London sits at the same latitude as Calgary. In January, no Londoner needs ski gear to go to the office, while in Calgary cars carry an external block heater plug because otherwise the oil freezes. Edinburgh is farther north than Moscow. Naples shares a parallel with Chicago. And yet palm trees grow on the coast of Galicia while the sea freezes over in Labrador. The difference has a name: the Atlantic Meridional Overturning Circulation, AMOC, a conveyor belt of warm water carrying roughly 1.3 petawatts of thermal energy toward northern Europe [1]. To put that in perspective: more energy than all of humanity consumes as electricity. And it has been losing strength for decades.
What this machinery is
Thermohaline circulation —”thermo” for temperature, “haline” for salinity— is the global system of currents driven by density differences. In the Atlantic, warm, salty surface water travels from the tropics northward. When it reaches the waters near Greenland, Iceland, and the Norwegian Sea, it releases heat into the atmosphere. It cools, and when the surface water freezes and forms sea ice, the ice expels salt into the surrounding liquid. The water left behind becomes cold and extraordinarily saline, gains density, and sinks to 2,000-4,000 meters depth in a process called deep convection. That sinking is the pump that pulls tropical water northward, completing the loop. Between 16 and 20 sverdrups —millions of cubic meters per second— more than a hundred times the combined flow of every river on the planet [2].
A fact that never loses its power to surprise no matter how often it’s repeated: the water sinking today off Greenland will take about 1,000 years to resurface in the North Pacific. What we’re altering now carries millennial inertia.
But it’s worth keeping in mind that Richard Seager and his team showed in 2002, using atmospheric simulations, that even if the ocean’s heat transport were shut off completely, Europe’s west coast would still be notably warmer than the North American east coast at the same latitude [3]. The main reason is the westerly winds: they carry mild maritime air toward Europe, while in North America the air arrives loaded with continental cold from Canada. The ocean stores heat in summer and releases it in winter, and that seasonal release weighs more in the balance than current transport does.
This doesn’t mean the AMOC is irrelevant. It means that “why is Europe warm today” and “what would happen if the AMOC collapsed” are not the same question, and they shouldn’t be confused. If the AMOC stops, we don’t just lose the current: we lose the global redistribution of heat, and that reorganizes the entire atmosphere.
The evidence for weakening: convergence from independent angles
In 2015, Rahmstorf and colleagues reconstructed the AMOC’s evolution over 1,600 years using surface temperature proxies (sediments, ice cores, corals). Their conclusion: the 20th-century weakening has no precedent anywhere in that record [4]. Three years later, the same group identified the AMOC’s “fingerprint” in observational data: an anomalous cooling in the subpolar North Atlantic —the famous cold blob south of Greenland— occurring alongside accelerated warming along the US east coast [5]. That spatial pattern is exactly what models predict when the current slows down.
Now come the complications. The Florida Current, which flows between Florida and the Bahamas and is a key piece of the system, has been measured by submarine cables almost continuously since 1982. When Volkov and his team reanalyzed that series in 2024, they found that the cable needed a correction for the secular drift of Earth’s geomagnetic field. Once corrected, the Florida Current has remained notably stable for four decades [9]. This doesn’t invalidate the AMOC’s overall weakening, which shows up in other parts of the system, but it’s a perfect example of how science in this field advances through corrections, not through definitive headlines.
The real debate isn’t whether it’s weakening — it’s when it breaks
But when will the AMOC stop or slow down, and how badly will it affect us? Can the AMOC cross a point of no return this century, or is a gradual decline without collapse more likely?
On one side, Peter and Susanne Ditlevsen (University of Copenhagen) extrapolated statistical early-warning signals in 2023 and estimated a collapse with a median date of 2057, with a 95% range between 2025 and 2095 [10]. A year later, van Westen and Dijkstra (Utrecht) forced a high-complexity simulation with the CESM model until they triggered an artificial collapse, and extracted a physical indicator —freshwater transport at the Atlantic’s southern boundary, at 34°S— which, applied to current data, suggests the real system is heading toward that same tipping point [11].
On the other side, Baker and his team at the UK Met Office analyzed a set of 34 CMIP6 models in 2025 under extreme CO₂ and freshwater forcing, and found that in every case a compensating circulation —fed by upwelling in the Southern Ocean, driven by persistent winds— keeps the AMOC from reaching zero [12]. Bonan and colleagues, also in 2025, used observational constraints to adjust projections and reached a similar conclusion: notable weakening, but without crossing the collapse threshold within the 21st century [13]. And in July 2026, Mehling and colleagues incorporated real Greenland meltwater into a high-resolution model for the first time and found that, while the meltwater worsens the weakening, the resulting change is neither abrupt nor irreversible on century timescales [14].
