The Death of the Leap Second: Versailles Vote in October, UTC Goes Seamless from May 2027

Forum topic · QianXun · 2026-09-01

Summary

The 28th General Conference on Weights and Measures (CGPM), meeting October 13–15, 2026 in Versailles, will vote on Draft Resolution C to abolish the leap second, with continuous UTC taking effect on May 20, 2027. Since 1972, 27 leap seconds (all positive) have been inserted to reconcile atomic time (UTC) with Earth's irregular rotation (UT1); the last was added on December 31, 2016, and the current TAI−UTC offset is 37 seconds. The urgency stems from Earth's unexpected rotational speedup around 2015, which made a never-tested negative leap second a real risk—experts estimated roughly 30% probability by 2035. However, recent IERS data show the acceleration has stalled: the shortest recorded day was July 5, 2024 (−1.65 ms), improving to +0.29 ms by 2026. The new resolution replaces the ±0.9-second tolerance with a 3,600-second 'leap hour' scheme, effectively eliminating corrections for centuries. Geopolitics matters too: Russia opposed the 2022 resolution because GLONASS embeds leap seconds natively, but this vote uses simple majority rules. Observatories will face added conversions, while financial and internet infrastructure benefit from a clean, monotonic timescale.

From October 13–15, 2026, at the Palais des Congrès in Versailles near Paris, the 28th General Conference on Weights and Measures (CGPM) will vote on a matter affecting every server on Earth: abolishing the leap second. A Nature piece dated August 31 headlined "could be scrapped as early as October"—but the wording matters. October's vote is only the decision; BIPM Draft Resolution C sets the effective date at May 20, 2027. Nothing changes the moment the vote passes, but afterward, the timestamp 23:59:60 will never appear on a calendar again.

How the mechanism works

Earth's rotation (UT1) is not uniform; atomic clocks (UTC) are. When the two drift apart by ±0.9 seconds, an extra second is inserted, announced via IERS Bulletin C. Since 1972 there have been 27 insertions—11 at the end of June, 16 at the end of December, all positive. The last was on December 31, 2016. The TAI−UTC offset currently stands at exactly 37 seconds. If the resolution passes, that record will be permanent.

1972 — first leap second inserted; all 27 insertions since have been positive
2015 — Earth's rotation begins speeding up; potential negative leap seconds enter the picture
2016-12-31 — 27th and (so far) final leap second
2022 — CGPM resolves to enlarge the tolerance by 2035; Russia votes against
2024-07-05 — shortest day ever recorded, 1.65 ms short of standard
2026-10-13 — Versailles vote on the continuous-UTC resolution
2027-05-20 — if passed, continuous UTC takes effect; the leap second retires

Why the rush

Around 2015, Earth's rotation inexplicably sped up, turning the negative leap second (dropping a second, jumping from 23:59:58 straight to 59:00) from pure theory into a genuine possibility. At a March workshop, CCTF and IERS experts put the probability of a negative leap second by 2035 at roughly 30%, based on three independent modeling groups; the report added that preparation costs would be "comparable to Y2K." Humanity has never inserted a negative leap second—and untested scenarios are precisely what infrastructure fears most. On New Year's Day 2017, Cloudflare was hit by a *positive* leap second: Go's time.Now() produced negative durations at 23:59:60, the RRDNS daemon panicked, and 0.2% of DNS queries failed at peak before engineers fixed it globally by 06:45 UTC. If a positive second caused that, nobody wants to gamble on a negative one.

But there's a twist

Earth has recently slowed back down. According to the IERS C04 series: the shortest day ever was July 5, 2024, at −1.65 ms below 86,400 seconds; July 9, 2025 came in at only −1.37 ms (the widely circulated "2025 record" is wrong—even English Wikipedia still carries the 2022 figure); by 2026, the average day length deviation was +0.29 ms, with UT1−UTC at just +0.013 s. The acceleration has stopped. So this sprint to abolish the leap second guards against an untestable tail risk, not an imminent certainty. Incidentally, the same experts gave roughly 50% odds of a *positive* leap second within a decade—far higher than for a negative one—but positive seconds are familiar, so nobody panics.

2024 shortest day: −1.65 ms
2025 shortest day: −1.37 ms
2026 daily average: +0.29 ms

The replacement is an old ghost: the "leap hour"

The new resolution widens the tolerance from 0.9 seconds to 3,600 seconds—a 4,000-fold increase—meaning UTC and Earth's rotation would only be reconciled after drifting a full hour apart, which at current drift rates won't happen for centuries. The US delegation first proposed this at ITU-R in 2004, where it was widely mocked; Steve Allen's leapsecs site documents the whole saga. Twenty-two years later, the proposal that survived the mockery has entered a CGPM draft resolution.

International politics adds drama: in 2022, Russia voted against (its counter-draft demanded "no earlier than 2040 with no deadline," sharply criticized by Poland's delegate) and Belarus abstained; GLONASS system time embeds leap seconds natively, so changing it is structurally painful. WRC-23 struck a compromise with a 2035 cap and radio stations maintaining broadcasts until 2040. This vote proceeds under Rule 18.5 by simple majority, with abstentions excluded from the denominator—Russia's opposition cannot block it, only be recorded again.

Old rules (0.9 s tolerance): UT1–UTC approaches 0.9 s → IERS issues Bulletin C → 23:59:60 inserted in June or December.

New rules (Draft Resolution C, 3,600 s tolerance): drift accumulates for an hour before correction → a leap hour, centuries away → UTC becomes continuous, with no 23:59:60.

Who wins, who loses

Observatories must relearn to consult tables for DUT1, adding a conversion step for telescope pointing. Financial systems gain a clean timeline with no 23:59:60 sorting headaches—though a reading of MiFID II RTS 25 finds no mention of leap seconds at all; the popular claim that "financial regulation requires abolishing the leap second" is weakly evidenced. The real beneficiaries are exchange calendars (ICE still maintains a leap-second page). And some of us will miss the feeling of those 27 time capsules. Oh well—progress always leaves someone feeling nostalgic.