The global robotics industry will consume as much electricity as France generates from nuclear power within a decade, according to Embodied AI How Robotics Are Accelerating Global Power Demand, new Wood Mackenzie research. The finding arrives as energy capacity planning is already under pressure from AI data centres, making robotics a second and largely unmodelled demand on global electricity infrastructure. Wood Mackenzie puts combined robotics power consumption at 363 TWh by 2035, with industrial robots accounting for 357 TWh in the upper range and humanoid robots adding a further 6 TWh annually.
The industrial robotic market’s growth by 2035 is drawing power and approaching data centre scale. The 5 million robots in operation consume 78 TWh a year, almost twice London’s electricity consumption. Wood Mackenzie expects the operational fleet to reach 16 million units by 2035, if it can maintain the current pace of expansion at 12% a year.
No single country shapes this market more than China. It accounts for over 70% of annual global industrial robot installations, and Chinese companies are responsible for nearly 90% of all humanoid units currently in deployment. China State Grid illustrates that position directly: it is spending US$1 billion in 2026 to buy 8,500 AI-enabled autonomous robots for more than 600 specialised tasks, from routine grid maintenance to live-line work on ultra-high voltage power lines.
Humanoid robots are still early in rollout but growing fast. If the global market reaches one billion units by 2050, power demand from humanoid robots alone could match South Korea’s entire electricity generation in 2026.
Increasingly competitive prices are pulling more buyers in. Average humanoid robot prices fell 93% between 2020 and 2025 to US$58,000. China’s Unitree Robotics’ G1 is now priced at US$16,000, with an estimated annual electricity cost of US$82 per unit at eight hours a day (based on a global average industrial tariff of US$0.14 per kWh.), significantly below human labour costs in most markets despite it being mostly used in limited settings.
“Power constraints are becoming a real brake on robotic adoption, and that matters because labour markets in developed economies are running short of alternatives,” said Robert Liew, Director, Integrated Energy Research, Wood Mackenzie. “Industrial robots already draw 78 TWh a year globally, and that is before humanoid robots reach any real scale. By 2035, combined demand could hit 363 TWh.”
Key findings:
- Combined global robotics electricity demand reaches 363 TWh by 2035 (Wood Mackenzie base case), approaching France’s nuclear generation of 373 TWh in 2025. Industrial robots account for 357 TWh with humanoid robots contributing 6 TWh annually. Both figures compound AI data centre demand already stretching global power infrastructure.
- The 5 million industrial robots operating in 2025 draw 78 TWh annually, equivalent to 20-25% of global data centre demand; the fleet reaches 16 million units by 2035 assuming 12% CAGR trend is maintained. Annual installations grew from 200,000 (2015) to 500,000 (2025) and are forecast to exceed 1 million by 2032. (Source: International Federation of Robotics, Wood Mackenzie)
- Humanoid robot stock grows at over 90% CAGR from 2025 to 2035, reaching 10 million+ units with annual shipments exceeding 4 million; average unit price fell 93% between 2020 and 2025 to US$58,000. Unitree G1 (2026 price: US$16,000) carries an estimated annual electricity cost of US$82 at eight hours a day, based on a global average industrial tariff of US$0.14/kWh. (Source: Unitree, Wood Mackenzie)
- China accounts for over 70% of annual global industrial robot installations and nearly 90% of all humanoid units deployed in 2025. China State Grid has committed US$1 billion in 2026 to procure 8,500 AI-enabled autonomous robots across more than 600 specialised tasks, signalling that robotics has become critical national infrastructure.
Background:
Robotics demand has grown for over a decade, driven by structural labour shortages, rising labour costs, and supply chain security concerns. AI advances in computer vision, large language models, and reinforcement learning have extended robot capability into logistics, healthcare, and field operations. Most industrial robots draw power directly from the local grid, without low-carbon sourcing mandates. Robotics training infrastructure is captured within data centre demand estimates and excluded here to avoid double counting.
