Skip to content
audiotale.net
Sign in
NewsNews

Leap seconds end: the CGPM votes to make UTC continuous from 20 May 2027

The General Conference on Weights and Measures voted on 15 October to stop adding leap seconds and let UTC drift up to one hour from Earth's rotation.

audiotale.net team

· 6 min read

Written by AI from the sources linked below. How we use AI

On this page
  1. Key points
  2. What happened
  3. Background
  4. The numbers
  5. What it means
  6. What we don't know yet
  7. What to watch next
  8. Sources

The General Conference on Weights and Measures (CGPM) voted on 15 October 2026 to stop adding leap seconds to Coordinated Universal Time (UTC). From 20 May 2027, according to Euronews, UTC will run continuously, and the gap between UTC and the time set by Earth's rotation will be allowed to grow to as much as an hour. For anyone who writes software that handles timestamps, runs servers or maintains time-sync infrastructure, it removes one of computing's most notorious special cases, and it replaces the risk of a first-ever negative leap second with a long-term, slower question: what to do when clocks and the Sun drift apart.

Key points

  • The CGPM adopted the resolution on Thursday 15 October 2026 at its 28th meeting in Versailles, which ran 13 to 15 October, according to Euronews and the BIPM's meeting page.
  • UTC is to run continuously, with no inserted leap seconds, from 20 May 2027, per Euronews and the draft resolution as described by the blog lilting.ch.
  • The allowed difference between UTC and UT1 (time based on Earth's rotation) rises from 0.9 seconds to 3,600 seconds, one hour, according to the draft as described by lilting.ch and The Register.
  • Leap seconds were added 27 times from 1972, most recently at the end of 2016, Euronews reports.
  • A negative leap second, which has never been applied, had become a real possibility because Earth has recently been spinning slightly faster; the decision was partly meant to avoid it.

What happened

The CGPM, the treaty body of the Metre Convention that meets roughly every four years, held its 28th meeting in Versailles from 13 to 15 October 2026, as the BIPM's meeting page states. Euronews reports that delegates voted on Thursday 15 October to end leap seconds, and that from 20 May 2027 UTC will run without interruption. The Register, writing on 19 October, says the conference voted to stop introducing them and that the replacement scheme, which it calls Continuous UTC, comes into force next year. Neither report gives a vote tally, and we could not open the adopted text on the BIPM site, which on its documents page still listed only the July 2026 draft resolutions. The details below about the text therefore come from reports of that draft; we mark where that matters.

Ending the insertions does not end UTC. As lilting.ch explains in its walk-through of the draft, UTC stays the world's civil time scale, but the difference between it and UT1 is now allowed to widen instead of being pulled back to within 0.9 seconds by a leap second. The draft sets the new limit at 3,600 seconds, and the blog says that is enough to keep UTC continuous for several centuries.

Background

UTC is built from atomic clocks, which tick at a steady rate. UT1 follows the planet's actual rotation, which varies. To keep the two within 0.9 seconds, the international timekeeping system has occasionally inserted an extra second at the end of 30 June or 31 December. Euronews counts 27 such insertions since 1972, the last at the end of 2016. Announcements come from the IERS, which publishes its Bulletin C roughly twice a year.

Leap seconds have always caused trouble for software. Lilting.ch recalls the 2012 insertion, when the Linux kernel's high-resolution timer code confused machines and locked up CPUs. Large cloud operators avoid the 23:59:60 problem by "smearing" the extra second: the blog says Google has smoothed it over 24 hours, noon to noon UTC, since 2008, AWS uses the same 24-hour window, and Meta uses 17 hours starting at 00:00 UTC. A smeared clock can sit up to half a second from true UTC, and clocks with different windows disagree with each other. The draft's own considerations, as described by lilting.ch, say these differing workarounds make distributed time inconsistent.

In 2022 the CGPM's 27th meeting adopted Resolution 4, which, as lilting.ch and search listings describe it, called for raising the maximum allowed UT1-UTC difference in or before 2035. Lilting.ch says the 2022 text asked for UTC to be continuous for at least a century. The October 2026 vote supplied the missing numbers and a start date that comes years earlier than 2035.

The numbers

Why act earlier than planned? Earth has recently been rotating slightly faster, which means that instead of adding a second the world might have had to remove one. Euronews says a negative leap second has never been applied and it was unclear how reliably technical systems would cope. The Register quotes the resolution as describing a negative leap second as high risk. Lilting.ch reports that the draft's considerations cite a March 2025 workshop of the CCTF (the BIPM's time and frequency committee) and the IERS, which estimated that the probability of a negative leap second would rise rapidly, reaching 30 percent by 2035. That is a workshop estimate relayed through a draft, not a measurement, and the blog gives no method for it.

Allowed UT1-UTC difference, before and afterseconds
Allowed UT1-UTC difference, before and after (seconds)
seconds
Old limit0.9
New limit3600

Source: Draft Resolution C as described by lilting.ch; The Register

The Register adds a cost argument: according to its article, the CGPM estimated that preparing industry for a negative leap second would cost about what it spent on the millennium bug, which the article says historians put at $300 to $600 billion worldwide and which many saw as poorly spent. We have not seen the underlying estimate, so treat that comparison as the Register's report.

What it means

For developers: the practical effect, after May 2027, is that no new leap second will appear. Libraries and operating systems that already ignore leap seconds (counting a smooth timeline) or smear them are unaffected. Code that handles 23:59:60 or reads leap-second tables will no longer be exercised by real events, so it becomes dead code that can rot; those tables still matter for converting historical timestamps correctly, because the 27 existing insertions stay in the record.

For infrastructure operators: NTP daemons and cloud time services that smeared the leap second no longer need to schedule that work. The Register notes that billions of IoT devices may not handle leap seconds at all, many of them insecure or hard to update, and that digital certificates carry a "not before" time, so a device whose clock is wrong can have a certificate rejected. Removing the event removes that failure mode for devices that never get a patch.

For astronomers, geodesists and satellite users: Euronews says these fields must track the growing gap between clock time and astronomical time themselves when needed. Everyday civil time does not change: Euronews notes that in the UK, for example, GMT and BST continue and the clocks go back on 25 October 2026 as usual.

Who loses: anyone who built a product on the assumption that UTC and Earth rotation stay within a second. After a long enough period the gap could reach a full hour, though Euronews says experts expect it will be some time before any correction is needed, and calls the measure essentially precautionary.

What we don't know yet

We have not read the adopted resolution, so we cannot confirm the vote count, the exact wording of the 3,600-second limit or how often it will be reviewed. The Register and Euronews differ in emphasis: one gives only "next year", the other a specific date of 20 May 2027, which matches the draft. We also do not know what the standards bodies behind NTP, PTP and POSIX time will publish, how GNSS operators will update their time messages, or whether cloud providers will keep smearing after 2027. The 30 percent figure is a workshop estimate, as noted, and a 2024 Nature paper by Duncan Agnew that lilting.ch cites extrapolates a negative leap second as early as 2029; we did not open that paper.

What to watch next

Watch for the BIPM publishing the adopted text, for the IERS announcing how its bulletins change before May 2027, and for operating system, kernel and time-service maintainers noting the change in release notes. Anyone running systems that parse leap-second files should decide now whether to keep that code path. We will update this article when the official text appears.

Sources

SaveStart the discussion

Written by

audiotale.net team

The official account of the people who run audiotale.net: announcements, conversation starters and the newsroom's drafts. Posts written with AI say so. Every other account here is a real person.

Read next