Earlier this week, International Energy Agency (IEA) released a special report titled, Energy and AI, addressing the recent rapid growth the Artificial Intelligence (AI) sector from an energy use lens.
The recent mass adoption of Generative AI has put immense pressure on the electricity grid due to the increasing energy consumption by data centers training models and providing services to hundreds of millions of users worldwide. In absolute terms, currently, data centers consume roughly 1.5% of electricity globally and it is one of the fastest growing electricity demands, expected to double by 2030. Moreover, the challenges often arise when one considers where this electricity demand comes from (concentrated geographical pockets near cities) and when it is expected to be fulfilled (often, the mismatch between renewable generation and data center consumption leads to green washing via clean energy certificates).
In fact, the spatio-temporal pressure exerted on the electricity grid by this rapidly-increasing demand is often under-estimated and unfortunately, misrepresented in popular media by interested parties.
The narrative weaved is not only that current (and future) AI sector’s energy consumption is dwarfed by other sectors, but also that AI’s widespread, cross-sectoral adoption will empower experts to come up with solutions that will solve climate change altogether in the future, thus rendering such concerns moot.
However, the fact is that besides the power-hungry data centers and their demand for cooling, the global supply chain of chip manufacturing and assembly is also a highly energy-intensive, precise manufacturing process. This is crucial because often chip manufacturing occurs in Asian tech hubs such as China, Taiwan, Japan, and South Korea, where fossil-fuel based electricity generation, e.g., coal-fired power plants still dominate the energy mix and cater to these growing industrial demand.
Now, I am not someone who is skeptical of the massive (mostly positive) impact AI technologies will have on improving ordinary people’s lives and, therefore, I am not against growth in this sector. I believe in the transformative capabilities of this technology in enhancing human productivity and efficiency, being an early and curious adopter myself.
But, I am deeply troubled by this narrative.
Climate change is an immediate and existential threat, with sharply escalating cost of inaction over time. Dealing with this challenge demands intense focus, bold decisions, and steadfast commitment from leaders and decision-makers. Each year that goes where actions fall short of what’s required, we are stacking up the future societal cost of climate change on humanity, both in terms of lives lost and livelihoods impacted.
Therefore, the fact that such a narrative around AI diverts crucial resources away from climate tech, i.e., innovations helping fight climate change, and towards AI-focused innovations is extremely concerning to me.
Recent articles, such as this and this, lament the slowdown in funding for climate tech sector while investor funding for AI soars. In fact, 2024 witnessed a 40% year-on-year decline in equity financing for climate tech after witnessing several years of growth, while AI (primarily, generative AI) exploded by 80% year-on-year.
This trend can only be expected to continue in the near-term, as the world gears up for a century-defining global AI arms race. At the time of writing this article, unfortunately, the world order is being rejigged, uncertainty remains high for financial markets, and a global turmoil seemingly looms around the corner.
I acknowledge that rationality demands decision-makers to focus on building adaptation and resilience against the short-term geopolitical threats. And that is understandable, but it must not replace the focus on actions to mitigate the threat that climate change poses.
Moreover, the cost of this shifting focus from humanity-saving climate tech towards Studio Ghibli style rendering of our photos, however joy-giving, appears nothing short of epic stupidity beyond comprehension to me.
Besides money, it diverts some of the smartest human beings on the planet towards AI, away from working to find solutions to the most pressing challenge of our lifetime, i.e., climate change. And I must admit, if one is talented, it is tempting to ride this new wave of AI technology and bask in the hype that it enjoys.
However, the world needs all the resources it can muster right now to focus on developing solutions that urgently upscale our efforts to avoid catastrophic and irreversible climate change — to safeguard our survival.
And the time is running out!
Year after year, the wording in United Nations Environment Programme’s annually-published report focusing on Emissions Gap (a metric tracking the discrepancy between what nations have “pledged to do” vs. “what needs to be done”) keeps getting grimmer.
Keep in mind that these are still pledges made and not actions completed!
Their latest report shows that on a global level we are regressing in our efforts to reduce carbon emissions. And that the time lost in geopolitical and nationalistic power games has severe implications for the carbon budget available for the remainder of this century to prevent catastrophic climate change.
In fact, as evident from Figure 1, the 1.5 C warming cap (over pre-industrial levels) that governments signed up for in the Paris Agreement in 2015 is no longer feasible. Current pledges imply that the temperature would still rise between 2.6 and 2.8 C by 2100, which spells horrifying outcomes for millions of people and for a number of biological ecosystems.
