A competition nobody asked for: heat vs. guns?

Welcome to climate change. Let's hope AI is good for something.

Disclaimer: So instead of talking only about the AI era in this column, let’s combine both: AI and the climate change era today, because I can’t take the current heatwave in Europe without talking about it.

It’s World Cup time, and instead of debating whether Brazil will finally make a comeback (if you didn’t know, yes, I’m Brazilian), let’s dive into a much more unfortunate competition. What claims more lives: heat in Europe or guns in the US?

I don't pay much attention to graphs that go viral online, because it never comes with source. This time, though, I'm willing to give my two cents because the analysis comes from Hannah Ritchie, one of the best data storytellers around, using official and verifiable sources (please, don't trust every chart that shows up on your feed).

What makes Hannah’s work particularly valuable is that she doesn’t stop at a single headline number. She compares different definitions of what counts as a heat-related death and also presents the results on both an absolute and a per capita basis.

You can read Hannah’s full analysis here, but here’s the bottom line:

In absolute numbers, more people die from heat in Europe than from guns in the US. Europe records an average of about 60,500 heat-related deaths per year (based on 2022–2024), while the US recorded 44,400 total gun deaths in 2024.

But Europe as a whole has more people than the US, so you would also expect more deaths overall. That’s why Hannah also shows the numbers adjusted for population size. Looking at deaths per 100,000 people is simply a fairer comparison because it puts both on the same scale, and here the story changes a bit.

When we adjust for population size, gun deaths in the US slightly exceed heat-related deaths in Europe. The US records 13.2 gun deaths per 100,000 people, compared with 11.2 heat-related deaths per 100,000 people in Europe.

Now for my two cents:

I’m not entirely convinced that grouping all gun deaths (homicides and suicides) is the right comparison for this question. From my perspective, they represent fundamentally different phenomena.

Heat-related deaths and gun homicides both involve people losing their lives as victims of external circumstances. They reflect failures in protection, infrastructure, public health, or public safety. Gun suicides, while equally tragic, belong to a different conversation with different causes, risk factors, and policy responses. Combining them into a single bar risks blurring those distinctions.

So, if we compare apples with apples (heat deaths in Europe versus gun homicides in the US), the conclusion becomes much stronger. Europe experiences roughly 60,500 heat-related deaths each year, while the US records around 15,400 gun homicides annually¹.

Whether you look at absolute numbers or per capita rates, heat claims considerably more lives than gun homicides.

That established, I refuse to accept this death toll from heat. It seems to me that providing infrastructure for the new times we live in (where climate change isn’t really contestable anymore, right?) should be a priority.

The more upsetting thing is that this isn’t just about life and death. There’s me, my coworkers, and everybody in between who simply live quite poorly during these heat waves. While most politicians and companies won’t pay attention to our well-being, I hope I can get through to them with something that often moves the narrative: the economic cost of heat.

Suffering from heat in numbers

Let’s start with our brain.

Above 16.5°C outside, every additional degree reduces your cognitive performance by about 0.13%² . Western Europe is sitting at 39–41°C this week. Doing the math that’s roughly 23 degrees above the threshold, which adds up to around a 3% drop in cognitive output than on a mild spring day.

For students, the numbers are even sharper. A degree Fahrenheit hotter school year reduces that year’s learning by 1% BUT the effect disappears almost entirely with air conditioning³. A follow-up study covering 58 countries and over 144 million young people confirmed that the same relationship between heat and reduced learning holds across very different climates, income levels, and education systems⁴. In other words, this is not only a European problem.

Now zoom out to the economy.

In Switzerland (one of Europe’s cooler, wealthier countries), heat already costs CHF 665 million per year in lost labour productivity (when it gets too hot, people work more slowly, need more breaks, make more mistakes, or simply cannot function—like me writing this article today to cool off). If we keep going in the worst direction on emissions, that figure could triple by the end of the century⁵. Switzerland isn’t even the worst-hit country—it is actually much closer to the best-case scenario.

Across Europe as a whole, heatwave damages ran at 0.3–0.5% of GDP in recent hot years, already exceeding 1% in southern regions like Croatia, Cyprus and Greece⁶. Looking further ahead, worst-case projections put European losses at €563 billion per year, or 1.15% of GDP, by the 2080s, with Southern Europe bearing the brunt at 3–5% of GDP⁷

Zoom out further to the global picture, and the ILO projects that by 2030 the world will lose the equivalent of 80 million full-time jobs worth of working hours every year (US$2.4 trillion in economic output) simply because it is too hot to work safely⁸. Already in 2024, heat exposure caused 640 billion lost work hours globally, 98% above the 1990s average, equivalent to roughly US$1.09 trillion in lost income⁹.

