DeepSeek plans Inner Mongolia data center ahead of new AI model launch

  • Ulanqab in Inner Mongolia is considered one of China's best cities for data centers, boasting abundant power and ideal natural climate conditions.
  • The company is pushing into the asset-heavy physical infrastructure sector as it prepares to launch its next-generation V4 AI model later this month.
DeepSeek plans Inner Mongolia data center ahead of new AI model launch
(DeepSeek logo)

Chinese AI star DeepSeek is planning to build a data center in northern China's Inner Mongolia to meet growing computing demands.

The artificial intelligence startup, known for its low-cost algorithmic miracles, has posted several senior job openings based in Ulanqab, Inner Mongolia, offering monthly salaries of up to 30,000 yuan ($4,400).

These roles are primarily responsible for full-cycle project management and technical support, from the initiation and construction of the data center to its delivery and final operation.

This marks a shift for DeepSeek, which is no longer relying solely on software-level architecture optimization, but is stepping in to directly manage massive physical computing power.

Ulanqab is widely regarded as one of the most suitable cities in China for establishing data centers, possessing abundant electricity and premium natural climate conditions.

The region's extremely low average annual temperatures can significantly reduce server cooling costs, thereby helping to improve the power usage effectiveness of new data centers.

The area is already a hub for computing giants, with major tech companies like Apple, Huawei, and Alibaba having previously made large-scale investments there.

This asset-heavy strategic move comes on the eve of DeepSeek's planned release of its next-generation flagship model V4.

DeepSeek founder Liang Wenfeng recently said in an internal communication that the highly anticipated V4 model will be officially launched in late April, expected to feature a trillion-parameter scale and a million-level context window.

The new model is expected to achieve deep compatibility with domestic AI chips like Huawei's Ascend for the first time.

As the marginal returns on algorithmic innovation begin to diminish, owning large-scale and highly efficient physical data centers has become a focal point of competition for AI companies.

Late last month, DeepSeek experienced a system crash lasting nearly 12 hours, further highlighting the severe challenges currently facing the entire industry.

Building its own data center will help DeepSeek gain more initiative in future model iterations.

In preparation for the release of new-generation models, domestic tech giants such as Alibaba, ByteDance, and Tencent have already begun actively expanding their underlying computing infrastructure.

DeepSeek's next-generation flagship V4 model will be released in late April, according to the Chinese AI company's founder.
Apr 10, 2026

($1 = 6.8180 yuan)

AI News Alert
Subscribe to receive email notifications immediately when new articles about AI are published.
AI
View more channels