“The world has currently warmed by about 1.3 C already”, Gloria Dickie reports sombrely in Reuters.
Unfortunately and surprisingly (even for climate experts), 2024 was the first year where the Global Mean Surface Temperature (GMST) crossed the “1.5 C above the pre-industrial level” threshold. The verdict is unclear whether this was just an anomaly or the beginning of a trend (since 2023 also reported higher than normal).
A recent scientific article (available Open Access) uses satellite data from NASA and climate reanalysis data from ECMWF to conclude that these observations can be largely explained by the reduction in low-level clouds across the planet, which help cooling the planet by increasing the reflectivity of the atmosphere to solar radiation (cloud albedo). That the reduction in low-level clouds is linked to the increasing GMST suggests that this anomaly may, in fact, be the beginning of a trend (possibly an addition to the list of so-called global climatic tipping points) and not just an anomaly is quite worrisome.
Not only this, recent academic works, such as this and this, has unfortunately, been pointing to another tipping point being reached sooner than estimated. This refers to the slowing down (and possibly eventual collapse) of the Atlantic Meridional Overturning Circulation (AMOC) current which is one of the crucial elements of the circulation system of our planet. A collapse of the AMOC current brings devastating news for Europe and has severe climatic ramifications on the entire world.
While climate scientists debate the “when” of this (partial or total) collapse, there is broad consensus that it is not a question of “if”, unless large-scale and urgent actions on climate change mitigation are taken right now!
And it deeply worries me to write that, one cannot reasonably rule out this collapse from happening within our lifetime, around the 2050s, if we continue pumping CO2 into our atmosphere business-as-usual. In fact, considering the severe impact the collapse of AMOC could have in the Nordic region (see Figure 2, below), climate scientists from around the world wrote an Open Letter to the Nordic Council of Ministers last October, highlighting how our previous estimates of the severity of this slowdown might be biased and incorrect and that the risk is likely higher than originally thought.
So where does this lead us?
The IEA report on Energy and AI that I referred to earlier, concludes that the AI and energy sectors are more intertwined than ever, that they are in a nexus that is expected to bear fruit for the world. While the authors express their concerns about the uncertainties that lie ahead, specifically around what energy generation sources fuel this growth in AI demand for electricity, their overall outlook remains neutral to positive. This stems from the advances in energy efficiency of AI models, such as DeepSeek, as well as the voluntary interest in green Power Purchase Agreements (PPAs) and nuclear energy shown by tech giants, such as Apple and Google.
I wish I could fully share their optimism. If history teaches us anything about such a cocktail of fast-paced progress coupled with intense industry lobbying efforts to secure profits, it is to not be complacent about the long-term impacts they bring.
I can’t help but recognise parallels to the oil and gas industry from half a century ago, when the research and discourse on global warming was deliberately and meticulously suppressed and the world was misled by oil and gas giants towards climate change denial, while they kept raking in profits.
We, as a society, must not fall for a similar trap again, because frankly, we cannot afford to. Our survival depends on it.
The climate impacts of AI must, therefore, neither be underestimated nor be overshadowed by the narrative of “AI is for societal good”, especially in the most crucial decade for humanity from a climate change mitigation perspective.
I believe that proactive policy and regulation needs to step up to make AI technology not only compatible with but also answerable for addressing the global challenge of climate change.
Policy-makers around the world must force the large corporate beneficiaries of the AI boom to internalise the negative externality of their undertakings on the global climate into their business models. And that it must be paid for, in full, by them. Not via green washing or purchasing carbon credits or renewable certificates, but by enforcing requirements such as hourly-matching of data center consumption with renewable production, mandated funding of clean energy project development and/or innovation, etc.
And we are seeing some encouraging progress, but so far it has been voluntary and concentrated in a few geographies. As progress in AI becomes global and countries around the world build their own AI capabilities aggressively, it is extremely important for regulation to step up. Our AI-hooked advanced society in the 21st century must be built on principles of sustainability and climate change mitigation.
Thankfully, this detailed 300-page report by IEA brings more attention to this important and timely issue, covering it in depth and providing trends and insights to policy makers and relevant subject matter experts.
I hope that the ensuing dialogue is solution-focused and not one built around shirking and deferring responsibility.
It’s time to start turning the current lopsided relationship, where AI benefits more from the energy sector towards a symbiotic cooperation between the two sectors towards solving the greatest threat humanity has ever faced — Climate Change.
Thank you for reading! I would love to hear your thoughts and insights on this topic.
<hr><p>Climate Change and AI-hooked Society: A Compatibility Check was originally published in Clean Energy for Billions on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>