AI is all the rage, until we consider its carbon footprint

AI's carbon footprint has been skyrocketing in recent years, primarily due to the exponential growth in computational power and data requirements. We’ve all heard about the rising energy footprint of bitcoin, due to the computational power for each new block, but what about AI?

I’m interested here because of the necessary innovation in energy that will follow. As I look at the AI revolution, I’m thinking about the “picks and shovels”. We all know that NVIDA is one of those beneficiaries, but energy is the key component powering everything. That’s what fascinates me.

The training of AI models demands massive computational resources, often relying on powerful data centers with thousands of servers running continuously.

These data centers consume vast amounts of electricity, contributing to a significant portion of the carbon emissions. According to a study published in Nature, training a single large AI model can emit as much carbon dioxide as five cars during their entire lifetimes.

The rapid expansion of AI applications and the proliferation of connected devices generate an ever-increasing volume of data. Processing and analyzing these colossal data sets necessitate extensive computational infrastructure, further intensifying energy consumption.

As more businesses and industries embrace AI technologies, the demand for data centers and cloud computing services surges, leading to a corresponding surge in carbon emissions.

A report by the International Energy Agency estimates that data centers account for approximately 1% of global electricity consumption, and their carbon footprint is projected to continue rising.

The energy requirements for training and deploying AI models are not limited to the computational infrastructure alone. Consumption of AI is also tied to the life cycle of the hardware used, including the production and disposal stages.

Manufacturing the specialized hardware, such as graphics processing units (GPUs) and application-specific integrated circuits (ASICs), requires significant amounts of energy and raw materials.

Additionally, the disposal of outdated or malfunctioning AI hardware poses environmental challenges, as e-waste management often involves complex processes and can contribute to pollution if not handled properly.

The surge in AI's carbon footprint is attributed to the substantial computational resources needed for training AI models, the exponential growth of data-driven applications, and the energy consumption associated with the life cycle of AI hardware.

As AI continues to advance and permeate various sectors, addressing its environmental impact becomes crucial. Efforts are underway to develop more energy-efficient algorithms, optimize data center operations, and promote sustainable practices in the AI industry to mitigate the carbon footprint and foster a greener future.

This feeds into my overall investment thesis (Missing the forest for the trees). I’m becoming even more convinced that the transition away from fossil fuels will gather pace in the coming years and countries in the gulf, like Saudi Arabia, see the writing on the wall. I wouldn’t be surprised to see oil at $20-30 per barrel in the near term, despite supply constraints.

Falling oil prices, energy innovation and AI, are all massively deflationary in the medium to long term.

Peter Esho is an economist and Founder of Esho Group. He has 20 years of experience in investments and markets.