Global timekeepers vote to ditch leap second, replacing it with future 'leap hour'

Twenty-seven times in the past hundred years, our clocks have stopped for a single second. That pause—a leap second—was never meant to be permanent. It was a fix, a small adjustment to keep the machines that measure time in perfect sync with the planet's actual rotation. But this week, delegates gathering near Versailles will likely vote to end the practice altogether, replacing it with something that won't matter for another thousand years or more.

The problem began with a mismatch between two different ways of measuring time. For centuries, humans defined a second as one eighty-six-thousandth of a day—a straightforward division of Earth's rotation. But Earth doesn't rotate like clockwork. Our planet spins slightly faster or slower depending on atmospheric conditions, ocean currents, melting glaciers, and the slow gravitational tug of the Moon. Over the long term, Earth's rotation is slowing. Over shorter periods, it can speed up. Since around 2020, it has been accelerating.

In the late 1960s, the International Bureau of Weights and Measures adopted a far more precise definition of a second based on the vibrations of caesium atoms. This atomic second became the foundation of Coordinated Universal Time, or UTC, the global standard that underpins GPS, financial transactions, and the internet itself. The atomic second is stable in a way Earth's rotation simply is not. But that stability came at a cost: UTC and Earth's actual day began to drift apart. To keep them tethered, timekeepers introduced the leap second. When Earth's rotation fell behind atomic time, an extra second was inserted into the clock. The first leap second arrived on June 30, 1972. The most recent came at the end of 2016.

For decades, this seemed like a reasonable compromise. But the digital age exposed a fundamental problem: software was not built to handle time moving backwards. Leap seconds broke assumptions that programmers had baked into their code—that time only moves forward, that a minute always contains sixty seconds. When a leap second was inserted, systems crashed. Cloudflare went down. Reddit went down. The irregularity of leap seconds made the problem worse; they were scheduled only six months in advance, giving engineers little time to prepare. Unlike daylight saving time or time zones, which follow predictable patterns, leap seconds arrived without warning.

The urgency to abandon them intensified recently. Because Earth has been spinning faster since 2020, the next leap second might need to be negative—time would jump forward instead of backward. This has never happened before. No software has ever been tested for it. The risk of cascading failures across systems that depend on precise timekeeping seemed too high to ignore. In 2022, the BIPM's governing body decided to retire leap seconds by 2035. This week's vote will formalize that decision and approve the replacement system.

The leading proposal is elegant in its simplicity: let UTC drift from Earth's rotation. Allow the two to separate gradually, year after year, until they are an hour out of sync. Then, centuries or millennia from now, observe a single "leap hour" to bring them back into alignment. Given that Earth's rotation currently loses less than a second per year, the first leap hour won't be necessary for more than a thousand years. By then, the timekeeping systems we rely on today will almost certainly be obsolete, replaced by technologies we cannot yet imagine. One researcher at Australia's National Measurement Institute noted that in a hundred years, software will likely be so automated that the kinds of problems leap seconds create today simply won't exist.

There is a philosophical dimension to the change. Some argue that abandoning leap seconds severs our connection between the way we count time and our actual experience of it—the rotation of the Earth beneath our feet. But clocks themselves are artificial constructs, designed to serve human needs. The atomic second serves those needs far better than Earth's variable spin ever could. For the timekeeping authorities gathering this week, the choice is clear: solve the problem now by accepting a drift that won't matter for a thousand years, or keep patching a system that breaks every few years. The vote will almost certainly go one way.