Here is where I want to pause and connect two research agendas that, as far as I can tell, nobody has formally put in the same room yet. This week I wrote about job displacement because of AI (go read it here if you missed it), and now I am checking the numbers on job loss because of climate change. Sincerely, as workers, it feels like we are doomed either way. Unless, of course, we use AI to actually solve climate change and pull off a plot twist on this whole story (is this even possible?).

Can AI actually help? An honest balance sheet

Before we talk about AI, let’s address the elephant in the room: air conditioning, therefore, energy.

The standard argument against expanding AC in Europe is environmental. More cooling means more electricity, more electricity means more emissions, and we end up making the heat problem worse while trying to survive it. It is a legitimate concern but it rests on an assumption that is becoming less true every year: that electricity is dirty. In 2024, renewables accounted for nearly 30% of global electricity generation, and in several European countries the share is already above 50% in many hours of the day¹⁰.

So the relationship between cooling and carbon is not fixed. It depends on what powers the grid and whether the grid is designed for the climate reality we already live in, not the one from fifty years ago.

Research from ETH Zürich¹¹ shows that whether climate change helps or hurts your electricity grid depends almost entirely on the choices made when building it — how much solar versus wind, and how much of your heating and cooling runs on electricity. So there isn’t one answer to whether higher electricity demand on hotter days helps or hurts the electricity grid. The answer depends largely on how we choose to design it, because the same increase in electricity demand can produce completely different outcomes depending on the system it hits, which means that the most important variable in this story is not the heat itself but the design of the electricity system.

If it’s a design problem with too many possible options, maybe AI can help.

Let’s start with what AI can already do, then look at what it costs, and finally be honest about whether the maths actually adds up.

What can AI do for our energy problem?

The IEA’s 2025 Energy and AI¹² report finds that AI applied to sensors on transmission lines could unlock up to 175 GW of additional capacity in existing infrastructure (without building a single new cable). To put that in perspective, 175 GW is roughly the entire electricity consumption of Germany and France combined on a hot summer day. This is not new power; it is power that already exists but gets wasted because the grid cannot manage it smartly enough. AI-based fault detection can also cut the duration of power outages by 30–50%, meaning when the grid does fail during a heatwave, the lights come back on faster. During a heatwave, when cooling demand spikes simultaneously across an entire region, both of those things matter directly to whether your AC keeps running or not.

China offers the most concrete city-scale examples of what this looks like in practice. Shanghai launched a citywide system using AI to coordinate capacity from 47 operators at once treating them as a single flexible resource. During a trial in August 2024, the system absorbed a sudden demand spike by shedding 162.7 megawatts of load in real time (roughly enough electricity to power 130,000 European homes for an hour). In Shenzhen, similar AI tools were used over 150 times between 2023 and 2025 to reduce grid stress at critical moments¹³.

What AI costs

Now the other side (because there is always another side).

The same AI systems that forecast heatwaves and optimise grids need enormous amounts of energy and water to train and run. Global data centre electricity demand is projected to roughly double by 2030, from 485 TWh in 2025 to around 950 TWh, growing at about 15% per year, four times faster than all other electricity demand combined¹⁴. That is the equivalent of adding the entire electricity consumption of Japan to global demand in five years. Today, about 30% of that electricity comes from coal and only 27% from renewables. The renewable share is growing but not fast enough to offset the growth.

We cannot forget the water consumption (here, there is so much noise on the internet, so please always check the source). A study published in Nature Sustainability estimated that AI server deployment in the US alone could generate a water footprint of 731 to 1,125 million cubic metres per year by 2030 (roughly the annual household water use of 14 to 21 million Europeans)¹⁵.

Can we net them out?

The IEA tried. Their estimate is that well-documented AI applications could save 13 exajoules of energy globally by 2035 (roughly 3% of all final energy consumption worldwide). Data centres will consume around 3.4 exajoules per year by 2030. On paper, the savings dwarf the cost by almost four to one.

Buttttttttt those savings require broad adoption of AI efficiency tools across industry and buildings and that is not happening automatically or fast enough. Meanwhile, data centres are already consuming more energy every year, most of it still from fossil fuels.

So can AI save us from the heat?

There is a philosophical trap that haunts every technology we have ever built. David Collingridge named it in 1980: when a technology is young enough to change, we don’t yet know what needs changing. By the time we understand the consequences, the technology is so embedded in everything (the economy, the infrastructure, our daily lives) that changing it is slow, expensive, and politically painful.

AI and energy are a perfect example. We can already see both sides of the ledger clearly enough to know we should be paying attention. The savings are real, but they are conditional. The costs are real too, and they are already happening without the transparency we need to keep them in check. I believe the window to get the design right is still open, but it won’t stay open forever.

