Total investment in embodied AI across China's power sector could top 10 billion yuan in 2026.
Photo from Jiemian News
by LI Jiaqi
China's power grid operators are preparing a large-scale rollout of robots, marking one of the first industrial deployments of embodied AI.
State Grid Corp. of China has drawn up a plan to procure about 8,500 robots in 2026, with total investment of roughly 6.8 billion yuan ($9.4 billion), according to people familiar with the matter. The rollout will focus on power inspection, live-line operations, emergency response and logistics, as the grid shifts toward autonomous operations.
Including potential follow-up procurement by China Southern Power Grid and regional energy groups, total investment in embodied AI across China's power sector could top 10 billion yuan in 2026, a person familiar with the embodied AI industry told Jiemian News.
Procurement will be phased, with pilot orders in the first quarter, large-scale purchases in the third, and follow-up purchases in the fourth quarter.
The largest category is quadruped inspection robots, with 5,000 units budgeted at 1.5 billion yuan, mainly for substations, transmission lines and mountainous terrain.
Humanoid robots — used for live-line work — carry the highest unit cost, with 500 units allocated 2.5 billion yuan. Another 3,000 dual-arm robots, designed for equipment handling and fault response, are budgeted at 1.8 billion yuan.
Of the total, about 5.8 billion yuan will go to equipment procurement, with a further 1 billion yuan earmarked for R&D and training.
All equipment must meet State Grid technical standards, with preference for systems compatible with its in-house AI model and capable of local deployment for data security.
Suppliers include domestic robotics firms such as Unitree, Agibot, UBTECH Robotics and Fourier Intelligence, according to the same people.
The plan is also expected to benefit longtime suppliers to the power sector including Xianheng International, Yijiahe Technology and Shenhaotech.
"We are already seeing a clear shift this year," an executive at Xianheng International told Jiemian News. "Orders from state-owned enterprises have started to come through in the first quarter."
The company reported 28.88 million yuan in revenue from embodied AI-related products in 2025, and expects a sharp increase this year.
For many in the industry, the power grid offers a rare combination of stable demand, strong funding and highly standardised applications.
China has roughly 100,000 substations. Equipping each with one or two inspection robots implies a market worth hundreds of billions of yuan.
"The grid's controlled operating environment, strong demand for labor and ample funding give it a natural advantage for large-scale adoption," the source said.
The shift points to a broader move toward industrial deployment.
At a recent partner conference, Agibot co-founder PENG Zhihui said embodied AI is moving from technological breakthroughs to real-world deployment, with the focus shifting from digital content to physical-world use cases.
The power sector is emerging as one of the first large-scale testing grounds.
State Grid has incorporated embodied AI into its five core AI capabilities and identified more than 600 application scenarios across planning, operations and customer service.
The business case is also becoming clearer.
A human inspector costs about 300,000 yuan a year. A robot priced at around 1 million yuan becomes economical once costs fall to 600,000–700,000 yuan, industry executives said.
Official estimates suggest each unit could save 500,000–800,000 yuan annually, with a payback period of two to three years. Inspection efficiency could rise fivefold, while fault response times could fall by 60%. More importantly, robots could reduce exposure to high-risk work by more than 90%.
The push is backed by heavy capital spending. State Grid plans 4 trillion yuan in investment during the 2026–2030 period, up 40% from the previous cycle, as it builds a more digital and resilient grid.
Under the plan, robot penetration in key regions is expected to reach 30% by 2026, rising above 80% by 2027 and covering most high-risk tasks. By 2030, utilities aim to integrate robotics with digital-twin systems to enable mostly autonomous operations.
Pilot projects have already laid the groundwork.
State Grid signed a 1 billion yuan framework deal with robotics firm Deep Robotics in 2023, covering more than 1,000 substations. By 2025, deployment projects had exceeded 600.
The key shift now is scale. The new procurement plan marks a transition from small-batch trials to system-wide rollout, turning a traditionally conservative industry into one of the first large-scale testing grounds for embodied AI.