On July 6, a quiet bulletin arrived from an observatory in Paris that almost nobody outside the timekeeping community reads — and immediately lit up Hacker News. Bulletin C number 72, issued by the International Earth Rotation and Reference Systems Service (IERS), carried a single consequential line: no leap second would be introduced at the end of December 2026.

The announcement landed at a charged moment. Scientific American reported on July 8 that international timekeeping authorities are preparing to vote at this October's General Conference on Weights and Measures (CGPM) on a proposal to replace the leap second — a 54-year-old mechanism for keeping clocks aligned with Earth's rotation — with a far rarer adjustment called a leap hour, targeting implementation as early as 2027. That vote, largely ignored outside specialist circles, could determine whether the global technology infrastructure faces its most consequential and least-tested clock event in history — or sidesteps it entirely.

The reason the stakes are this high comes down to an anomaly in Earth's behavior. Since roughly 2016, the planet has been spinning measurably faster than it has at any point in the atomic-timekeeping era. Earth set a new rotation speed record on July 4, 2024, completing its daily spin 1.66 milliseconds faster than the standard 86,400 seconds. That record was broken again by July 2025, when multiple days — including July 9 and July 10 — clocked in more than a millisecond shorter than usual. If this trend continues, the world's timekeepers will face something that has never happened in 54 years of the leap second system: a negative leap second — one in which a second is removed from the clock rather than added.

Nobody has built software to survive that.

Why Earth Is Running Ahead of Its Own Clock

To understand the problem, consider that two entirely different systems currently define what "time" means — and that they are slowly drifting apart.

The first is Coordinated Universal Time (UTC), which runs on atomic clocks. One second in atomic time is defined as exactly 9,192,631,770 oscillations of a caesium-133 atom. This value never varies. The second system is Universal Time 1 (UT1), which tracks how long Earth actually takes to complete one full rotation. UT1 changes constantly, influenced by the gravitational pull of the moon, seasonal atmospheric shifts, ocean currents, and motions in the planet's liquid outer core.

For most of history, Earth's rotation has been gradually slowing — tidal friction from the moon drags on Earth's oceans, braking the planet's spin by roughly two milliseconds per century. This created a familiar one-way problem: atomic time kept advancing steadily while Earth fell slightly behind, so the IERS periodically added a "positive leap second" to UTC to keep the two systems within 0.9 seconds of each other. Since the system was introduced in 1972, 27 such seconds have been inserted — all of them positive — with the most recent on December 31, 2016.

Around 2016, that direction reversed. Earth began spinning faster, most likely because its liquid outer core began slowing down. By a mechanism rooted in conservation of angular momentum — the same principle that causes a spinning ice skater to speed up when drawing in their arms — a decelerating core allows the solid outer layers to rotate faster. Atmospheric seasonal patterns and the redistribution of water mass from polar ice melting also play roles. Climate change has provided a modest brake: meltwater spreading toward the equator acts like a skater extending their arms, slightly counteracting the acceleration. Duncan Agnew, a professor emeritus of geophysics at UC San Diego's Scripps Institution of Oceanography, calculated in a 2024 paper in Nature that without polar ice melt, Earth would have required a negative leap second by 2026 rather than 2029.

"We've been on a trend toward slightly faster days since 1972," Agnew told CNN. "But there are fluctuations. It's like watching the stock market, really."

Why a Negative Leap Second Is Categorically Different

Positive leap seconds pause the clock for one tick: instead of the sequence 23:59:59 → 00:00:00, UTC adds a 23:59:60 step. This extra timestamp has no place in the normal 60-seconds-per-minute model, and for half a century it has tripped up software that wasn't designed for it.

A negative leap second would do the opposite: instead of the clock pausing, it would skip forward. The sequence would run 23:59:57 → 00:00:00, with 23:59:58 and 23:59:59 simply not existing. The last minute of the affected UTC day would contain only 59 seconds.

The engineering implications are different in kind, not just degree. The documented failures from positive leap seconds — Reddit offline for 30 to 40 minutes in 2012, LinkedIn and Mozilla crashing the same night, more than 400 Qantas flights delayed when Amadeus Altéa's reservation system buckled — all involved software that couldn't handle time appearing to pause. The patches written after those incidents were designed for the pause case, not for the jump case. When Cloudflare's DNS resolver failed at the December 2016 leap second, it was because Go's time.Now() function used only a real-time clock source and returned a negative elapsed-time value — a situation fixed in Go 1.9 by adding a monotonic clock source, but fixed specifically for the scenario where a timestamp repeats, not for the scenario where a timestamp is skipped.

A negative leap second produces a different class of failure. Distributed databases that rely on timestamp ordering to determine transaction sequence could find a record timestamped 23:59:59 arriving after a record timestamped 00:00:00, even though the first transaction was initiated first. Scheduled task queues that check "how long until my next deadline" could find their deadlines appeared to have already passed. Any system that uses elapsed-time calculations to rate-limit operations, generate sequence numbers, or verify audit logs built on the assumption that time only moves in one direction — and only stops momentarily, never jumps.

