Negative Leap Second 2026: System Impact &

Are Systems Ready for First Negative Leap Second in 2026?

On July 8, 2026, Scientific American reported that international timekeeping authorities are preparing to vote on a proposal to replace the leap second with a leap hour. The target for implementation is as early as 2027. The same article explains that Earth’s rotation has been speeding up since 2016 and that a negative leap second, the first in history, could be needed before 2035. Patrizia Tavella, director of the Time Department at the International Bureau of Weights and Measures (BIPM), told Scientific American that BIPM estimates a 30 percent risk of a negative leap second occurring if no action is taken by 2035. Tavella also told Scientific American that stakeholders told BIPM even a 10 percent risk is too much.

This is not an abstract timekeeping debate. A negative leap second would be the first time the world’s civil time standard moves backward. Every system that relies on monotonic clock behavior, continuous timestamps, or strict UTC alignment faces a failure mode that has never been tested at scale.

Why This Matters Now in 2026

The leap second was introduced in 1972 by BIPM to keep Coordinated Universal Time aligned with Earth’s gradually slowing rotation. Every time UTC drifted one second ahead of astronomical time, timekeepers added a leap second. All have been positive leap seconds, adding one second to the UTC day.

Then Earth started spinning faster. A 2024 study published in Nature found that the acceleration is significant enough that timekeepers now face a scenario with no historical precedent: subtracting a second instead of adding one. The Nature paper projects that a negative leap second may be required as early as 2029, driven by changes in Earth’s core angular velocity, with melting of polar ice sheets delaying the event by about three years compared to what would otherwise occur.

The timing compounds the risk. Digital networks now demand processing speeds measured in microseconds, not milliseconds. High-frequency trading platforms execute orders in nanoseconds. Distributed databases coordinate writes across continents. Cloud services manage certificate validation, session expiry, and replication lag with sub-second precision. A single backward jump propagates through every layer of this stack.

What Breaks When Time Goes Backward

A negative leap second means that at a designated moment, UTC skips one second. For an operating system kernel that assumes time advances monotonically, this creates an instantaneous reversal. The kernel’s clock functions, which power log timestamps, TLS certificate validation, and database transaction ordering, return a value that is smaller than the value returned one second prior.

This breaks the monotonic clock contract that most software implicitly trusts. When that contract is violated, consequences cascade through multiple system layers.

Network Time Protocol (NTP)

NTP is the backbone of internet time synchronization. It handles positive leap seconds through a well-documented mechanism: the kernel inserts an extra second at 23:59:60, and the clock stalls for one second before advancing. This code path has been tested across more than two dozen real events over five decades, as documented in the leap second record. The negative case is different. NTP implementations must handle a clock that jumps forward by one full second to skip the missing tick. No standard code path exists for this in most NTP distributions. The protocol’s reference implementation has never been validated against a live negative leap second announcement from the International Earth Rotation and Reference Systems Service (IERS).

Operating System Kernels

Linux, Windows, and BSD kernels all implement monotonic clocks specifically to protect applications from time reversals. The Linux kernel’s CLOCK_MONOTONIC is guaranteed to never go backward, even during a leap second. But the real-time clock (CLOCK_REALTIME) does reflect UTC. When UTC jumps backward, every application using real-time clock values for ordering decisions encounters a problem. A database that timestamps a write at 23:59:58.500 and then sees a subsequent timestamp of 23:59:58.000 may reject the second write as already applied, or worse, apply it out of order.

Distributed Consensus and Databases

Raft, Paxos, and other consensus protocols use timestamps and logical clocks to establish event ordering. A backward time jump can cause leader elections to fail, log replication to stall, and snapshot isolation to break. Databases that rely on wall-clock time for MVCC (multi-version concurrency control) may see phantom reads or lost updates during the reversal window. The exact behavior depends on the implementation, but no major database vendor has published a tested mitigation for the negative leap second scenario.

Historical Lessons From Positive Leap Seconds

The track record of positive leap seconds is not reassuring. Even with decades of advance notice and well-documented handling procedures, each insertion has caused measurable disruptions.

All of these incidents occurred with positive leap seconds, where time moves forward by one extra tick. The negative case introduces a fundamentally different failure mode: time reversal. Google’s leap smear technique, which gradually adjusts time over a window to avoid instantaneous jumps, is designed for positive insertions. A negative smear would require the clock to slow down gradually, which is the inverse of the current implementation and has not been deployed or tested in production.

