Why It’s Still So Hard to Count Heat Wave Deaths

More than 5,700 excess deaths were recorded in France during this summer’s record-breaking heat wave, the country’s health agency announced today. That makes the event — which ran, by the official reckoning, from June 17 to July 2 — the country’s deadliest heat wave in more than 20 years.

That’s in line with other estimates we’ve heard. EuroMOMO, a network of European public health agencies that track excess mortality, found that the continent saw more than 10,000 excess deaths during the same period. Roughly 90% of those victims were older than 65, it said. (France’s cohort seems similar: Adults older than 75 made up about two-thirds of the victims, the government said.)

These numbers are staggering — and much larger than some astute Heatmap readers might anticipate. If you read my colleague Jeva Lange’s piece on why it’s so hard to estimate heat deaths last week, she cited a much smaller estimate: Roughly 2,700 died in France during the most recent heat wave. That tally came from Christopher Callahan, an Indiana University scientist who studies climate change’s economic and social costs.

Why is there such a gap between the figures? I emailed Callahan to find out. He shared a few thoughts. First, he uses a different (and theoretically more rigorous) method than the French government: “Our approach uses a statistical relationship between temperature and mortality to explicitly quantify how many additional deaths are associated with a given day’s temperature,” he wrote. “France’s report of excess deaths is just based on how many more people died in late June compared to previous Junes - but we don’t know if those people died because of the heat or some other factor.” (Carbon Brief recently published a Q&A on these varying approaches.)

That might mean his estimate is right, in which case France has misidentified roughly nearly 3,000 deaths. But it could also mean his model, which is trained on data from 2004 to 2019, is “missing something,” he said, like a post-Covid change to public health risk. Last year, Callahan and his colleagues used a similar model to estimate deaths from France’s worst-ever heatwave, a 2003 episode that overwhelmed morgues and killed about 16,000 people. Even 23 years ago, global warming helped make that disaster larger than it needed to be: Some 6,000 of those deaths were due to climate change, their paper found.

Either estimate of the 2026 heat wave, of course, is shattering. As Jeva wrote, even the lower figure would mean the 2026 heat wave killed as many people as died in three years of French homicides. But the divergence in estimates tells us something else too: Even as climate change breaks records and alters our world, we’re never going to quite agree on where it ends and normal randomness begins.

Google blows out expectations

The AI data center boom does not seem close to ending. Google’s parent company, Alphabet, announced its second quarter results this evening, and it beat Wall Street’s expectations, nearly quadrupling its profit on a year-over-year basis. Among the drivers: Its cloud business grew 82% compared to the same quarter last year. (As I’ve written, that rapid growth is helping to turn Alphabet and other hyperscalers into light industrial firms.)

The company’s AI bets seem to be paying off so far — so Google is now planning on spending even more on data centers, energy infrastructure and AI development this year than it once anticipated. It raised its estimates of 2026 capital expenditure to $195 billion to $205 billion, which is above earlier projections and twice as much as it spent in the same category last year. 2027 could be even bigger, it signaled. The company’s shares fell slightly on the news in after-hours trading, but from an energy and climate wonk perspective, the message is clear: For now, the AI demand surge transforming the power sector — and the real economy — continues to chug along.