Leap Seconds Are Dead. Here Is What Breaks Next

The vote that ends an era, and quietly reshapes the infrastructure of the modern world

The International Committee for Weights and Measures has killed the leap second. Starting May 20, 2027, the world will no longer add or subtract a single second to keep atomic time aligned with Earth’s rotation. Instead, it will absorb the growing drift by adjusting the clock by one full hour every 3,600 seconds of accumulated difference — a leap hour, roughly once every few thousand years.

On the surface, this is a technical decision about a pedantic adjustment to Coordinated Universal Time. In practice, it is the most consequential change to global timekeeping since UTC was established in 1972. And the fallout will not be uniform.

Who wins and who loses

The winners are obvious: every system that parses timestamps. Aviation scheduling platforms, nuclear grid controllers, database engineers, high-frequency trading firms, and the engineers who maintain the NTP infrastructure that keeps the internet from drifting apart. All of them have spent decades living with the anxiety of an unpredictable second insertion that arrives without warning and can corrupt state transitions, log sequences, and transaction boundaries.

The losers are less visible but real. Russia benefits from this change less than anyone expected, despite being the country most invested in maintaining leap-second support in its satellite infrastructure. Its GLONASS navigation system is one of the only major global positioning networks that still explicitly processes leap seconds in its signal. The United States’ GPS and Europe’s Galileo have both moved past that requirement. Under the new regime, GLONASS will need an additional software update to remain interoperable with the rest of the satellite navigation ecosystem — a cost Russia may resist politically but will eventually have to absorb.

Then there is the astronomical community. The original rationale for leap seconds was preserving a tight link between civil time and solar time, so that noon would remain roughly when the sun is highest. That linkage is now severed in any meaningful sense. Within a century, apparent solar noon will drift several minutes away from UTC noon. Within a millennium, the gap becomes unignorable for anyone who cares about the old relationship. That trade-off was accepted deliberately and with open eyes.

The real disruption is not the year 2027. It is the year 2026.

The decision was announced on October 15, 2026, and takes effect roughly one and a half years later. That compressed timeline is the problem most observers are missing.

Every major platform on earth — every trading engine, every telecom network, every satellite operations center — has been running code that assumes leap seconds may arrive at any moment. The new rule changes that assumption, but the migration window is extraordinarily narrow. Patching timestamp logic across financial exchange architectures, telco radio access networks, cloud provider schedulers, and embedded systems in automotive and aviation is not a weekend task. It is a multi-phase rollout that typically takes quarters, sometimes years.

A significant portion of the world’s critical infrastructure will not finish migrating before the new rule activates. That means the first real leap-hour transition — whenever it occurs, possibly centuries from now — will happen in an environment where legacy code paths still assume leap-second behavior. The gap between the law and the codebase is where incidents will emerge, not on the day itself.

Finance is the canary

High-frequency trading firms already treat sub-microsecond timestamping as a competitive moat. Their clocks are disciplined by GPS and dedicated atomic references, and they build logic that explicitly accounts for leap-second anomalies: duplicate timestamps, gaps, and out-of-order events. Removing leap seconds simplifies the model but does not eliminate the need for careful monotonic-clock design. Firms that coded defensively will benefit. Firms that assumed leap-second handling was a solved problem may find that their defensive path was actually the correct one, and that the removal forces a rewrite of assumptions built into settlement engines, order books, and audit logs.

The same logic applies to banking core systems, clearing houses, and insurance ledgers. These platforms already handle leap-second edge cases at various levels of sophistication. The shift to leap hours means they can stop scheduling leap-second drills and redirect resources elsewhere. But the migration itself introduces risk during the transition period, and the transition period is short.

Telecom and 5G networks

The telecommunications industry has been pushing for this change for years. Precise time synchronization is now a hard requirement for 5G and upcoming 6G deployments. Time Division Duplex cells, coordinated multi-site scheduling, and edge-computing workflows all depend on microsecond-level alignment provided by protocols such as IEEE 1588 PTP and, increasingly, 5G’s own time-sensitive networking extensions.

Leap seconds introduce jitter and occasional clock restarts that break these alignments. Network operators have dealt with it by isolating leap-second adjustments to non-critical paths or applying them during maintenance windows. The new standard removes that burden entirely. Networks can lock to a pure atomic timescale without anticipating irregular second insertions. That is a real engineering win, and it will improve reliability for timing-sensitive services.

But the win is conditional on operators completing their migrations. Network function virtualization means the affected code runs on commodity hardware across distributed clouds. Upgrading NTP stack behavior, internal scheduler logic, and timestamp formatting across thousands of microservices is not trivial. Any provider that skips the migration cycle will inherit the same breakage patterns that plagued early internet systems before proper time handling became routine.

The satellite navigation divergence

GLONASS is the outlier here. Unlike GPS and Galileo, which encode leap-second information in a way that receiver software handles transparently, GLONASS builds leap-second awareness into its navigation message structure. The Russian system has explicitly supported leap seconds since its inception. Under the new rule, GLONASS either adopts the same abstracted approach or diverges from the other constellations.

Divergence is unlikely to be intentional. It is more likely a consequence of development priorities and geopolitical friction. Russia’s veto or abstention, if it occurred, signals discomfort with moving away from a system that has been part of GLONASS identity. But compatibility with GPS and Galileo is non-negotiable for receivers that claim global coverage. Russia will almost certainly update GLONASS eventually, and the cost of that update will be borne by satellite operators and receiver manufacturers alike.

The negative leap-second scare was the accelerating force

The decision did not come purely from accumulated frustration. It came from a specific, tangible threat: a negative leap second. Since the mid-2010s, Earth’s rotation has been speeding up slightly. Scientists projected that a negative leap second — removing one second from the clock rather than adding one — could occur around 2029. That scenario terrified engineers more than positive leap seconds ever did.

A negative leap second introduces a duplicate 23:59:59 hour. Systems that do not handle it can produce double-counted timestamps, corrupt log sequences, and transaction duplicates. The Y2K panic of the late 1990s was about a data representation flaw; a negative leap second is about temporal overlap. Many experts have warned that handling it correctly requires assumptions that far fewer systems satisfy than the community publicly acknowledged. The possibility of that event arriving sooner rather than later gave the reform movement the urgency it needed.

What happens next

The resolution passed with a majority, though not unanimity. The exact vote counts have not been released publicly. Sources present at the meeting indicate Russia likely voted against. France’s Noël Dimarcq, who formalized the proposal, called it a decisive step toward a stable international timescale that will persist for thousands of years. That characterization is accurate within its frame: the new system eliminates the recurring administrative overhead of leap-second decisions and the operational risk they carry.

But the stability it promises is deferred, not eliminated. The next adjustment will be an hour, not a second. That hour could arrive decades or centuries from now, depending on tidal friction and Earth’s rotational dynamics. When it does, the same class of migration problems will reappear at a larger scale. Engineers will need to patch timestamp systems for a 3,600-second jump rather than a one-second jump. The difference is one of degree, not kind.

For now, the immediate concern is the migration window. The world has approximately 18 months to move its critical timestamp infrastructure away from leap-second assumptions. Organizations that treat this as a scheduling item will complete the work. Those that treat it as background noise will inherit the simplest form of technical debt: the kind that reveals itself only when something breaks.

The leap second served its purpose. It kept civil time anchored to Earth’s rotation while atomic clocks ran free. That anchoring is now officially optional. The question is no longer whether Earth will continue to slow down — it will, gradually, under tidal friction — but whether future generations will care enough to pull the lever again. The new system suggests they will not, at least not for a very long time.