I started this article wanting to tell you that AI can help with heat. The honest answer is: it can, it is already doing so in some places, and it is also part of the problem in ways we are only beginning to measure. If you’re waiting for a definitive answer, I’m afraid the data just isn’t there yet.

I promise to keep an eye on it and keep you posted, so follow along.

1. Pew Research Center, What the data says about gun deaths in the US, April 2026. https://www.pewresearch.org/short-reads/2026/04/28/what-the-data-says-about-gun-deaths-in-the-us/

2. Krebs, B. (2024). Temperature and cognitive performance: Evidence from mental arithmetic training. Environmental and Resource Economics, 87, 2035–2065. https://doi.org/10.1007/s10640-024-00881-y. The 0.13% figure reflects the effect of a 1°C increase in daily average outside temperature on performance in a single cognitive session — not a cumulative hourly effect. The study uses data from the Lumosity brain-training platform, matching each game session to local weather data on the day it was played. The threshold of 16.5°C is the temperature above which the negative effect begins; below it, temperature has no measurable impact on performance.

3. Park, R. J., Goodman, J., Hurwitz, M., & Smith, J. (2020). Heat and learning. American Economic Journal: Economic Policy, 12(2), 306–339. https://doi.org/10.1257/pol.20180612. The study uses fixed-effects models on approximately 10 million students who retook the PSAT, matching their scores to school-year temperatures. The air conditioning finding comes from the first national school-level dataset on AC penetration across the US, schools with cooling infrastructure showed no meaningful heat effect on learning.

4. Park, R. J., Behrer, A. P., & Goodman, J. (2021). Learning is inhibited by heat exposure, both internationally and within the United States. Nature Human Behaviour, 5, 19–27. https://doi.org/10.1038/s41562-020-00959-9.

5. Stalhandske, Z., Nesa, V., Zumwald, M., Ragettli, M. S., Galimshina, A., Holthausen, N., Röösli, M., & Bresch, D. N. (2022). Projected impact of heat on mortality and labour productivity under climate change in Switzerland. Natural Hazards and Earth System Sciences, 22, 2531–2541. https://doi.org/10.5194/nhess-22-2531-2022

6. García-León, D., Casanueva, A., Standardi, G., Burgstall, A., Flouris, A. D., & Nybo, L. (2021). Current and projected regional economic impacts of heatwaves in Europe. Nature Communications, 12, 5807. https://doi.org/10.1038/s41467-021-26050-z

7. Szewczyk, W., Mongelli, I., & Ciscar, J. C. (2021). Heat stress, labour productivity and adaptation in Europe — a regional and occupational analysis. Environmental Research Letters, 16, 105002. https://doi.org/10.1088/1748-9326/ac24cf

8. International Labour Organization. (2019). Working on a warmer planet: The impact of heat stress on labour productivity and decent work. https://www.ilo.org/publications/major-publications/working-warmer-planet-effect-heat-stress-productivity-and-decent-work

9. Lancet Countdown on Health and Climate Change. (2025). Explore our data — labour. https://lancetcountdown.org/explore-our-data/

10. International Energy Agency. (2025). Renewables 2025. IEA, Paris. https://www.iea.org/reports/renewables-2025

11. Bloin-Wibe, L., Fischer, E., Göke, L., Knutti, R., De Marco, F., & Wohland, J. (2026). Climate change impacts on net load under technological uncertainty in European power systems. Environmental Research Letters, 21, 084004. https://doi.org/10.1088/1748-9326/ae5727.

12. International Energy Agency. (2025). Energy and AI — executive summary. IEA, Paris. https://www.iea.org/reports/energy-and-ai/executive-summary.

13. Walton, R. (2025, August 15). Virtual power reaches global: China claims new aggregation record. Microgrid Knowledge. https://www.microgridknowledge.com/distributed-energy/virtual-power-plant/news/55310131/virtual-power-reaches-global-china-claims-new-aggregation-record

AI News. (2025, December 23). Inside China’s push to apply AI across its energy system. Artificial Intelligence News. https://www.artificialintelligence-news.com/news/inside-chinas-push-to-apply-ai-across-its-energy-system

14. International Energy Agency. (2026). Key questions on energy and AI — executive summary. IEA, Paris. https://www.iea.org/reports/key-questions-on-energy-and-ai/executive-summary.

15. Zhang, C., & Rao, P. (2025). Environmental impact and net-zero pathways for sustainable artificial intelligence servers in the USA. Nature Sustainability. https://doi.org/10.1038/s41893-025-01681-y. The European household water comparison is calculated using Eurostat’s figure of approximately 53 m³ per person per year for household water use from public supply. Source: https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Water_statistics