The difference is methodological. The first camp extracts statistical signals from short time series or forces models into artificial collapse to study precursors. The second relies on large ensembles of full models under realistic forcing. The IPCC AR6 assigned “medium” confidence to the idea that there will be no collapse before 2100, precisely because it acknowledges that models tend to underestimate the system’s instability [15].
What would happen if it stopped: it already happened once
This isn’t science fiction. About 12,900 years ago, a massive pulse of freshwater from North American ice melt stopped the AMOC and triggered the Younger Dryas, a cooling episode so sudden that Greenland ice cores show temperatures dropping by up to 10°C within a decade. Steffensen and his team’s analysis of annual ice layers leaves no room for doubt: it was abrupt, not gradual [16]. The planet took more than a thousand years to recover.
With that reference point, current models project the following: an AMOC collapse would cool the North Atlantic by 5 to 10°C within a few decades. Western Europe would lose between 5 and 15°C of average winter temperature. The tropical rain belt would shift southward, disrupting monsoons in West Africa and South America. Sea level on the US east coast would rise an additional 30-50 cm from gravitational and dynamic effects [17]. Van Westen and Baatsen simulated combined scenarios of AMOC collapse with different levels of global warming in 2025, and under an intermediate-emissions scenario, the cooling from the collapse outweighs background warming: one in every ten London winters could approach -20°C, and Oslo could register -48°C, driven by sea ice extending southward [18].
What would that temperature drop mean in practice? As a baseline, winter heating demand across Europe would rise by 50% to 70%; in the Nordic countries this might be manageable, but in central and southern Europe, the power grid as it currently stands would collapse. Scandinavian hydropower, which generates 90% of the region’s electricity, depends on spring snowmelt, which would no longer occur.
Agriculture, meanwhile, would be the biggest loser: fruit trees and crops that need summer growing seasons would disappear or would need to shift 1,000 to 1,500 km south, affecting 500 million people who depend on them. Mortality rises exponentially with temperature extremes, especially from cold; European epidemiological studies estimate that for every degree the temperature drops, mortality rises by between 1% and 3%. Doing the math, it’s easy to work out the implications of a 10-degree drop in winter temperatures.
The AMOC would also disrupt the Sahel and parts of South America, generating food crises in regions already under severe strain, which would likely translate into greater migratory pressure toward Europe.
Solutions: the uncomfortable truth
The only lever that’s been proven to reduce the risk is cutting CO₂ emissions to limit warming and, with it, the melting of Greenland, which currently dumps about 270 billion tons of freshwater into the North Atlantic every year [25]. Every tenth of a degree we don’t warm is salinity we don’t lose. There’s no shortcut.
What is actually being done is monitoring. Besides RAPID, the OSNAP program (Overturning in the Subpolar North Atlantic Program), running since 2014 with participation from the US, UK, France, Germany, the Netherlands, Canada, and China, deploys sensors between Labrador and Scotland. And here’s a finding that surprised the community itself: most of the overturning happens in the eastern Atlantic basin (the Irminger and Norwegian Seas), not in the Labrador Sea as had been assumed since the 1990s [29].
What we know and what we don’t
The AMOC is weakening. This is supported by paleoclimate proxies, the surface thermal fingerprint, direct measurements at four independent latitudes, and physical models. It’s not up for debate.
The future pace, and whether there’s a near tipping point (2050-2070) versus a distant one (post-2100), is an active disagreement between modelers and statisticians. The former tend to be more conservative; the latter, more alarmist on timing.
What isn’t in question: if we cross that threshold, there’s no going back on human timescales. The system’s hysteresis is brutal. Models show that once collapsed, restoring the AMOC would require reducing atmospheric CO₂ well below the level that caused the collapse in the first place. Returning to the starting point isn’t enough. You have to undo more ground than you covered.
The Atlantic has spent a thousand years moving the water that keeps us warm. We’ve spent two hundred altering the boundary conditions. Physics doesn’t negotiate.
References
[1] Bryden, H.L., Longworth, H.R., Cunningham, S.A. (2005). “Slowing of the Atlantic meridional overturning circulation at 25° N.” Nature, 438, 655-657. Reliable. Methodology: direct hydrographic measurements of heat transport across an Atlantic transect at 25°N, comparing cruises from 1957, 1981, 1992, 1998, and 2004. Classic geostrophic calculation supplemented with current meters.
