Sponsored Content

AI May Accelerate the Great Energy Transition – But Not the Way You Think

Sponsored by 
Native Article Lead Image Update 070224_inst_articleimage_2000x1126_ai2.jpg

Artificial intelligence (AI) is on everyone’s minds these days, and for good reason – it has the potential to dramatically change society in ways we don’t yet fully understand. That includes how we approach the great energy transition and the US$4.5 trillion in investments needed annually by 2030[1] to get the world to net zero. Like any new technology, there are certainly reasons to get excited about the impact AI may have on our lives and the energy transition, but there are some significant concerns to consider as well.

In an ideal but highly unlikely scenario, AI will allow us to conjure up some new source of energy we haven’t yet imagined. What’s more plausible is that it will expose the inefficiencies in our energy grid, forcing utilities, energy consumers, financial companies, governments and others to invest more heavily in electrification.

Indeed, the diminished state of the global electricity grid is already coming into focus. Bankers are starting to extend their AI excitement from software and chip manufacturing to the inadequacies of our existing power infrastructure, which is something renewable developers, electric vehicle makers and heat pump proponents have so far failed to do themselves. If AI refocuses industry stakeholders and investors on rebuilding the global electricity grid – that will be its greatest contribution to our future.

But improvements in the grid need to happen quickly. Over the past decade, global electricity demand has been growing at just over 2% per year. So far, generation capacity and the global grid system have mostly been able to meet this growing demand. The next decade will test this resiliency. Investments in renewables have been encouraging, but the investments in transmission infrastructure haven’t kept pace. Almost 1 terawatt of renewables projects – about 7 times Canada’s total generating capacity – cannot even secure grid connections[2].

Power demand from AI will compound this conundrum. The International Energy Agency (IEA) estimates that data centres alone consumed around 460 terawatt hours of electricity in 2022, about 1% of global demand. Generative AI searches, however, can be 10 to 35 times more energy intensive than traditional searches. As AI becomes more widely used, it’s estimated the technology could eventually account for between 2.5% to 6% of total global generation.

Energy consumption from AI will have even greater regional implications. Bloomberg New Energy Finance suggests the Chicago area might see a 900% jump in data centre power demand, requiring the electricity output of four new nuclear reactors. In the same study, power utility Southern Co. predicted AI would increase its electricity sales by 6% per annum with 80% of that for data centres.

In addition, market research group Dell’Oro estimates annual global data centre CAPEX to nearly double to US$400 billion from US$240 billion between 2022 and 2027. And these estimates could be low. Amazon, for example, is projecting US$150 billion in data centre investments over the next 15 years, while Microsoft, in combination with OpenAI, is planning a single next-generation super data centre project with a price tag that could be more than US$100 billion[3].

All this demand will not only add more pressure to the grid, but could also make electricity more expensive. To satisfy the energy requirements needed to run these data centres, Microsoft, Amazon and many others – all of which are flush with cash – will do whatever it takes to connect to the grid and purchase electricity. That’s a positive in that all that spending will allow utilities to make significant infrastructure investments, but given that transformer stations are currently on five-year backorders, it’s going to take years before supply keeps up with demand. Not only will rising electricity prices pinch consumers, they will also price out critical developments in green hydrogen, ammonia and cement, processes that require electricity, and that will slow down the energy transition.

Other Impacts From AI

AI’s insatiable need for power will impact other areas, too. If renewable energy can’t support the increase in demand, utilities will likely need to turn back to traditional fuels – the Washington Post recently reported [4]that utility executives in the U.S. are lobbying the government to delay closing fossil fuel plants.

Companies also need vast water resources to help cool the many computers used in their data centres. Microsoft, in its 2022 environmental report[5], revealed that its global water consumption jumped by 34% in 2022 over a year earlier, while Google reported a 20% increase over the same time period. Shaolei Ren, a researcher at the University of California, estimates[6] that ChatGPT uses 500 millilitres of water for every five to 50 prompts or questions it’s asked.

At the same time, there’s a growing concern that all the hype surrounding AI will distract capital from the very infrastructure that enables it to function. The IEA estimates that if we’re going to reach international climate goals, 80 million kilometres of grid wires must be added and refurbished by 2040. Annual investment globally in grid-related upgrades will need to reach US$600 billion by 2030 from US$300 billion today and then increase to US$800 billion per year between 2040 and 2050. Imagine what might happen if all that money went into AI itself – a technology that doesn’t yet have a solid business case.

That may already be happening. Last month, The Economist claimed the market cap of the top 100 listed AI companies cumulatively added US$8 trillion in market cap over the past year. It’s an astonishing figure, greater than the market cap of the global airline, pharmaceutical and automobile businesses combined. For a technology that is mostly misunderstood and whose business model is unclear, this could be a historic capital destroyer. It’s essential that AI attracts capital to, and not away from, rebuilding our global electricity grid.

One can only hope that all this demand results in the creation of a larger, more resilient grid. If AI doesn’t just continually gobble up all the electricity needed to support the transition, then we could see utilities overinvesting, similar to how network companies did in the late 1990s and early 2000s. And while all that exuberance didn’t work out as well for the early networking companies – see Nortel – it was a good thing for the digital economy and the companies that tapped into an already-built infrastructure.

However, we need to ensure the world makes it to a place where we benefit from those long-term upgrades. If investments pour into AI instead of into transition-related innovations, if we damage our water supply or move backward with fossil fuels, if electricity becomes too expensive for consumers and others, if we fail to invest in the grid in the ways we should, will the AI revolution really be worth it?


Issued by Mackenzie Financial Corporation. For institutional investor use only. This material is for marketing and informational purposes only and does not constitute an offer of investment products or services (or an invitation to make such offer).

This material, including any references to specific issuers, is not intended to constitute investment advice or any form of recommendation.

[1] International Energy Agency, net zero roadmap, 2023

[2] CER – Canada’s Energy Future 2021 Fact Sheet : Electricity (cer-rec.gc.ca)

[3] Microsoft, OpenAI plan $100 billion data-center project, media report says | Reuters

[4] The Washington Post, 2024

[5] Microsoft, Sustainability Report, 2022

[6] Making AI Less “Thirsty”: Uncovering and Addressing the Secret Water Footprint of AI Models, October 2023