"A negative leap second has never been added or tested, so the problems it could create are without precedent," Patrizia Tavella, director of the time department at the International Bureau of Weights and Measures (BIPM), has warned. According to BIPM's 2022 Resolution 4, the international timekeeping community confirmed that the negative leap second's insertion "has never been foreseen or tested" — a finding that has since been echoed across the metrology and infrastructure engineering communities.

Meta engineers writing in 2022 flagged the specific risk to code that handles elapsed time: "Depending on how [elapsed time] is used, we may end up in a situation relying on a negative value during a leap second event." The comparison to Y2K has been raised by multiple researchers — with the critical caveat that Y2K had decades of warning and extensive testing, while a negative leap second has had neither.

How Three Different Clocks Are Already Running on Different Time

The engineering challenge is compounded by a fragmentation problem that has quietly grown worse over the past decade: there is no longer a single authoritative time that all systems share.

UTC, distributed via the Network Time Protocol (NTP), is what most internet-connected servers receive. GPS time is something different: GPS satellites run on atomic time but do not incorporate leap seconds, instead broadcasting a UTC-GPS offset that receivers can use to convert. As of 2026, GPS time leads UTC by exactly 18 seconds. A GPS-synchronized system and an NTP-synchronized system claiming to display UTC may handle a leap second event in fundamentally different ways.

Meanwhile, the largest cloud providers have implemented their own divergent workarounds called "leap second smearing." Google has spread each leap second across a 24-hour window centered on the leap event — running clocks slightly slower than usual during the smear period — since 2008, though the smear window's shape and duration have evolved over that time. Amazon, Microsoft, and Meta have used similar but non-identical approaches. These smear windows are unstandardized: different providers start at different times and use different algorithms. During a smear window, a server using one provider's approach and a server using another's may be measurably apart in time — sufficient to corrupt timestamp ordering in distributed systems that span cloud providers.

This fragmentation means that even if a negative leap second were handled perfectly by one provider's infrastructure, systems communicating across provider boundaries during that window would face the same ordering ambiguity that causes failures in simpler scenarios. The negative leap second does not arrive into a unified, well-defined timekeeping ecosystem. It arrives into one that is already running on at least four loosely synchronized time standards simultaneously.

What Does 30% Probability Actually Mean for Infrastructure?

Agnew has calculated roughly a 40 percent probability that a negative leap second will be required between now and 2035. Tavella put the BIPM's figure at 30 percent under the same condition. Both figures are contingent on Earth's rotation not decelerating again before the 0.9-second threshold is breached — a possibility that some researchers consider likely. Leonid Zotov of Moscow State University has predicted the spin may slow again before a negative leap second becomes necessary.

But infrastructure engineers cannot plan around the optimistic scenario. Tavella described the reaction when her stakeholders were asked whether a 30 percent risk was acceptable: "They said, 'No, even 10 percent risk is too much.'"

The documented cost of even a routine positive leap second provides useful context. According to published research on the economics of leap second events, the U.S. military reportedly places its nuclear command-and-control network into a special precautionary mode for one hour before and one hour after each scheduled leap second — a procedure cited in that research at a "two-digit-million-dollar" cost per event. The Intercontinental Exchange, parent company to the New York Stock Exchange and 11 other exchanges, chose to shut down entirely for 61 minutes during the 2015 leap second rather than risk timestamp errors in trading systems.

A negative leap second, applied to systems that have been patched for positive events but never tested for negative ones, represents a materially larger and less-bounded risk than any of those precedents.

Has Any System Been Built to Handle This?

The short answer is that some systems have been patched to not fail catastrophically — and almost none have been tested at scale.

Go's standard library added a monotonic clock source in version 1.9 (August 2017), which prevents elapsed-time calculations from returning negative numbers during a positive leap second. That fix addresses one class of failure. It does not prevent the class of failures caused by a timestamp being skipped rather than repeated.

Google's NTP smear eliminates the need for its servers to handle the discrete jump at all, by spreading the change over time. But this only works if all systems a given application communicates with use a compatible smear — and as noted above, they do not. A microservice that sends a timestamped request during one provider's smear window may be talking to a downstream service that has not smeared and is at a slightly different moment in time, producing a timestamp inversion the upstream system was not designed to handle.

Meta engineers were direct in their 2022 public statement: "The impact of a negative leap second has never been tested on a large scale; it could have a devastating effect on the software relying on timers or schedulers."

October Vote Could Make Negative Leap Seconds Obsolete

This is the context in which October's CGPM meeting takes on an importance that its dry procedural framing does not convey. The 28th General Conference on Weights and Measures, meeting in Paris, is expected to vote on a proposal to replace the leap second with a leap hour — an adjustment that would allow UTC to drift from UT1 by up to a full hour before any intervention was required.

