Wood Mackenzie's 363 TWh Robotics Power Forecast
The global robotics sector will consume 363 terawatt-hours of electricity by 2035, according to a Wood Mackenzie report, roughly matching the 373 TWh generated by France's nuclear fleet in 2025. Industrial robots account for 357 TWh of the projected demand, while humanoid robots add 6 TWh over that period as they grow from a small base.
The consultancy describes robotics as a second and largely unaccounted-for source of electricity demand at a time when AI data centres are already straining energy capacity planning. In 2025, five million industrial robots worldwide consumed 78 TWh, equivalent to 20-25% of data centre electricity demand. Wood Mackenzie projects the industrial robot fleet to reach 16 million by 2035, assuming average annual growth of about 12%. Annual installations rose from 200,000 in 2015 to 500,000 in 2025 and are expected to pass one million in 2032.
China dominates humanoid deployment: almost 90% of humanoid robots now in operation were produced by Chinese companies. Average humanoid prices fell 93% between 2020 and 2025 to $58,000. Unitree Robotics' G1 now sells for $16,000 and, according to the report, uses about $82 of electricity a year when run eight hours a day at average global industrial power prices, far below the cost of human labour on most markets. Wood Mackenzie's director of integrated energy research, Robert Liew, says energy supply constraints are becoming a real brake on robotics adoption even as developed economies face labour shortages.
Why Robotics Is Becoming a New Load on Global Power Grids
A second data-centre-scale load
The comparison to data centres is the report's central warning. Power planners already treat AI computing demand as a structural load; Wood Mackenzie argues robotics should be modelled separately, not hidden inside industrial consumption. The 2025 baseline of 78 TWh from five million industrial robots equals one-fifth to one-quarter of data centre demand, and the projected 363 TWh by 2035 would be nearly five times that baseline. That is not a marginal addition but a national-scale power requirement.
China's pricing advantage in humanoid robots
The numbers point to a sharp cost-driven shift in the humanoid market. With Chinese firms supplying nearly 90% of operational units, the average price has collapsed 93% in five years. Unitree's $16,000 G1 is less than a third of the $58,000 average, suggesting Chinese manufacturers are already setting the price floor. For Western competitors, this is a margin problem as much as a technology race: a robot that costs a few tens of dollars a year to power competes on labour cost, not just capability.
Energy limits as the adoption bottleneck
Liew's framing shows that industrial robots already consume 78 TWh before humanoid robots reach true scale. If grid connections and power supply do not expand in step, robotics adoption in factories and warehouses will be constrained by electricity availability rather than by machine price or software. That is especially relevant in developed labour markets where automation would otherwise fill shortages. The projected 90%-plus annual growth in humanoid units assumes industrial and commercial users can actually connect and power them.
What the 2050 scenario reveals
The long-run benchmark is striking: if the global humanoid fleet reaches one billion units by 2050, their power demand alone could match South Korea's 2026 electricity generation. This is a hypothetical, not a central forecast, but it shows how a sharp price decline plus labour-substitution economics could create a demand shock much larger than the 2035 figure.
What the Forecast Means for Grid Planners and Robot Buyers
- For grid planners and utilities, model robotics as a separate load category in 10-15-year capacity plans: 363 TWh by 2035 is nearly all of France's 2025 nuclear generation, and robotics is arriving as a second structural demand source alongside AI data centres.
- For industrial companies assessing humanoid robots, update total cost of ownership models with the new price points: Unitree's G1 at $16,000 and about $82 of annual electricity at eight hours per day is below most market labour costs, but only where the site has reliable power and grid capacity.
- For businesses in developed markets facing labour shortages, begin power-supply due diligence when planning large robot deployments, because Wood Mackenzie identifies energy constraints as the main bottleneck to adoption, not machine cost.
- For energy investors and infrastructure developers, the 2025 baseline of 78 TWh, equal to 20-25% of data centre demand, indicates a large and still largely unmodelled demand pool for generation, grid connections and energy-management systems.
Risk & Opportunity Assessment
| Commercial Risk | Medium | A 363 TWh demand pool by 2035, concentrated in industrial robots, creates revenue and capacity-planning risk for utilities and industrial buyers if power supply is not expanded in parallel. |
| Competitive Risk | Medium | China supplies almost 90% of operating humanoid robots, and Unitree's $16,000 G1 against a $58,000 average shows aggressive Chinese pricing could compress margins for non-Chinese robot makers. |
| Regulatory Risk | Medium | Energy permitting and grid connection rules may become the binding constraint, and policymakers have not yet counted robotics in demand models, creating approval risks for new capacity. |
| Reputation Risk | Low | Direct reputational exposure is limited today, but utilities that fail to model robotics demand may face reliability criticism if grid constraints slow automation. |
| Technology Disruption | High | A projected humanoid fleet compound annual growth rate above 90% to 2035, combined with a 93% price decline, could rapidly shift automation adoption and electricity load profiles. |
| Commercial Opportunity | High | Robotics becomes a second large load after AI data centres, creating multi-decade demand for generation, grid connections and energy-efficiency solutions. |
Comments 0