The Leap Hour Proposal and What It Means

The BIPM’s 2026 General Conference on Weights and Measures (CGPM) is considering a radical change: replacing the leap second with a leap hour. Under this proposal, timekeepers would allow UTC to drift up to one full hour away from astronomical time before making a single one-hour correction. According to Scientific American, some experts estimate that a leap hour would not need to be added for hundreds of years, effectively abolishing the leap second system.

The proposal emerged from direct consultation with stakeholders. When BIPM asked users, cloud providers, financial exchanges, and telecommunications operators whether they could accept a 30 percent risk of a negative leap second, the answer was unanimous: even 10 percent was too high. Tavella told Scientific American that BIPM estimated the risk at 30 percent if the system waits until 2035, and that stakeholders said even 10 percent risk is too much. The leap hour proposal aims to eliminate that risk entirely by removing the need for frequent, irregular time corrections.

If CGPM votes to adopt the leap hour, the target implementation date is 2027. That timeline is aggressive by international standards, but the urgency reflects a growing consensus that the current system is unsustainable. Tavella told Scientific American that there is “sufficient urgency to put leap hour into effect as soon as possible.”

The 2026 AI infrastructure buildout adds another layer of complexity. Massive GPU clusters running training and inference workloads depend on precise time synchronization for checkpointing, distributed data loading, and job scheduling. A backward time jump in a multi-epoch training run could corrupt gradient accumulation or cause checkpoint timestamps to overlap, forcing a full restart.

What Operators Should Do Now

System readiness for the first negative leap second varies widely across the industry. Here is a practical checklist for engineering teams managing production infrastructure in 2026.

Audit Your Time Synchronization Stack

Document every NTP source, stratum level, and fallback configuration in your infrastructure. Identify which systems use CLOCK_REALTIME versus CLOCK_MONOTONIC for critical operations. Any system that uses wall-clock time for ordering, expiry, or sequencing needs attention.

Test Monotonic Clock Assumptions

Run integration tests that simulate a backward time jump of one second. Observe whether your application crashes, produces errors, or silently corrupts data. Most testing frameworks do not include this scenario. Add it.

Evaluate Smear Implementations

If your organization uses Google-style leap smear for positive leap seconds, verify whether the implementation also handles negative adjustments. Most smear implementations are one-directional. Plan for a gradual slowdown approach if a negative leap second becomes necessary before the leap hour transition.

Monitor IERS Announcements

The IERS announces leap seconds with several months of lead time. If a negative leap second is scheduled before the leap hour takes effect, that announcement window will be the only opportunity to deploy mitigations. Ensure your incident response team has a runbook for time synchronization events.

Engage With Industry Groups

The BIPM, NIST, and other timekeeping bodies are actively seeking feedback from technology stakeholders. The leap hour proposal was shaped by user input. Continued engagement ensures that the engineering community’s concerns are represented in the final decision.

Key Takeaways

• The BIPM estimates a 30 percent risk of a negative leap second occurring before 2035 if no action is taken, and stakeholders say even a 10 percent risk is too much.
• A negative leap second introduces a fundamentally new failure mode: time reversal. Positive leap seconds have already caused outages at Meta, Cloudflare, Reddit, and Qantas.
• The BIPM’s 2026 CGPM is voting on a leap hour proposal that would replace leap seconds entirely, with a target implementation date of 2027.
• NTP implementations, OS kernels, and distributed databases lack tested code paths for negative leap second handling.
• Engineering teams should audit time synchronization stacks, test for backward time jumps, and monitor IERS announcements now, not when the event is imminent.

The question posed in the Ask HN thread is whether systems are ready for the first negative leap second. The honest answer is that most are not. The leap hour proposal represents a structural fix that would eliminate the problem, but it is not yet approved, and the transition timeline is aggressive. Until the leap hour is implemented, the risk of a negative leap second event remains real, and the systems that underpin modern digital infrastructure have never faced this particular failure mode. The time to prepare is before IERS issues the announcement, not after.

Related Reading

More in-depth coverage from this blog on closely related topics:

• QuadRF: Open-Source Radio Visualizes
• The $725B 2026 AI Infrastructure Buildout
• Parliament Greenlights Ransomware Control
• GLM-5.2 and AI Innovation in 2026
• GLM 5.2 Approaches Human Accuracy

Sources and References

Sources cited while researching and writing this article:

• International timekeepers to vote on changing the leap second to a leap hour | Scientific American
• leap second record documents
• A 2024 study published in Nature
• Meta Engineering
• Cloudflare Blog
• Brookings
• arXiv
• leap smear technique