Johann Páll Jóhannsson, Iceland’s climate minister, didn’t sound alarmist. He sounded procedural. In Reuters’ reporting, he called the AMOC a direct threat to national resilience and security.⁴ The unsettling detail isn’t the phrase “existential threat.” It’s that he used it the way governments use any other planning term—calmly, without drama.³
Once the ocean enters your national security meetings, it has already left the realm of “environment” and entered the realm of “state.”
The AMOC is not a curiosity. It’s infrastructure made of physics.
It runs on density. Warm, salty water flows north, sheds heat to the atmosphere, and in key regions becomes dense enough to sink, driving the deep return flow. Add freshwater—especially from Greenland melt—and the surface dilutes, sinking becomes harder. Warm the air above the ocean and winter heat loss weakens. Nothing has to break. The machine just has to lose the conditions that once kept it predictable.
This is where the argument turns uncomfortable.
Public debate often splits into caricature: collapse tomorrow versus collapse never. The science is narrower and less soothing. It’s about definitions—collapse versus weakening—and about thresholds that don’t announce themselves politely. It’s also about risk with a long tail: outcomes that are unlikely, but catastrophic if they occur.
The IPCC’s position reflects that caution. A substantial weakening of the AMOC this century is considered very likely; an abrupt collapse before 2100 is judged very unlikely.² That does not make a large weakening harmless, and it leaves open what follows if emissions remain high beyond mid-century.
In 2025, a prominent Nature study—widely covered by the Associated Press—pushed back against the most dramatic near-term collapse headlines. In those simulations, a full shutdown this century looked unlikely, even under severe forcing, partly because of stabilizing mechanisms in the Southern Ocean.⁶ But the same reporting made something else clear: major weakening would still be deeply disruptive. “Unlikely” is not the same thing as “safe.”⁶
At the same time, early-warning studies have been flashing quieter signals. In 2023, Peter and Susanne Ditlevsen argued that statistical indicators of lost resilience could imply a tipping transition sometime around mid-century, with a wide and uncomfortable uncertainty range.⁷ The work is debated for good reasons. It resonates for a simpler one: systems like this don’t fail gracefully. They fail, and then they stay failed.
This isn’t a forecasting exercise with a clean answer. It’s closer to underwriting risk on a machine with multiple operating modes and no reset button.
Which is why “Gulf Stream collapse” remains such a tempting phrase. It promises clarity. The ocean offers none.
The impacts, though, aren’t abstract. They just don’t arrive as a single disaster.
Along the U.S. East Coast, a slowing circulation can raise local sea levels without waiting for global averages to catch up. The IPCC is explicit about this.² Storm surge rides in on a higher baseline. “Nuisance flooding” crosses into structural damage more often than planners expect.