Zeke Hausfather

Check out my full analysis of expected impacts of the 2026-27 El Niño over at The Climate Brink: https://www.theclimatebrink.com/p/the-impacts-of-this-years-record

Additionally, I've added a fully interactive version of these maps to my Climate Dashboard, where you can see the up-to-date predictions from different models as well as explore impacts during past strong events El Niño events: https://dashboard.theclimatebrink.com/#impacts

Europe has some of the weakest teleconnections of any region, with few consistent signals during past El Niño events. Impacts there depend more on the development of the North Atlantic Oscillation, though models lean wet for the UK and central Europe this winter.

In Africa we expect to see sharp reductions of rainfall in the southern part of the continent, which threatens a maize crop on which tens of millions depend. Around the horn of Africa there is likely to be heavier rains and flooding during the short fall rainy season.

We expect drought in northern South America, over the Amazon (with corresponding high fire risk), and in Northeastern Brazil. There will be heavy downpours and flooding in Peru and Ecuador, and wetter conditions in Central Chile.

I'm not sure folks have realized just how crazy the second half of 2026 and 2027 will be for global temperatures – on the back of a record-smashing El Niño event. Here is my latest estimate of where both years will end up compared to global temperatures since 1850.

I looked at three different lines of evidence: the literature on El Niño teleconnections, what occurred during the past 8 strongest events since 1950, and the predictions of this year's event from 13 different dynamical models initialized in September.

In North America we expect to see wet winters from the Gulf Coast to Cuba. We also will likely have a wetter conditions in the Southwest, particularly in Southern California. Central America is likely to have drought conditions in countries south of Mexico.

The rise in Niño 3.4 region sea surface temperatures over the past few weeks has been nothing short of astonishing. We are now hotter than any days during the 1997/1998 super El Niño event, and we are still many months from the peak.

So while we think we are heading toward ~3C today, we could really end up at 4C if we role 6s on the proverbial climate dice. Its more useful to treat climate outcomes as a set of risks – the odds of the world warming by 2C, 3C, or 4C over time – rather than a deterministic outcome.

Many of the most reliable El Niño impacts are clustered around the tropical Pacific. Drought in Indonesia, the Philippines, the western Pacific islands have shown up in essentially every strong event since the 1950s, and this year’s models agree.

These arise from two factors: carbon cycle feedbacks – how the biosphere and ocean's ability to absorb carbon changes as we emit more and the world warms, and climate sensitivity – how the interplay of feedbacks like water vapor, clouds, snow and ice, etc. respond to warming.

We hear a lot about central outcomes of climate change. For example, under the new "medium" scenario (e.g. current policies continuing), the world warms around 2.9C by 2100. But these numbers mask some massive uncertainties in how the climate actually responds to our emissions.

And on the path we are on today those odds are deeply worrying. Its virtually certain we reach 2C by 2100. By 2150 we have a 77% chance of being above 3C, a ~30% chance of being above 4C, and a 6% of >5C warming: https://www.carbonbrief.org/explainer-the-cmip7-emissions-scenarios-and-how-they-explore-future-climate-change

2026 is on track to be the warmest (~35% chance) or second warmest (~63%) year on record, while 2027 is very likely (92% chance) to be the warmest on record on the back of a growing super El Niño event. For more, check out my new Carbon Brief State of the Climate Q2 update: https://www.carbonbrief.org/state-of-the-climate-rapidly-developing-el-nino-raises-chance-of-record-warm-2026

As always, you can treat real-time updates to this El Niño event (and the dynamical models of it) over at the Climate Dashboard: https://dashboard.theclimatebrink.com/#enso

Europe has been warming at a much faster rate than the world as a whole: roughly twice as fast as the global average, and 40% faster than the global land average.

Fires in France have now burned over 91,000 acres, by far the most of any year since the satellite record started in 2006. Current area burned is about 34x larger than this date in a typical (median) year:

Annualized, my agentic AI energy use is around 1.1 MWh, and is responsible for as much emissions as a clothes dryer or half a round trip flight from SF to NYC. Not enormous compared to my overall footprint, but far from trivial!

