Your cart is currently empty!
Figure 1 shows various forecasts from multiple sources for power consumption by AI data centers through 2030, revealing a wide range of outcomes and highlighting several ongoing uncertainties:
- Data Centers demand will depend on the successful adoption and widespread use of AI applications at reasonable cost levels. Future demand for data center services—such as AI, video streaming, IT infrastructure, and communications—will largely be a function of these AI fundamentals.
- Software and data processing technologies may improve over time and require less power; for example, newer AI models could become more efficient and need fewer computational resources.
- Computer hardware, particularly chipsets used for AI data processing, is continually advancing and may also reduce power requirements.
- Cooling technologies for data centers are expected to become more efficient. Additionally, colocation with industrial facilities that utilize waste heat for processes could further improve overall energy efficiency.
- By 2030, projections range from around 206 TWh (EPRI historical) to approximately 970 TWh (BCG). Incorrys assesses that power consumption is likely to be near the lower end of this range.