How to Invest in AI's Demand for Energy

How much energy does AI use?

The math on this differs based on the source, but the consensus is that it's a lot. Goldman Sachs quantified this: a ChatGPT query needs nearly ten times as much electricity to process as a Google search, making it very inefficient.

Why is this? According to the World Economic Forum (WEF), 80% of this demand comes from the "inference" phase of the AI lifecycle where the model solves problems, while the learning phase where they analyze the data takes up 20%.

The WEF further notes that the energy required to run AI has an annual growth rate between 26% and 36%. By 2028, AI could use more power than Iceland did in 2021.

Forbes reports that one of the most popular generative AIs commercially available, GPT-4, requires 50 gigawatt-hours. To put it in perspective, Forbes noted that this was 0.02% of the electricity California generates in a year and 50 times the energy requirements of GPT-3.

The culprit behind this immense energy consumption is the backbone of AI infrastructure: data centers. According to Forbes, data centers account for 3% of global energy consumption and emit the same amount of CO2 as Brazil.

Goldman Sachs concurs, predicting an overall increase in data center power consumption from AI to be on the order of 200 terawatt-hours per year between 2023 and 2030. By 2028, AI is expected to represent about 19% of data center power demand.

Additionally, the International Energy Agency (IEA) forecasts that electricity consumption from data centers and AI will double by 2025, globally consuming an estimated 460 terawatt-hours (TWh) in 2022, rising to 1,000 TWh by 2026, which is equal to Japan's total electricity consumption.

Where will the energy to power AI come from?

There is a significant push toward "sustainable AI," aiming to ensure its energy comes from renewable, clean sources. This aligns with broader global trends towards renewable energy adoption.

For instance, the IEA projects that low-emission sources—including nuclear and renewables like solar, wind, and hydro—will cover half of the world's electricity generation by 2026, up from 39% in 2023.

They also believe that renewables are set to provide more than one-third of total electricity generation globally by early 2025, largely replacing fossil fuels like thermal coal, thanks to lower solar costs.

However, solar is only part of the equation. A big driver of this transition is nuclear generation, which is forecast to grow by close to 3% per year on average through 2026 due to new reactors in China, India, Korea, and Europe, and the recommissioning and restarts in France and Japan.

"Many of the technology companies that are building the large data centers required to power AI have clean energy goals and are increasingly turning to nuclear energy to secure reliable baseload power," says Steve Schoffstall, Director of ETF Product Management at Sprott Asset Management USA.

Moreover, the IEA notes that during the December 2023 COP28 climate change conference, more than 20 countries signed a joint declaration to triple nuclear power capacity by 2050. This highlights the increasing acceptance of nuclear energy as a stable and significant source of clean power.

"The uranium needed to feed nuclear reactors may likely be in a supply deficit for the foreseeable future, with a cumulative deficit of one billion to 2.1 billion pounds through 2040," Schoffstall explains. "In our view, uranium miners are likely to benefit from the increased demand for nuclear energy."

The benefit of this shift therefore extends to the materials sector rather than the traditional energy sector companies involved in oil and gas or thermal coal.

Key raw inputs include not only uranium, but also lithium, copper, nickel, silver, manganese, cobalt, graphite, and rare earth elements, which are essential for clean energy technologies such as solar panels, nuclear reactors, wind turbines, batteries, and cabling.

"Popular in consumer electronics and electric vehicles (EVs), lithium-ion batteries may see increased demand from AI data centers as the need for energy storage increases," Schoffstall notes. "Though lithium is currently in a supply surplus, we see the potential for large lithium supply deficits in the coming years from both EVs and grid battery storage." 

Which ETFs can I use?

The buy-it-all thematic ETF to watch here is the Sprott Energy Transition Materials ETF

. For a 0.65% expense ratio, it holds companies producing all these materials, offering a globally diversified portfolio balanced between large, mid, and small caps.

But, if you prefer to slice-and-dice your exposure, Sprott also has numerous other ETFs corresponding to select materials found in SETM – including the two highlighted earlier by Schoffstall:

• Sprott Uranium Miners ETF
  URNM
  -1.72%
• Sprott Copper Miners ETF
  COPP
  -0.82%
• Sprott Lithium Miners ETF
  LITP
  +1.43%
• Sprott Nickel Miners ETF
  NIKL
  -0.68%

Please note this article is for information purposes only and does not in any way constitute investment advice. It is essential that you seek advice from a registered financial professional prior to making any investment decision.