[2] Buckley, M.W., Marshall, J. (2016). “Observations, inferences, and mechanisms of the Atlantic Meridional Overturning Circulation: A review.” Reviews of Geophysics, 54(1), 5-63. Reliable. Methodology: comprehensive review of observational and modeling literature on the AMOC. Synthesis of the field’s consensus state.
[3] Seager, R., Battisti, D.S., Yin, J., Gordon, N., Naik, N., Clement, A.C., Cane, M.A. (2002). “Is the Gulf Stream responsible for Europe’s mild winters?” Quarterly Journal of the Royal Meteorological Society, 128(586), 2563-2586. Reliable. Methodology: atmospheric model (GCM) simulations in which oceanic heat transport in the Atlantic is selectively removed. Comparison of resulting temperatures in Europe and North America. Widely replicated in its qualitative conclusion.
[4] Rahmstorf, S., Box, J.E., Feulner, G., Mann, M.E., Robinson, A., Rutherford, S., Schaffernicht, E.J. (2015). “Exceptional twentieth-century slowdown in Atlantic Ocean overturning circulation.” Nature Climate Change, 5, 475-480. Reliable. Methodology: multi-archive proxy reconstruction (marine sediments, ice cores, corals, instrumental records) of North Atlantic surface temperatures over 1,600 years. AMOC index derived from the spatial pattern of thermal anomalies. Cross-validated against GCMs.
[5] Caesar, L., Rahmstorf, S., Robinson, A., Feulner, G., Saba, V. (2018). “Observed fingerprint of a weakening Atlantic Ocean overturning circulation.” Nature, 556, 191-195. Reliable. Methodology: analysis of observational SST data, 1900-2017. Identification of the spatial pattern predicted by models under a weakened AMOC: subpolar cooling + US coastal warming.
[6] Rahmstorf, S., Jendrkowiak, J., Gou, R., Cheng, L., Ruiz-Angulo, A., Björnsson, H. (2026). “Multidecadal Atlantic ‘Warming Hole’ Heat Content Variations Are Caused by Ocean Heat Transport, Not by Surface Fluxes.” Geophysical Research Letters, 53, e2025GL118383. Recent. [Not currently cited in the body text — see note above.]
[7] Smeed, D.A., McCarthy, G., Rayner, D., Moat, B.I., Johns, W.E., Baringer, M.O., Bryden, H.L. (2014). “Observed decline of the Atlantic meridional overturning circulation 2004-2012.” Ocean Science, 10, 29-38. With reservations. [Not currently cited in the body text.]
[8] Xing, Q., Elipot, S., Johns, W.E., et al. (2026). “Meridionally consistent decline in the observed western boundary contribution to the Atlantic Meridional Overturning Circulation.” Science Advances, 12(14). Recent. [Not currently cited in the body text.]
[9] Volkov, D.L., Smith, R.H., Garcia, R.F., Smeed, D.A., Moat, B.I., Johns, W.E., Baringer, M.O. (2024). “Florida Current transport observations reveal four decades of steady state.” Nature Communications, 15, 7780. Reliable. Methodology: reanalysis of the Florida Current submarine cable series (1982-2023) with correction for secular drift of Earth’s geomagnetic field. The correction removes much of the previously reported downward trend. Transparent methodological correction.
[10] Ditlevsen, P.D., Ditlevsen, S. (2023). “Warning of a forthcoming collapse of the Atlantic meridional overturning circulation.” Nature Communications, 14, 4254. With reservations. Methodology: fitting a stochastic model (Ornstein-Uhlenbeck with multiplicative noise) to the North Atlantic SST series (1870-2022). Probabilistic extrapolation of time-to-collapse. The reservation: the series is short relative to the AMOC’s variability timescales, and the model assumes parameter stationarity that may not hold under increasing forcing. Several groups (Potsdam, Met Office) have challenged the central ~2057 estimate.
[11] van Westen, R.M., Kliphuis, M., Dijkstra, H.A. (2024). “Physics-based early warning signal shows that AMOC is on tipping course.” Science Advances, 10(6), eadk1189. Reliable. Methodology: simulations with CESM (Community Earth System Model) under progressively increasing CO₂. Identification of AMOC freshwater transport at 34°S as a causal indicator of proximity to the tipping point. Validated against observational salinity data.
