Generative AI and Derived Demand

The demand for power to support generative AI is an excellent example of derived demand in economics. Derived demand occurs when the demand for a good or service arises not from its direct use but because it is required for the production of another good or service. The increasing demand for power is well explained in this new research piece from ING Economics.

Here’s how the demand for power to support generative AI fits into this concept:

1. Indirect Demand for Electricity

• Generative AI does not directly create demand for electricity for its own sake. Instead, the demand for electricity comes from the need to operate data centers, run AI models, and process vast amounts of data. In other words, the need for electricity is derived from the growing use of AI technologies that require immense computing power to function effectively.

2. Infrastructure as a Prerequisite

• The rise of generative AI models—used for tasks like text, image, and video generation—requires a substantial infrastructure backbone, including data centers that consume vast amounts of electricity. As generative AI becomes more embedded in various industries, the demand for electricity is a necessary input to keep these systems operational. The more AI expands, the greater the need for energy resources.

3. Demand Chain Linked to AI Usage

• The demand for power is directly linked to the expansion and application of AI technologies. As businesses and industries adopt AI models to enhance productivity, creativity, and automation, their need for electricity to run these systems grows. This makes electricity a secondary but crucial component in the chain of production for AI-driven services.

4. Economies of Scale and Efficiency

• In economic terms, large data centers supporting AI benefit from economies of scale, reducing the per-unit cost of computing power. However, these economies also drive up the total demand for electricity because as data centers grow in size to meet AI's increasing needs, their energy consumption rises.

• AI is expected to double its energy consumption by 2026 due to the growing sophistication and application of generative models. This trend showcases how energy demand is derived from AI advancements rather than energy being the final product of consumption.