What is a leap second?

Why does the leap second exist?

The leap second exists to keep two different ways of measuring time from drifting too far apart. Coordinated Universal Time (UTC) ticks at the steady rate of atomic clocks — the international second has been defined since 1967 by the resonant frequency of caesium atoms, and an atomic clock ticks at that rate forever. The Earth's rotation, by contrast, is slightly irregular. Tides slow it gradually over centuries; the seasonal redistribution of air and water across the planet speeds it up and slows it down by milliseconds within a single year; large earthquakes nudge it by smaller amounts still.⁷

Civil time has historically been tied to rotational time — the Sun should be roughly overhead at noon — but a clock that ticks at the rate of the Earth's actual rotation cannot be a stable physical reference. UTC reconciles the two by ticking at the atomic-clock rate while inserting an extra second every so often to keep the cumulative drift bounded. The agreed limit, set by an international standard for time-signal broadcasts, is that UTC must stay within 0.9 seconds of UT1 — the modern realisation of mean solar time on the Greenwich meridian, computed from observed star transits with corrections for the wobble of the Earth's spin axis.¹ Each leap second adds one to the integer offset between UTC and the underlying continuous atomic-time scale, called International Atomic Time (TAI). The ticks themselves never change — only the count of how many extra seconds have been inserted along the way.¹

How is a leap second inserted?

A positive leap second is inserted by lengthening the last minute of a UTC day by one second. The clock counts the extra second as 23:59:60, sitting between the usual 23:59:59 and 00:00:00 of the following day, so the minute containing the leap second runs sixty-one seconds rather than sixty.¹ Outside that one minute, every UTC second is an SI second of identical length; the insertion is a one-off adjustment to the count of seconds, not a change to their rate.

Insertions are restricted to four candidate dates per year, in order of preference: the end of 31 December, the end of 30 June, with the end of 31 March or 30 September allowed in principle if either of the first two would not be enough.¹ In practice, every leap second since 1972 has fallen at the end of June or the end of December.² The decision is announced by the body responsible — the International Earth Rotation and Reference Systems Service, or IERS — through a short two-paragraph bulletin issued from the Paris Observatory at least six months in advance.¹⁵

A negative leap second, were one ever needed, would shorten the last minute by one second instead — the clock would jump from 23:59:58 straight to 00:00:00, skipping 23:59:59 entirely.¹ The mechanism has been part of the standard since 1972 but has never been invoked.⁶

How many leap seconds have been added?

Twenty-seven leap seconds have been added to UTC between 1972 and 2017. UTC therefore currently runs 37 seconds behind International Atomic Time. That 37 is the sum of two parts: a 10-second seed value applied at the start of 1972 — which carried forward the drift that had accumulated between atomic clocks and the Earth's rotation since the atomic-time scale itself was anchored in 1958 — and the 27 leap-second insertions since.⁷² The first true leap second was inserted at the end of 30 June 1972; the most recent at the end of 31 December 2016.⁶²

The cadence has not been uniform. Through the 1970s the IERS announced a leap second nearly every year, often at both candidate dates. Through the 1990s the rate slowed to about one every eighteen months. Since 1999 leap seconds have arrived in scattered pairs and singletons, and there has now been an unbroken nine-year gap since the last one. The reason is that the Earth's rotation has been faster, on average, in the most recent decade than in any prior decade since the atomic clock was invented; the body that monitors the Earth's actual rotation has noted a recent acceleration that owes partly to changes in the planet's liquid core and partly to a redistribution of mass at the surface.²⁶

The latest issue of the IERS bulletin that announces upcoming insertions — Bulletin C 71, issued from the Paris Observatory on 6 January 2026 — confirms that no leap second will be inserted at the end of June 2026. The current TAI–UTC offset of −37 seconds therefore continues "until further notice."⁵

Who decides when a leap second happens?

Three international bodies share the work, by treaty.

The body that names the candidate dates and the 0.9-second tolerance is the international agency responsible for radio time-signal broadcasts. Its standing recommendation on standard-frequency and time-signal emissions is the operational specification of how leap seconds work; the current edition was issued in 2002 and remains the governing document.¹

The body that announces each individual leap second is the International Earth Rotation and Reference Systems Service. The IERS measures the Earth's actual rotation by combining observations from very long baseline interferometry, satellite laser ranging, and global navigation satellite systems, and projects the gap between UT1 and UTC forward six to twelve months. When that projection is about to break the 0.9-second band, the IERS issues a numbered bulletin from the Paris Observatory — Bulletin C — declaring whether a leap second will be inserted at the next candidate date. Bulletin C is the operational record of when leap seconds happen and the file every NTP server, every operating system, and every astronomical software package consults.¹⁵

The body that computes UTC itself is the Bureau International des Poids et Mesures (BIPM), an intergovernmental measurement bureau outside Paris. The BIPM averages the readings of more than 300 atomic clocks at over 80 national time laboratories around the world to produce International Atomic Time, then derives UTC from it by subtracting the integer leap-second offset published by the IERS.⁴ Each national laboratory in turn keeps its own working realisation of UTC — UTC(NIST) at the U.S. National Institute of Standards and Technology, UTC(USNO) at the U.S. Naval Observatory — synchronised against BIPM's official UTC to within about twenty nanoseconds.⁴

What happens to leap seconds in 2035?

