Infrastructure · Trend
Virtual Power Plants for AI Data Center Energy
Grid integration and flexible energy management needs driven by spiky and oscillatory AI training loads at data centers.
Forecast horizon
Grid architecture and coordinated load design become first-order constraints on AI cluster growth.
Connections
Connections · 4
How this node ties into the rest of the map, and the evidence behind each link.
to · informs 4/10
UAE Data Center Advantage for AI Infrastructure Both trends address the energy and infrastructure constraints shaping where AI compute capacity is built globally.
+4 growth from · supports 4/10
Nuclear Power Milestones Supporting AI Energy Demand Laser enrichment and nuclear fuel-cycle advances are covered alongside broader efforts to secure energy for compute growth.
+4 growth from · informs 3/10
Rising Risk of Weather Data Sabotage Weather-data integrity risk affects grid and infrastructure operations that also underpin AI energy planning.
+3 growth from · publishes 3/10
MIT Technology Review MIT Technology Review How-To explains signing up for virtual power plants amid flexible energy demand.
+3 growthSignal sources
Signal sources
Dated facts from primary sources in this direction.
Compute scaling 2024
Training compute for frontier AI models grew ~4–5× per year from 2010 to 2024 — a trend Epoch projects to continue toward 2030.
Epoch AI → Data-center energy 2025
Global data-centre electricity use is projected to roughly double from ~485 TWh (2025) to ~950 TWh by 2030 — about 3% of world electricity.
International Energy Agency → Inference cost collapse 2022–2024
The price to query a GPT-3.5-level model fell from $20.00 to $0.07 per million tokens between Nov 2022 and Oct 2024 — a 280× drop in ~18 months.
Stanford HAI — AI Index 2025 → Open-weight frontier Jan 2025
DeepSeek disclosed its open-weight R1 reasoning model's final training run cost just $294,000; R1 was released under open weights on 20 Jan 2025.
Nature / DeepSeek →