TLDR

• According to the International Energy Agency's (IEA) 2024 electricity report, electricity consumption from AI, data centers and cryptocurrency is projected to more than double from 460 terawatt-hours (TWh) in 2022 to 1000 TWh in 2026.

• Putting things into perspective, between 2010 and 2018, data center energy usage has been fairly stable, accounting for around 1-2% of global consumption. However, this trend has shifted dramatically, with data centers now contributing to 3% of global consumption in 2023 and expected to rise to 4% by 2030. 1% difference accounts for electricity supply that can power 40 million US households for a year. This surfaces concerns for carbon emissions, energy shortages, and the broader implications of rising living costs for consumers.

• On a positive note, renewable energy is set to provide more than 1/3 of total electricity generation globally by early 2025, overtaking coal. Energy experts remain optimistic that technology advancements in both renewable energy production and the optimization of AI’s energy utilization will reduce costs and preserve the environment. However, these benefits may not be felt equally across all regions. In places with less sophisticated energy infrastructures, such as Taiwan, we can expect both direct and indirect impacts on daily life.

Estimated read time: 5mins

The Unfold

“Where are you going tonight, Frank?”

Currently, AI products are widely employed for generating code, text, and images, analyzing data, automating tasks, improving online platforms, and much more, with usage anticipated to grow even further. However, these advanced models demand substantial computing resources for both training and inference. This computing power relies on extensive arrays of sophisticated hardware located in industrial-scale facilities, which in turn require significant access to energy, water, broadband, and other essential infrastructure to operate effectively.

Interestingly, energy is often the most overlooked asset in this AI boom. While many focus on data and chips as the primary drivers of AI innovation, it's startling to consider how little attention is given to energy—the fundamental resource that makes all of this possible. Addressing this "elephant in the room" is essential, as the energy demands of AI operations cannot be ignored if we wish to sustain this technological advancement.

The Big Chess Moves

Summer of 2024 has turned into a bustling energy shopping spree season for hyperscalers as they scratch their heads to grapple with the insane energy demands driven by the AI Revolution. Major players like Amazon, Google and Microsoft have made significant commitments to nuclear to brace for the impending surge in demand.

Zooming in to Taiwan, the land of semiconductors, we are witnessing pressing energy deployment challenges. In 2023, the total electricity production in Taiwan reached 335.6 TWh, with a concerning distribution as we approach 2025:

• 81.8% of electricity comes from fossil fuels (oil, gas, and coal).

• Only 18.2% is generated from clean, renewable sources, including nuclear, hydro, wind, and solar.

This imbalance will worsen as Taiwan plans to phase out all nuclear power plants by 2024-2025, with no clear strategy to incorporate new renewable energy sources into the mix.

In light of these developments, it’s no surprise that hyperscalers like Google are investing into solar energy solutions in Taiwan to revitalize this overburdened energy system, with Amazon and Microsoft following suit. This strategy is wise for several reasons:

1. Diversifying energy sources helps companies minimize their vulnerability to fluctuations in traditional energy prices and enhances their ability to adapt to energy market volatility, ultimately stabilizing energy supply and reducing business risks.

2. Procurement of renewable energy opens up access to incentives such as tax credits and government subsidies, alleviating some economic pressure related to carbon emissions.

3. Supporting sustainability initiatives like RE100 is crucial. Currently, Google’s data center in Taiwan only has a 18% hourly match rate with carbon-free energy, in comparison to 76% on average in the EU.

As Amazon and Microsoft establish their data centers in 2025 and 2026, respectively, we can expect further strategic initiatives from these companies to build moats around their energy supplies.

Taking us back to the TLDR. While global consumption was 460 terawatt-hours (TWh) in 2022, the AI trend is projected to consume more than 1000 TWh in 2026, which is effectively equivalent to the Japan’s annual consumption of electricity. Are we ready to embrace this wave? The answer may vary widely depending on what stance you are taking.

We are not only witnessing a pivotal moment in technology, as described by OpenAI CEO Sam Altman during this AI Revolution, but we may also be at a crucial crossroads where environmental preservation and the future of humanity could become secondary concerns.

It’s a Wrap!

I must admit, my concerns about rising energy costs first arose when I discovered that my electricity bill for this summer increased by over 50% compared to last year 🤯🤯🤯 So yes, there is stream that I need to let out. 💨

While objectively speaking, the likely causes may be due to rising fossil fuel costs and the impending closure of the nuclear power plants, I can’t stop noticing the looming challenges on the horizon. With a population of only 24 million, Taiwan ranks #21 globally in terms of energy consumption per person, and is #10 in growth rate. The staggering increase combined with TSMC’s jaw-dropping performance this year, raises more questions than answers. Will AI eat into my electricity bill, or even into my living as a whole?

For those who know me, it’s clear that I am a passionate advocate of technology and AI. This passion directed me to create the newsletter you are reading now, Hungry Strategist, to capture my IO (input-output) as I closely observe these developments. But strides must be made with caution and discipline, especially when we are all now interconnected as one given globalization and the internet. AI should ultimately serve to enhance our lives, and nothing more. Period.

Food of Thought

Why so serious?