top of page

AI needs electricity: where will it come from?

In fact, the increased demand for electricity is driven by many factors: electrification of transportation, buildings, colder weather in certain parts of the world .

AI is a massively power-intensive workload.

in 7 years, US data center power demand increased by 80%

According to Stephen Oliver, vice president of corporate marketing and investor relations at Navitas Semiconductor, the demand for power by US data centers increased from 19 GW to 35 GW.

The initial demand for additional power will be driven by training workloads. Training AI algorithms is not a perfect science. Expect more demand for more power.

Dominion Energy serves the largest datacenter market in the world in Loudoun County, Virginia, about 30 miles west of Washington, DC.

According to Dominion Energy, electricity demand from datacenters in Virginia increased by about 500% from 2013 to 2022.

The demand for power as a result of wider deployment of AIU applications will be difficult to forecast.

Let's explore why.

Two types of AI workloads: exploratory and maintenance

Exploratory AI workloads are created by humans seeking answers from the data. The same question can be asked many times, affecting 1 million rows of data or 10 billion rows of data.

It's likely that the cost of developing and supporting exploratory AI workloads will increase.

Maintenance AI workloads are created by humans to perform a narrow, well defined task, such as forecasting, predictive maintenance, detection of certain objects.

Both will drive exponential demand for electricity and power - and commodities, such as metals and rare earths. More servers and batteries will require more commodities.

The US government identified critical 50 minerals

These 50 minerals are critical to both the US economy and national security.

According to Carlos De Alba, managing director at Morgan Stanley,

"Since 1990, the US has seen a significant increase in both the variety of imported minerals and the level of dependence on these imports. In 1990, the US was fully reliant on imports for the supply of 9 minerals, and imports exceeded 50% of consumption for 27 minerals (including all minerals, not only critical ones)."

"These figures have risen to 15 and 51, respectively, as of 2022. China currently dominates the global mineral processing/refining industry, and the US is heavily reliant on China (>50% of net import reliance) for imports of 26 minerals."

The investment levels in the US metals and mining sector industry have reached their lowest level in decades.

60% of electricity generated in the US comes from fossil fuels

According to the US Energy Administration,

"In 2023, about 4,178 billion kilowatt-hours (kWh) (or about 4.18 trillion kWh) of electricity were generated at utility-scale electricity generation facilities in the United States.1 About 60% of this electricity generation was from fossil fuels—coal, natural gas, petroleum, and other gases. About 19% was from nuclear energy, and about 21% was from renewable energy sources."

Greater demand for electricity = continued demand for commodities

Greater use of AI application will drive greater demand for more power, electricity, and commodities.

Commodities ran, run, and will run the world

TerraManta is the only company which applies machine learning to forecast commodity prices by continuously learning how geopolitical events influence fundamentals (supply, demand) and result in a price action as evidenced by markets behavior.

TerraManta welcomes serious inquiries from investors who want to see machine learning solve real problems in today's world.

4 views0 comments

Recent Posts

See All


Commenting has been turned off.
bottom of page