The roles cover the full lifecycle of a data‑center project: initiation, construction, delivery, and ongoing operation.
Strategic shift to physical efficiency
Founder Liang Wenfeng promotes “small‑force miracles”, emphasizing architectural and engineering efficiency over brute‑force hardware scaling. DeepSeek is extending this philosophy to the physical data‑center level, aiming to lower total cost of ownership (TCO) by optimizing cooling, power consumption, and overall infrastructure.
Location: Ulanqab, Inner Mongolia
Ulanqab is one of the eight key nodes of China’s “East‑Data West‑Compute” strategy and a core component of the Inner Mongolia compute cluster. By the end of 2025 the region is projected to host 237,000 P of compute capacity, with more than 92 % dedicated to intelligent computing, making it the nation’s largest compute hub.
The cold climate provides natural “free air‑conditioning”, enabling new data‑centers to achieve Power Usage Effectiveness (PUE) below 1.2. Over 65 % of the regional power mix comes from clean energy.
Existing cloud ecosystem
Huawei Cloud built one of the world’s largest cloud data‑centers in Ulanqab in 2013. Alibaba Cloud plans to invest more than 10 billion CNY in a 114,000 m² campus. This mature ecosystem provides ready‑made infrastructure, supply chains, and operational expertise, dramatically lowering the risk and cost for DeepSeek’s own facility.
Upcoming V4 model and compute requirements
Internal sources indicate DeepSeek’s next flagship model, V4, will launch in late April. Rumored specifications include a trillion‑parameter scale and a context window of up to one million tokens. The model’s performance will depend on the large‑scale, efficient data‑center described above.
DeepSeek’s web interface already offers “Quick Mode” for simple queries and “Expert Mode” for long‑context, complex tasks, reflecting the need for expanded compute capacity.
Code example
梦晨 发自 凹非寺
量子位 | 公众号 QbitAI
本文经AI新媒体量子位（公众号 ID: QbitAI）授权转载，转载请联系出处IT Services Circle
Delivering cutting-edge internet insights and practical learning resources. We're a passionate and principled IT media platform.
How this landed with the community
Was this worth your time?
0 Comments
Thoughtful readers leave field notes, pushback, and hard-won operational detail here.