[12] Baker, J.A., Bell, M.J., Jackson, L.C., Vallis, G.K., Watson, A.J., Wood, R.A. (2025). “Continued Atlantic overturning circulation even under climate extremes.” Nature, 638, 987-994. Reliable. Methodology: analysis of 34 CMIP6 models under extreme CO₂ and freshwater forcing. In every case, a compensating circulation fed by Southern Ocean upwelling (driven by westerly winds) keeps the AMOC from reaching zero.
[13] Bonan, D.B., Thompson, A.F., Schneider, T., Zanna, L., Armour, K.C., Sun, S. (2025). “Observational constraints imply limited future Atlantic meridional overturning circulation weakening.” Nature Geoscience, 18, 479-487. Reliable. Methodology: use of observational constraints (the relationship between internal variability and AMOC sensitivity) to adjust and narrow the spread of CMIP6 projections. An “emergent constraint” approach.
[14] Mehling, O., Bellomo, K., Fabiano, F., Devilliers, M., Petrini, M., Corti, S., von Hardenberg, J. (2026). “Limited impact of Greenland meltwater on abruptness and reversibility of future Atlantic overturning changes.” Science Advances, 12(25), eaed2633. Recent. Methodology: first incorporation of real (rather than idealized) Greenland meltwater into a high-resolution climate model (CMCC-CM2). Result based on a single model; the authors themselves call for replication.
[15] IPCC (2021). Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report. Cambridge University Press. Chapter 9. Reliable. Methodology: evaluated synthesis of climate modeling literature (CMIP6). Multi-model projection under SSP scenarios. “Medium” confidence in the absence of abrupt collapse before 2100.
[16] Steffensen, J.P., Andersen, K.K., Bigler, M., et al. (2008). “High-resolution Greenland ice core data show abrupt climate change happens in few years.” Science, 321, 680-684. Reliable. Methodology: analysis of annual ice layers in Greenland ice cores (NGRIP) at sub-annual resolution. Quantification of the rate of thermal change during Younger Dryas transitions. Classic, widely cited study.
[17] Yin, J., Schlesinger, M.E., Stouffer, R.J. (2009). “Model projections of rapid sea-level rise on the northeast coast of the United States.” Nature Geoscience, 2, 262-266. Reliable. Methodology: simulations with GFDL CM2.1 under an AMOC collapse scenario. Calculation of the gravitational and dynamic adjustment effect on regional sea level.
[18] van Westen, R.M., Baatsen, M.L. (2025). “European temperature extremes under different AMOC scenarios in the Community Earth System Model.” Geophysical Research Letters, 52(12), e2025GL114611. Reliable. Methodology: scenario simulations with CESM2 (NCAR) combining complete AMOC collapse with different levels of global warming. Calculation of winter temperature extremes and sea ice extent. A hypothetical experiment, not a date prediction.
[25] Mouginot, J., Rignot, E., Bjørk, A.A., van den Broeke, M., Millan, R., Morlighem, M., Noël, B., Scheuchl, B., Wood, M. (2019). “Forty-six years of Greenland Ice Sheet mass balance from 1972 to 2018.” PNAS, 116(19), 9239-9244. Reliable. Methodology: Greenland mass balance via satellite data (GRACE, ICESat), regional climate models (MAR, RACMO), and in-situ glacier discharge measurements. A 46-year series.
[26] Soons, J., Dijkstra, H.A. (2026). “The effects of a constructed closure of the Bering Strait on AMOC tipping behavior.” Science Advances, 12(17), eaeb7887. Recent. [Not currently cited in the body text.]
[27] Keefer, B., Wolovick, M., Moore, J.C. (2023). “Feasibility of ice sheet conservation using seabed anchored curtains.” PNAS Nexus, 2(3), pgad053. Reliable. [Not currently cited in the body text.]
[28] Dijkstra, H.A., Weijer, W. (2005). “Mechanisms of multiple equilibria in the thermohaline circulation of a coupled ocean-atmosphere model.” Journal of Physical Oceanography, 35(1), 48-66. Reliable. [Not currently cited in the body text.]
[29] Lozier, M.S., Li, F., Bacon, S., et al. (2019). “A sea change in our view of overturning in the subpolar North Atlantic.” Science, 363(6426), 516-521. Reliable. Methodology: OSNAP array data (2014-2016), the first direct measurement of overturning transport in the subpolar region between Labrador and Scotland. Current meters, profilers, gliders, and hydrography. Finding: most of the overturning occurs in the eastern basin, not the Labrador Sea.