AI robot working in the office

Artificial Intelligence May Soon Require as Much Power as a Whole Nation

April 18, 2024

A.I. saw a boom when OpenAI’s ChatGPT burst into the scene, amassing an estimated 100 million users in just two months. The technology runs on thousands of specialized computer chips behind the scenes. (that is fantastic news for NVIDIA’s shareholders) They will likely consume enormous amounts of electricity in the upcoming years.

Some preliminary estimations are presented in a newly published, peer-reviewed analysis. According to a moderate estimate, artificial intelligence (AI) servers may require 85–134 terawatt hours (Twh) per year by 2027. 

That is roughly 0.5 percent of the world’s current electrical use, and is comparable to the annual usage of Sweden, the Netherlands, and Argentina .

The energy required for artificial intelligence (A.I.) will likely increase global carbon emissions, contingent on whether renewable energy or fossil fuels, power data centers.  

Here is a great plug for Solar 😉 

  • The most plentiful energy source on Earth is solar energy, which constantly bombards the planet with 173,000 terawatts of energy. That is over ten thousand times the amount of energy used worldwide.

If you are reading between the lines, it’s easy to see the necessity for current electrical grids to be supported by homeowners’ individual power generation.  There will have to be a major push to require homeowners to supply some of their own power or pay a premium.  

In 2022, data centers consumed 1 to 1.3 percent of the world’s electricity, which runs all computers, including Amazon’s cloud services and Google’s search engine. That does not include the 0.4 percent utilized for mining cryptocurrencies; however, some of those resources are currently being redistributed to operate AI.

The precise amount of energy used by artificial intelligence is impossible to measure because firms such as OpenAI reveal relatively little information, including the number of specialized processors required to run their software.

Keep Reading