DeepSeek is now the brain of Chinese state-owned firms

Asia Times

27-02-2025

Major state-owned enterprises (SOEs) in China have connected to DeepSeek's artificial intelligence model after the central government launched an 'AI+' program to call for boosting efficiency.
Xinhua reported that at least 20 companies owned by the central government have integrated DeepSeek into their operations. These companies are engaged in the energy, telecommunication, automobile, financial and construction industries.
In the energy and chemical sectors, Sinopec, PetroChina, CNOOC, Sinochem, China National Nuclear Power Co, China General Nuclear Power Group, China Southern Power Grid and China Oil and Gas Pipeline Network Corp (PipeChina) said they have already linked their AI models to DeepSeek-R1, which was launched on January 20.
'With DeepSeek and other large language models (LLMs), PipeChina's oil and gas control centers can set up their production plans within minutes instead of four hours,' Xu Kun, deputy general manager of Beijing Zhiwang Digital Technology Company, a unit of PipeChiona, told the China Central TV in an interview. 'They can also improve accuracy with AI models by 10%.'
LLMs refer to AI models like ChatGPT, which can understand human language or achieve natural language processing (NLP).
'Besides, PipeChina can use DeepSeek to reduce the time needed to simulate the construction of salt caverns for gas storage from tens of days to just one hour,' Xu said.
PipeChina, controlled by the State-owned Assets Supervision and Administration Commission (SASAC) of the State Council, is the construction contractor for the China–Russia East-Route Natural Gas Pipeline, or the Power of Siberia. The pipeline transfers 38 billion cubic meters of natural gas annually from Russia's eastern Siberia to northern China.
In 2022, PipeChina set up Beijing Zhiwang Digital Technology to digitalize its operations.
Last December, PipeChina launched its 'Pipeline Network' AI model using Huawei Cloud. It said it had deployed the model across more than 20 application scenarios and planned to use it in 80 more.
On February 17, Chinese President Xi Jinping met corporate leaders at a symposium in Beijing. These company heads included Huawei Technologies' Ren Zhengfei, Alibaba's Jack Ma, Tencent's Pony Ma, DeepSeek's Liang Wenfeng and Unitree's Wang Xingxing.
On February 19, the SASAC met with a group of central SOEs to summarize the results of their AI development and plan for future work. SOEs refer to all state-owned firms controlled by the central and local governments, while central SOEs are those owned by the central government.
'Central SOEs should seize the strategic window period for the development of the AI industry. They should improve LLMs, focus on applying core technologies, build an open ecosystem, develop original innovations 'from 0 to 1' and commercialize scientific achievements,' the meeting said.
'The central government should increase efforts to support Central SOEs to use AI technology, highlight the development of AI technology in the coming 15th Five-Year Plan (2026-2030), support the creation of more leading AI enterprises and start-ups,' it said, adding that the government and Central SOEs will increase capital investment to ensure that talents can focus on research and development over the long run.
'AI competition is all about data scale and quality,' Zhou Lisa, a researcher at the China Enterprise Reform and Development Society, a research unit of SASAC, told Xinhua. 'Central SOEs have substantial data resources, so they can use DeepSeek to share and marketize their data.'
'Many SOEs in the energy sector have already deployed their AI applications to promote intelligent transformation. Connecting with DeepSeek can allow them to provide even more solutions,' she said.
She said, for example, that Sinochem used DeepSeek's capabilities in complex information processing, logical reasoning and open-domain knowledge to meet the needs of different business scenarios; and that China Southern Power Grid used DeepSeek to upgrade its AI model called Big Watt, which was launched in September 2023 to collect data about facility damages.
China has been developing AI models for industrial use in the past few years.
In May last year, China Mobile launched its Jiutian AI model, which has professional knowledge in 15 industries, including telecommunications, electricity generation, logistics, energy, metal, construction, transportation and aviation. It said it would tailor-make AI models for 40 more industries.
In July, China Telecom launched a generative AI model called TeleChat2, which is good at video editing and graphic design. The company said TeleChat2, which understands different Chinese dialects, can be widely used in public services across different cities.
However, challenges remained when Chinese firms tried integrating their Internet-of-Things and AI technologies.
Dong Zhaojie, an AI expert of China Southern Power Grid, said in June 2024 that the company deployed drones to check its electricity transmission poles and took over ten thousand pictures of split pins on them.
He said he thought the AI model would be smart enough to identify the damaged split pins from the pictures, but it was not. He said engineers finally had to go to the sites to collect data and teach the AI model that there could be 500 different kinds of split pin damage.
He said that once the AI model started working, the company saved about 10 million yuan (US$1.37 million) in annual expenses for finding damaged split pins.
While state media celebrate China's advancement in AI technologies, a Jiangsu-based commentator called Qianqian warns that AI could replace tens of millions of jobs in China, from factory workers and deliverers to medical professionals and civil servants. She says people should keep learning new skills to avoid losing their jobs.
Yong Jian is a contributor to the Asia Times. He is a Chinese journalist who specializes in Chinese technology, economy and politics.
Read: China connects everything to DeepSeek in nationwide plan