Check out the full analysis over at The Climate Brink: https://www.theclimatebrink.com/p/the-real-energy-use-of-agentic-ai

2027 is now expected to reach 1.7C in my model – exactly where Hansen predicted it would land (albeit in the GISTEMP dataset) back in December 2025:

The biggest lever we have is to decarbonize electricity – and push AI companies to invest more heavily in clean energy like solar+storage, advanced nuclear, enhanced geothermal, etc. Using energy isn't inherently bad, its all about the impacts of producing that energy.

I've previously done 2026 and 2027 global temperature estimates for the Earth's surface, but a lot of people also seem interested in what will happen in the lower troposphere (TLT). Historically TLT has had a much bigger response to El Niño and La Niña than we see on the surface:

This one was somewhat novel; I can't recall seeing a visualization designed like this before (but let me know if any of you have!).

2026 El Niño models continue their inexorable march upwards. CanSIPS (which used to be among the cooler ones!) released their August runs today, and now show a Nov/Dec peak of 3.6C (CanESM5) and 4.2C (GEM-NEMO). For reference, the prior record El Niño peaked at 2.8C.

Folks skeptical of human-driven warming like to point to a period of somewhat less rapid warming from 1900-1945. In a new analysis at TCB, I dig into the different factors driving it, and find that humans, natural factors, and variability all played an important role: https://www.theclimatebrink.com/p/what-drove-the-early-20th-century

The 2026/2027 El Niño event has already grown faster than any prior events (at least on an ONI basis). Here its observed and projected future evolution compared to the strongest prior El Niño events in recorded history.

The Niño 3.4 region of the tropical Pacific has been hovering above 2C for the past week. This is the threshold for a very strong ("super") El Niño event, and it was reached more than a month before the prior record set during 1997/1998:

I took a look at my own Claude Code use over the past 8 weeks to calculate agentic energy use. Using three different methods I find that the typical agentic prompt uses 180 watt-hours (45 to 360 Wh). This adds up to around 3 kWh per day, around the energy of running two fridges.

The claim that "pink elephants cannot fly" is not an evidence-based claim because studies have not ruled out the possibility that pink elephants can fly. 🙄

Currently 13 out of 14 dynamical models expect a record setting event based on the Niño 3.4 region sea surface temperature anomalies (ONI), with overall odds of a record sitting at 91% across all 667 model ensemble members.

Credit where credit is due: my projections have moved increasingly toward Hansen's over time, on the back of what appears increasingly likely to be the strongest El Nino on record later this fall. Year to date temperatures remain less exceptional (third warmest on record).

This warming has been fastest in the winter months – driven in part by greater absorption of sunlight with less winter snow cover – but has been rapid year-round:

If we instead look at the new relative oceanic niño index (RONI) that subtracts out warming across the tropical oceans, 11 of 14 models predict a record strong 2026/2027 El Niño event, with odds of a record sitting around 77%.

One potential silver lining: net-zero needs ~3x more clean electricity, and the main barriers are permitting, transmission, interconnection, not tech. If AI's capital and urgency fund firm clean power and grids, it could help clean the grid. Behind-the-meter gas, and it won't.

The bad news is that usage is more than outpacing increases in efficiency (Jervons Paradox in action!), and AI compute demand is rapidly rising with no end in sight. While we can try and avoid using agents for trivial things, behavior changes will only go so far.

So what can we do about it? The good news is that AI chips are getting massively more efficient, with the energy use per unit of intelligence falling in half every two years:

I'd still argue that this doesn't tell us all that much about bigger questions like climate sensitivity – its just not really possible use short surface temperature trends to constrain it much. But there are other lines of evidence that do point in that direction.

But in summer months, we've seen the hottest summer days warming nearly a degree faster than the monthly average temperatures in Western Europe:

If we look at RONI – which differences tropical sea surface temperatures from the ENSO3.4 region temperatures – we see an event reaching 3.3C in December, and are about 0.2C warmer than the June forecasts.

This is a pretty generous approach to attributing temperature to solar forcing, as it uses an instantaneous equilibrium response of 0.5 to 1.2C for each w/m^2 increase in forcing (equivalent to a climate sensitivity of 2.0 to 4.7C). Using a proper climate model gives a somewhat smaller response: