The last leap second was inserted at the end of 2016. There has not been one since, and that is not an administrative lapse. The Earth simply stopped needing it.
Which is a strange sentence, so it is worth unpacking — along with the fact that the next correction, if one comes, may have to run the other way.
Two kinds of time, quietly disagreeing
There are two ways to answer “what time is it”, and they do not stay in step.
Atomic time is counted. Since 1967 the second has been defined as 9,192,631,770 periods of a caesium transition, and caesium clocks tick that out with monstrous consistency, indifferent to anything happening on the planet they sit on.

Solar time is observed. It is the actual rotation of an actual Earth — a lump of rock with a liquid core, oceans that slosh, an atmosphere that shifts, and a Moon that has been gently applying the brakes for four billion years.
The rock is not a good clock. Its day is not exactly 86,400 atomic seconds, and by how much varies. So atomic time drifts away from the sky, and if nothing intervened, the accumulated difference would eventually become visible.
Leap seconds are the intervention: an extra second inserted into UTC to keep it within 0.9 seconds of the Earth’s actual rotation. Twenty-seven have been added since the system began.
Worth correcting one popular framing: they do not exist to keep noon at noon. Left entirely alone, civil time and solar time would separate by a little over a minute across an entire century. Leap seconds serve systems that need the sky and the clock to agree tightly — astronomy, navigation, satellites — not people wondering whether lunch is late.
Not always in December
The first arrived on 30 June 1972, and 1972 had two of them.
The common belief that they all land at the end of December is wrong: eleven of the twenty-seven have been at the end of June. They are announced by the International Earth Rotation and Reference Systems Service in a document called Bulletin C, which appears twice a year and mostly says nothing is happening. Bulletin C 72, issued on 6 July 2026, is a recent example: no leap second at the end of this year.
What one extra second does to a computer
The trouble is that a great deal of software has been written on the quiet assumption that minutes have sixty seconds.
The 30 June 2012 leap second is the famous case. The travel booking platform Amadeus went down for more than two hours, and Qantas — one of the airlines running on it — fell back to checking passengers in by hand. Reddit went offline. Mozilla, Gawker and others reported failures. The underlying cause was not one bug but a stack of them — Reddit pointed at Java and Cassandra, while much of the wider damage traced to a Linux kernel timer bug that a repeated second could trigger.
It happened again, more narrowly, at the start of 2017, when Cloudflare’s DNS resolver was hit: code that assumed time never runs backwards was handed a negative interval and behaved badly.
The pattern is consistent. The leap second rarely breaks the clock. It breaks the assumption, several layers up, that time is monotonic and uniform.
Large operators mostly gave up on handling the discontinuity honestly and now smear it — spreading the extra second across a whole day so no clock ever repeats a value. It works. It also means the seconds those systems hand out during a smear are not, strictly, SI seconds.
The decision to stop
In November 2022 the General Conference on Weights and Measures (CGPM) resolved to end the practice: the tolerance between atomic and solar time will be widened, in or before 2035, so that corrections effectively stop for centuries.
It was close to unanimous rather than unanimous — Russia voted against, Belarus abstained.
A further proposal is in front of the 28th CGPM, which meets on 13–15 October 2026 — at the time of writing it is still a draft resolution and has not been adopted. The draft would widen the tolerance to a full hour with effect from 2027, which in practice means nobody alive will see another correction. Until that meeting happens, the honest description is proposed, not agreed.
And then the Earth sped up
Here is where it turns strange.
Since around 2020, the Earth has been rotating faster, not slower — enough that recent days have been among the shortest ever recorded since atomic timekeeping began. The cause is thought to lie in the motion of the liquid core.

If that continues, UTC would eventually need the opposite of a leap second: a negative one, in which a minute lasts fifty-nine seconds and 23:59:59 never happens.
One has never been done. Nobody particularly wants to find out what breaks. Every piece of software that stumbled over a repeated second in 2012 would be meeting, for the first time, a second that simply is not there.
Melting ice is buying us time
The genuinely lovely part is in a 2024 Nature paper by the geophysicist Duncan Agnew, with a title that reads like a joke and is not: “A global timekeeping problem postponed by global warming.”
Ice at the poles is melting, and that water redistributes toward the equator. A spinning body with mass moved outward from its axis slows down — the physics a skater uses when they extend their arms to come out of a spin. The planet is doing the same thing, and it is partially cancelling the speed-up from the core.
Agnew’s estimate is that this has pushed the likely need for a negative leap second out to around 2029 — later than it would otherwise have arrived. Melting ice sheets are, incidentally, delaying a timekeeping crisis.
That projection deserves its caveats, and they matter. It assumes UTC keeps its current definition, which the 2026 proposal would change first. The core’s behaviour may shift again — one analysis already suggests the Earth has begun decelerating. A former chief scientist of the US Naval Observatory has publicly doubted any date can be given, on the grounds that the Earth is too unpredictable. The formal assessment puts the probability of needing a negative leap second before 2035 at around 30 per cent.
So the most likely outcome is that leap seconds end not with a dramatic negative one, but by quietly ceasing to be required — a rule retired shortly before the case that would have tested it.
What this is really about
Underneath the engineering it is a question about what a clock is for.
Atomic time is more accurate in every sense that can be measured. Solar time is less accurate and happens to be the one the sun keeps. For fifty years we have insisted the two stay reconciled, at the cost of a correction that periodically breaks computers. The decision now being taken is to let them separate — to accept that civil time is a human agreement that no longer has to be checked against the sky every few years.
That is a bigger change than one second sounds like. It is the same shift the railways made when they replaced local noon with a shared standard, only larger: cutting the last formal tie between the clock on the wall and the rotation of the planet.
None of it touches an elapsed time. When you time something with TiCaNo Stopwatch, you are measuring a difference between two moments, and a difference does not care whether the calendar behind it gained a second or lost one. It is the same reason a stopwatch survives crossing a time zone without noticing: the interval between two events is one of the few things in timekeeping that nobody has to agree on.
Sources
- Leap seconds and UT1-UTC information — NIST
- IERS EOP 20 C04 series — International Earth Rotation and Reference Systems Service, the daily length-of-day record the chart above is drawn from
- IERS Bulletin C and Leap_Second.dat — International Earth Rotation and Reference Systems Service, the canonical record of every leap second
- Resolution 4 of the 27th CGPM (2022) — BIPM, on ending leap seconds in or before 2035
- Draft Resolutions of the 28th CGPM (2026) — BIPM; not yet decided at the time of writing
- Network-crashing leap seconds to be abandoned by 2035 — Ars Technica, on the vote and smearing
- Leap second glitch causes site and software crashes — CNET, 1 July 2012
- Leap second bug hits Amadeus, Qantas and Reddit — The Guardian, 2 July 2012
- How and why the leap second affected Cloudflare DNS — Cloudflare postmortem
- A global timekeeping problem postponed by global warming — Duncan Agnew, Nature 628:333–336 (2024)
- Melting ice is slowing Earth’s rotation — NBC News, March 2024