​The Industrial Revolution Of Intelligence: AI Energy Infrastructure

Bill Wong - AI Research Fellow, Info-Tech Research Group.

​​In my previous article, we examined the emergence of a new data center asset, the AI factory, which is optimized to run AI workloads. Compared to traditional data centers, AI data center campuses require up to 10 times the total facility power capacity, driven by GPU server racks that consume 12 to 28 times more power per rack and thermal density that requires advanced liquid cooling to handle five times the heat generation.

This new industrial revolution is driving an unprecedented demand for power, transforming the utilities industry from a relatively stable and quiet industry to a growth sector.​ Nations and organizations alike will need to optimize their AI energy strategy in order to flourish during and after this revolution.

National AI Energy Infrastructure Strategies

After reviewing several national AI energy infrastructure strategies around the world, four distinct strategies stand out. These strategies will shape how nations are planning to build the required energy infrastructure for AI workloads.

• Sovereign Regulation-Driven: These jurisdictions introduce legislation to ensure that AI developers must fund their own power requirements, maintain the reliability of the grid and provide the supporting water infrastructure required for these data centers to prevent rising utility costs. In Canada, the country’s recently published AI strategy champions their grid’s 83% investment in clean power. Australia will require data center providers to generate as much energy as they consume. The EU has introduced a sovereign AI compute platform, EuroHPC JU, which aims to “develop innovative energy-efficient HPC software technologies tailored to exascale and post-exascale supercomputers.”

• State-Driven: China’s government is planning to connect renewable energy resources from its western region to its computing hubs in its eastern region.​