GenAI and Surging Energy Demand

A few weeks ago, I talked with my colleagues Angie (Evans) Gildea and Mark Gibson about the rapid increase in Generative AI use, and thus data centers, and how this is impacting energy demand. You can replay the conversation here or check out some of the highlights from our conversation below.

1. Energy demand from data centers is skyrocketing

Data center usage is booming, and therefore, so is energy consumption. Data centers are energy-intensive, consuming 10 to 50 times the energy per floor space of the average commercial office building. The International Energy Agency (IEA) estimates that currently, there are approximately 8,000 data centers worldwide. That number is changing – and fast. The IEA also estimates global electricity consumption by data centers could double between 2022 and 2026, effectively adding the equivalent power consumption of Sweden or Germany.

1. One energy source alone will not supply this demand

Electricity supply has become the new bottleneck for scaling AI. It is simply not possible for one energy source alone to meet this growing demand. One thing that also cannot be emphasized enough in this conversation is that people need to have reliable, affordable, sustainable energy. This is going to take a mix of traditional and renewable resources on the supply side. Each energy source comes with its own set of advantages and drawbacks. In the U.S., for example there has been a renewed focus on nuclear energy due to its density. However, educating Americans, who tend to prefer renewables and fossil fuels over nuclear, on its benefits is an imperative. While renewables, like solar and wind have broad appeal among consumers, these power sources are more feasible in certain locations and less feasible in others due to weather, permitting, and other constraints. So far, natural gas been seen as current as the most reliable, affordable, and sustainable option serving as a bridge to greener sources.

1. Permitting remains a major challenge

New energy sources require new permitting. Obtaining local, state and federal approvals to construct new energy projects and connect the new energy to the grid can take six to ten years – and in some cases, even longer. This can hinder the efficient integration of renewable generation into the grid.

Permitting challenges pose a significant obstacle to scaling renewable energy, both in the U.S. and globally. We discuss this at length in KPMG’s Turning the tide in Scaling Renewables. Governments are increasingly recognizing the need to intervene and implement legislative changes to address the climate crisis and remove barriers to renewable energy expansion.

1. Tech has entered the energy business

A recent KPMG survey showed that business leaders overwhelmingly see technology, media & telecommunications (TMT) as the sector leading in GenAI adoption (91%). Tech and data companies with large data centers are also quickly coming to the understanding that they are now in the energy business. Not only are they thinking about different forms of renewable energy to procure, but also how to make data centers more efficient or reduce their overall energy consumption. How data centers are designed, heated, and cooled at this scale has become a major focus. Significant amounts of venture capital is also flowing toward new tech platforms that show promise for providing energy efficiency solutions.

None of these problems we discussed are easy, but they are solvable. With the understanding that this problem is too big to solve alone, new partnerships between tech and energy companies are emerging, paving the way for a more sustainable future.