| | | | | |

AI Deployment – Contributing to Global Energy Crisis

ByBrainTrust

The energy consumption of Artificial Intelligence (AI) systems, both for training and operational use, has become a significant concern as these technologies continue to advance and expand. Current estimates indicate substantial power requirements for AI, with implications for global energy demand and environmental impact.

The Choice is in Your Hands

Rather than give you just the facts, I feel that since this subject is never mentioned when you make purchasing decisions, or signing up to AI services. The mantra “You must use AI in your everyday life” along with advertising every new PC as “AI enabled”, is akin to enticing you to use a drug but never telling you the side-effects.

In this article I will summarise the alarming “side-effects” for this phase in human achievement, for you. It is up to you to decide whether using AI conflicts with your drive to “be good to the environment” or “save our scarce energy reserves” or “going green!”.

Ultimately, we do not need AI to function as a prosperous society. The push by the World Economic Forum (WEF) to deploy AI into all aspects of your life should be looked upon with the view that this will increase consumption of our most valuable energy reserves. The addition to AI features into domestic products such as electric cars, PCs and household devices comes at an extra cost for both purchase and running which is beyond the reach of many people in our society. This effectively increases the gap between the classes within most western societies.

Remember that you can buy an older model iPhone and a refurbished PC running Linux and be perfectly able to function in a technological society. And this is for as little as US$300 in total! To complement this purchase, all you need is a desire to learn, read, explore and think for yourself. Your brain is many times more powerful than any AI today and takes a lot less energy to run. The are many articles on this website to help you get started.

The True Cost of AI Today (November 2024)

Training AI Models

Training large AI models is extremely energy-intensive:

  • Training a model like GPT-3 is estimated to use approximately 1,300 megawatt hours (MWh) of electricity, equivalent to the annual power consumption of about 130 US homes[1].
  • The more advanced GPT-4 is estimated to have used 50 times more electricity for training than GPT-3[1].
  • A single AI model’s training can use more electricity than 100 US homes consume in an entire year[3].

Running AI Systems

The operational energy use of AI systems is also substantial:

  • A single ChatGPT query requires about 2.9 watt-hours of electricity, compared to 0.3 watt-hours for a Google search[2].
  • Generating one image using a powerful AI model consumes as much energy as fully charging a smartphone[5].
  • Creating text 1,000 times with AI uses about 16% of a full smartphone charge[5].

Projected Energy Demand

The energy demand for AI is expected to grow significantly:

  • Goldman Sachs Research estimates that data center power demand will grow 160% by 2030, largely due to AI[2].
  • Wells Fargo projects AI power demand to surge 550% from 8 TWh in 2024 to 52 TWh in 2026, and then increase another 1,150% to 652 TWh by 2030[4].
  • AI could represent about 19% of data center power demand by 2028[2].

Comparisons

To put these figures in perspective:

  • Data centers currently consume 1-2% of global power, expected to rise to 3-4% by 2030[2].
  • In the US, data centers are projected to use 8% of power by 2030, up from 3% in 2022[2].
  • Nvidia’s 3.76 million GPU shipments in 2023 could consume as much as 14.38 TWh annually, equivalent to the power needs of over 1.3 million US households[4].

Environmental Impact

The increased energy consumption has environmental implications:

  • The carbon dioxide emissions of data centers may more than double between 2022 and 2030[2].
  • The “social cost” of this rise in emissions is estimated at $125-140 billion (at present value)[2].

While these figures are concerning, it’s important to note that AI may also contribute to energy efficiency in some sectors, however, these are too few to make a difference. Many more sectors are adopting AI with an illusion that they will achieve efficiency improvements.

Citations:
[1] https://www.weforum.org/stories/2024/07/generative-ai-energy-emissions/
[2] https://www.goldmansachs.com/insights/articles/AI-poised-to-drive-160-increase-in-power-demand
[3] https://www.iea.org/commentaries/why-ai-and-energy-are-the-new-power-couple
[4] https://www.forbes.com/sites/bethkindig/2024/06/20/ai-power-consumption-rapidly-becoming-mission-critical/
[5] https://www.technologyreview.com/2023/12/01/1084189/making-an-image-with-generative-ai-uses-as-much-energy-as-charging-your-phone/
[6] https://www.eidosmedia.com/updater/technology/the-energy-demand-of-gen-ai-models-a-growing-concern
[7] https://think.ing.com/articles/ai-monthly-hungry-for-power/
[8] https://e360.yale.edu/features/artificial-intelligence-climate-energy-emissions

Similar Posts