At the rate Earth's rotation currently varies, a one-hour threshold would not be reached for somewhere between 50 and 100 years, effectively making both positive and negative leap seconds a concern for the distant future. The proposal would target implementation as early as 2027. When it goes into effect, the last-minute-of-the-day adjustment mechanism would be retired, and any remaining gap between UTC and UT1 would accumulate silently until a leap-hour correction — likely manageable as a smear rather than a discrete jump — becomes necessary generations from now.

The 2022 CGPM Resolution 4 had already established the mandate: increase the maximum permissible divergence between UTC and UT1 by 2035 at the latest. What the October 2026 meeting represents is the opportunity to accelerate that timeline before Earth's rotation makes the interim period dangerous. The same principle that led Russia to successfully lobby for a 2035 deadline rather than an earlier one — to give its GLONASS satellite navigation system, which unlike GPS does incorporate leap seconds, time to adapt — may resurface as GLONASS engineers push back on a 2027 implementation. Whether the 64 member states can agree on a timeline fast enough to matter is the central editorial question of the October meeting.

The irony embedded in this governance sequence is sharp: the original impetus for abolishing the leap second was tech-company frustration with positive leap seconds, which were well understood and patchable. The mechanism that is finally forcing the international community to act is the negative leap second — a scenario that is neither well understood nor patchable — arriving from a geophysical cause nobody had predicted when the 2022 resolution was drafted.

What This Means for Your Clock Tonight

For most people, none of this is immediately visible. Smartphones synchronized via NTP receive time from servers that absorb leap seconds — or try to. The clock on your wall will continue ticking normally through midnight on December 31, 2026, as Bulletin C 72 has now confirmed.

For engineers managing distributed systems, databases, financial applications, or any infrastructure that assumes timestamps are monotonically increasing: this is the period to audit. Specifically, it is worth confirming which time sources systems rely on (NTP, GPS, or a cloud provider's smeared clock); whether elapsed-time calculations depend on wall clock monotonicity; and whether scheduled task systems have failure modes triggered by a missed timestamp. Most developers are unaware that leap seconds exist. The engineering literature on the negative case is sparse by necessity — the event has never occurred.

The next IERS Bulletin C, covering a possible leap second at the end of June 2027, is expected in January 2027. The window between now and October's CGPM meeting is the best available advance notice the timekeeping community can provide. Whether it is long enough to matter depends on what the 64 member states decide in Paris — and on how fast the planet keeps spinning.

Frequently Asked Questions

What is a negative leap second, and has one ever been inserted?

A negative leap second would remove one second from Coordinated Universal Time (UTC) rather than adding one. In practice, this means the clock would jump from 23:59:57 directly to 00:00:00, with 23:59:58 and 23:59:59 simply not existing. No negative leap second has ever been inserted in the 54-year history of the leap second system. All 27 adjustments since 1972 have been positive — adding a second, not removing one.

Why is Earth spinning faster, and will it keep accelerating?

The leading explanation is a slowdown in Earth's liquid outer core, which by conservation of angular momentum causes the outer layers to spin faster — similar to a skater pulling in their arms. Seasonal atmospheric shifts and the redistribution of meltwater from polar ice also contribute. Climate change has paradoxically provided a slight brake: meltwater spreading toward the equator slows the surface spin. Whether Earth's rotation will decelerate again before a negative leap second becomes necessary is genuinely uncertain; some researchers predict it will, others argue the core-driven trend could continue for decades. Agnew's 2024 Nature analysis provides the most detailed current projection.

Why would a negative leap second be more dangerous than the positive kind, which has already caused outages?

Positive leap seconds pause time for one tick; most modern software patches handle this by either repeating the 23:59:59 timestamp or applying a smear. A negative leap second does the opposite: it skips time forward, eliminating an entire second. Software that holds a target timestamp — waiting for 23:59:59 to trigger a scheduled task, for example — may find that second no longer exists and respond unpredictably. Transaction ordering in distributed databases depends on timestamps; skipping a second can make a later transaction appear earlier than an earlier one. No production system has ever encountered this scenario, so unlike positive leap second failures — which were diagnosed and patched after the fact — there is no patch history to draw on. Meta's engineering team documented the risk in a widely cited 2022 statement.

What would the proposed "leap hour" change, and when would it take effect?

If the October 2026 CGPM meeting approves the leap-hour proposal, the leap second mechanism would be retired, likely as early as 2027. UTC would be allowed to drift from UT1 by up to one full hour before any correction was required. At Earth's current rate of rotation variation, that threshold would take 50 to 100 years to reach — effectively removing the leap second problem from the engineering agenda for the foreseeable future. When a leap-hour correction eventually became necessary, it would be applied as a long, gradual smear rather than a discrete one-second jump, avoiding the failure modes that have caused outages since 1972. The one remaining complication is Russia's GLONASS satellite system, which incorporates leap seconds in its timekeeping — unlike the U.S. GPS system — and would require adaptation.