Can AI Data Centres & Sustainable Energy Actually Coexist?
Representatives from EPRI, GE Vernova and Schneider Electric discuss the data centre power bottleneck, energy efficiency and the liquid cooling water myth
TL;DR
- Rack density is doubling annually, increasing power demands for AI data centers.
- Grid connection queues in the US and UK have a 500% backlog, with wait times up to 7-10 years.
- Data center construction takes 12-24 months, but grid connection can take up to 7-10 years.
- On-site power solutions like natural gas, fuel cells, and battery storage are being used to bypass grid bottlenecks.
- GE Vernova faces a 100GW gas backlog, with new gas turbines potentially shipping as early as 2031.
- EPRI's DC Flex initiative helps data centers act as flexible grid resources.
- Data centers can manage energy usage by categorizing workloads by priority or relying on on-site microgrids.
- European data centers face regulations like Germany's heat reuse law, though practical application is challenging.
- TeraWulf's Lake Mariner campus utilizes 89% zero-carbon hydropower from Niagara Falls.
- Liquid cooling systems can reduce data center water consumption by up to 53%, and closed-loop systems require zero water usage for cooling.
- High-efficiency closed-loop liquid cooling systems eliminate water consumption by radiating heat directly into the air.