DeepSeek and the Million Dollar Question about Data Centers and Distributed Energy

Source: Elisa Wood | · LINKEDIN · | February 9, 2025

AI is driving up energy demand, but how much and how fast remains under debate. DeepSeek heightened the discussion in January with its release of a new large language model.

The Chinese AI company purports that its model performs as well as US models but uses far less computing power and, therefore, less electricity.

This is a big deal because the power industry has been planning and investing based on the idea that AI will increase electric demand significantly.

Were the energy bears right all along in arguing that bullish energy forecasts failed to account for the inevitable efficiencies that occur as technology evolves? Are utility energy demand forecasts inflated? Exactly how much influence will DeepSeek have on electricity demand?

Answering this question requires a dive into the world of data centers.

So, for this episode of the Energy Changemakers Podcast, I spoke to Rich Miller former editor and founder of Data Center Knowledge and Data Center Frontier, about the meaning of the DeepSeek announcement. After 25 years of tracking the industry, he offers a breadth of insight into data center innovations and their increasingly important role in the energy arena.

Miller also talks about the threat data centers could pose to the utility model as highly capitalized tech companies increasingly choose onsite energy over utility contracts. And finally, he offers insight into what motivates energy decisions by data center operators.

Listen until the end, when Miller unveils the results of a poll he conducted with his data center industry followers about AI energy demand. The numbers surprised me.