DeepSeek's recruitment of civil engineering talent sends a signal. Recently, DeepSeek posted recruitment information for its IDC data center team on its official website, with positions including electrical, HVAC, automation, energy, and civil engineering, based in Beijing, Hangzhou, and Ulanqab in Inner Mongolia. In the posting, DeepSeek noted that every generation encounters its own infrastructure revolution-once railways, power grids, and the internet, and today the ultra-large-scale computing clusters that underpin AI development.
DeepSeek, known for its algorithms and models, has begun recruiting civil engineers and other traditional infrastructure roles, reflecting that leading AI companies are becoming 'heavier' in terms of computing power. For a long time, DeepSeek was known for its small team and high efficiency, training products competitive with Silicon Valley's top models under relatively limited computing resources. However, as model parameter scales continue to swell and user call volumes grow explosively, computing costs and supply stability have gradually become unavoidable bottlenecks.
In June this year, DeepSeek completed its first round of external financing, raising about 50 billion yuan. After the funds were in place, DeepSeek's strategy for acquiring computing power changed, and this recruitment signals its shift toward building its own computing infrastructure. According to media reports, DeepSeek plans to build a large AI data center with about 1 gigawatt of computing power in Ulanqab.
It is not just DeepSeek; leading large-model companies and internet giants across the industry are accelerating their computing infrastructure deployments. In July this year, Zhipu completed the acquisition of Zhongke Jiahe, a domestic AI heterogeneous computing software company. The latter originated from the compilation laboratory of the Institute of Computing Technology at the Chinese Academy of Sciences and is regarded as one of the country's top AI infrastructure teams. In addition, Zhipu AI has begun construction of a gigawatt-class domestic AI computing data center, using entirely domestic AI chips.