The leap second is on its way out. In November 2022 the global body that governs measurement units — the General Conference on Weights and Measures, which sets the definitions of the metre, the kilogram, and the second — passed a resolution to retire it. The maximum allowed gap between UTC and the Earth's rotation will be widened, in or before 2035, to a value chosen so that no further leap seconds will be needed for at least a century.³

The motivation is operational. Leap seconds are awkward in computer systems, and the resolution names that risk explicitly: the discontinuities they introduce "risk causing serious malfunctions in critical digital infrastructure including the Global Navigation Satellite Systems, telecommunications, and energy transmission systems."³ Different cloud providers, operating systems, and time-server vendors have adopted different ad-hoc strategies for absorbing each leap second, and the resulting inconsistencies have caused real outages — the largest being a multi-hour disruption at the social-news site Reddit at the moment of the June 2012 leap second, traced to a Java thread-scheduling bug exposed by the second's backwards step, and a global outage of Cloudflare's public DNS resolver at the moment of the December 2016 leap second, traced to a code path that assumed time only moves forward.⁸

The push to retire the leap second has been a long one — discussion within the international metrology community began in 1999 — and the 2022 vote settled a debate over twenty years old.⁷ The remaining work is to choose the new tolerance value and write the implementation plan; that is expected to come before the next General Conference, which meets in November 2026.³

The change does not affect existing timestamps. The 27 leap seconds inserted between 1972 and 2017 stay in the historical record, and code that handles past leap seconds correctly today will continue to do so. The change retires future insertions only.³

How do computers handle leap seconds?

Because Unix time — the way nearly every computer represents an instant in time internally — has no encoding for the leap second, every operating system that synchronises against UTC has to do something with the second when it arrives. Three strategies are in use.

The simplest is to step the clock. At the leap-second instant, the system's seconds counter is held for one second, so that 23:59:60 UTC and the following 00:00:00 UTC share the same Unix-time value. Programs reading the clock during the leap see a one-second backwards jump; ordering is no longer monotonic. This is the historical default behaviour of NTP-disciplined Unix systems.⁹

A more sophisticated approach is to smear the leap second across a longer interval by running the clock slightly slow. Google has used a smear since 2008, currently a 24-hour linear smear from noon UTC to noon UTC, with each smeared second running about 11.6 microseconds longer than an SI second — a frequency offset of roughly 11.6 parts per million.¹⁰ During the smear interval Google's clocks disagree with non-smearing UTC sources by up to half a second, but no client ever sees a non-monotonic clock or a 23:59:60 timestamp. Meta runs a comparable scheme over a 17-hour quadratic smear, and Amazon Web Services and Microsoft Azure run their own variants.⁸¹⁰ The internet engineering body that maintains the Network Time Protocol formally cautions that smeared and non-smeared NTP servers must not be mixed in a single client's configuration, and that smeared servers "MUST NOT be used for public-facing NTP servers" — the smear is an internal optimisation, not a UTC source.¹¹

A third route is to expose the underlying continuous atomic-time scale. POSIX systems with a CLOCK_TAI clock surface the count of SI seconds elapsed since the same 1970 epoch, with no leap-second adjustments — equivalent to Unix time plus the current TAI–UTC offset. Code that needs strictly monotonic SI-second behaviour reads CLOCK_TAI directly rather than CLOCK_REALTIME, sidestepping the leap-second question entirely.¹¹

The ISO 8601 standard for internet timestamps preserves 23:59:60 in the seconds field "at the end of months in which a leap second occurs", so a UTC leap-second instant can be written explicitly as 2016-12-31T23:59:60Z.¹² One detail worth knowing: the IANA Time Zone Database, which encodes every region's history of UTC offsets and daylight-saving rules and is used by every major operating system, does not encode leap seconds. Application code that needs to handle leap seconds correctly — typically scientific or astronomy software — uses TAI, GPS time, or a vendor-specific monotonic clock instead.¹³

Could there ever be a negative leap second?

In principle, yes. The 1972 standard provides for a negative leap second exactly as it provides for a positive one: if UT1 ever ran ahead of UTC by close to 0.9 seconds, the IERS would announce a one-second deletion at the next candidate date and the last minute of that day would run from 23:59:58 straight to 00:00:00, with no 23:59:59.¹

In practice it has never happened. Every leap second since 1972 has been positive — added rather than removed — because the long-term trend has been for the Earth to rotate slightly more slowly than the atomic-clock day. That trend reversed around 2020. The Earth has been rotating faster, on average, since then than at any time in the era of atomic measurement, and the IERS noted that without the partial offset of polar ice melt — which redistributes mass towards the equator and acts as a brake on the rotation — a negative leap second would already have been needed by about 2026.⁶

Whether one is ever called for now depends on how long the current acceleration persists and on how quickly the planned 2035 retirement of the leap second arrives.³ If the acceleration continues and the retirement is delayed, a negative leap second is possible before 2035. If the acceleration reverses, it is not. Either way, the leap-second-handling code that has been refined over fifty years to absorb a positive leap second is unlikely to be tested in the negative direction in production.

Frequently asked questions