| Report: How AI will shape the future of energy |
 Source: Unsplash |
| The International Energy Agency (IEA) today published an extensive (300-page!) report on the complex relationship between artificial intelligence and energy. |
| The work aims to inject data-based projections into a landscape that has been papered for years, now, in reports concerning the rapid electricity demand incited by the rise of artificial intelligence, and the subsequent surge in carbon emissions, water consumption and grid unreliability — against a backdrop of energy desperation — it is causing. |
| The major findings: The focal point of the report revolves around the IEA’s “base case” for AI advancement and AI adoption; this base case holds that the global electricity demand of data centers is set to more than double by 2030 to 925 terawatt-hours (TWh), a number that is slightly more than the entire electricity consumption of Japan today. |
| In 2024, data centers consumed 415 TWh, equating to roughly 1.5% of the world’s electricity consumption. |
| AI, according to the report, “will be the most significant driver of this increase” — electricity demand from data centers specifically optimized for AI is projected to “more than quadruple” by the end of the decade. And in the U.S., the power consumption of data centers is set to make up nearly half of the country’s growth in electricity demand between now and 2030. |
| The electricity demand associated with data centers began surging in 2017, a spike that was driven by a number of non-AI (depending on how you define it) factors, including the growth of the cloud, the expansion of social media and streaming and the rise, of course, of both traditional and modern forms of AI. |
| Still, the relationship between energy and AI is complicated; there is a clear cost, in both dollars and emissions, of rising AI demand, according to the IEA, but there remain significant uncertainties regarding how capable the technology will become, how quickly it will advance and how quickly it will get adopted on a mass scale. |
| Keeping these differing scenarios in mind, the range of the IEA’s headwinds to tailwinds cases see data center electricity demand somewhere between 700 and 1,700 TWh by 2035. |
| AI and emissions: While the report found that data centers are “among the largest sources of growth in emissions,” the peak and decline of those emissions after 2030 will remain at around 1% of the total emissions coming from the energy sector. |
| Global carbon emissions from fuel combustion are expected to reach 35 billion tons in 2024, according to the report; in 2024, data centers accounted for around 180 million tons of carbon emissions. |
| Then, there are the “rebound effects” associated with the proliferation of AI. For instance, “cheaper oil and gas could directly induce greater demand and, therefore, higher emissions.”Likewise, a surge in autonomous vehicles could drive people away from public transport, increasing vehicular emissions; more efficient generative AI models could “lead to significantly higher use in daily life”; and an expansion in robotics could, likewise, boost energy demand. |
| In many ways, the IEA said, AI and automation might be able to help; instantaneous statistical analysis and insights could help energy companies waste less, while aiding grid operators in the challenging task of grid balancing. The technology could also drive innovation in materials generation — as we’ve discussed — which could lead to breakthroughs in battery technology and renewable energy technology, reducing AI’s net emissions. |
| These efficiencies could power emission reductions that are larger than the initial emissions themselves. However, the report added that there is “currently no existing momentum of Al adoption that would unlock these emissions reductions to this degree.” |
| “AI is a tool, potentially an incredibly powerful one, but it is up to us — our societies, governments and companies — how we use it,” IEA Executive Director Fatih Birol said in a statement. |
| The landscape: The report shortly follows new executive orders from President Donald Trump designed to boost the coal industry in order to meet mounting AI energy demands. This push comes as the cost of coal remains far higher than renewables and natural gas. |
| And on the individual company scale, the electricity demand, water consumption and carbon emissions associated with driving AI have been on a constant increase among the hyperscalers, an increase that has pushed these companies into the arms of nuclear energy. |
 |
| The poin that I feel often gets missed here is that, as much as it is a good thing for renewables to power AI data centers, their allocation to this additional strain on our planet and our energy grids means that renewables are not being applied — at least, as much as they could be — to reduce our base energy demand, and our base carbon emissions. |
| As a share of the whole, AI emissions and demand might not be enormous, but what it is doing is leading to increases in demand that we don’t have enough green capacity to respond to, at a time when we really need clean energy to be the bulk of what we deal with. |
| Unfortunately, governments are incentivizing companies to do this quickly. |
| Not cleanly. |
|
canisgallicus 