BeijingChina targets 5 AI dispatch models in its grid by 2027.
The International Energy Agency projects AI-optimised grids could cut outages 50%; China’s data centres will consume 3-5% of generation by 2030.
Whether AI saves more energy than data centres consume is China’s central grid trade-off.
Sources: Dialogue Earth
WHY GRIDS NEED DISPATCH
Electricity cannot be stored at grid scale cheaply. Supply must match demand second-by-second, or frequency drifts and equipment trips offline. Dispatch is the continuous decision of which plants to run, at what output, in what order — the nervous system beneath every wall socket.
THE FLEXIBILITY CEILING
A grid built around coal baseload can absorb only so much intermittent solar and wind before stability breaks. Coal plants take hours to ramp; solar can drop 80% in minutes when clouds pass. The mismatch is why China curtails (wastes) tens of TWh of renewable output every year — the grid physically cannot accept it.
WHAT AI ACTUALLY DOES
Machine-learning dispatch forecasts load and renewable output minutes-to-hours ahead, then pre-positions thermal plants and battery storage to absorb the swings. The gain is not invention — it is reaction speed. Human dispatchers update schedules every 15 minutes; AI updates continuously.
THE JEVONS PROBLEM
Every efficiency gain in energy history has been swallowed by rising demand. Watt's steam engine used less coal per horsepower and Britain's coal use exploded. LED lighting cut bulb wattage 80% and global lighting electricity rose. AI optimising the grid while AI data centres consume the grid is the same pattern, compressed.
WHY CHINA MOVES FIRST
China owns the world's largest grid by load (over 2.5x the US), the largest installed solar and wind capacity, and a centralised State Grid Corporation that can mandate national protocols. Western grids are fragmented across regulators and private utilities; a single AI dispatch standard takes years of approvals. Beijing can decree one.
THE HISTORICAL ECHO
The 1965 Northeast Blackout cascaded across eight US states in 13 minutes because relays acted faster than humans could coordinate. The response was SCADA — supervisory control and data acquisition — the first generation of automated grid control. AI dispatch is the third generation, after SCADA and the 2000s-era smart grid.