Since 2019, the kilogram has had an exact definition based on a constant of nature. In March 2026, laboratories nevertheless began using an updated international reference for putting that definition into practice.
That can sound contradictory. If a kilogram is exact, what is left to adjust?
The answer is the difference between defining a unit and measuring a physical object in that unit. I think the distinction is worth a little attention, partly because the less tidy version of the story is more impressive.
The object stopped being the definition
Before 20 May 2019, the kilogram was defined by the mass of the International Prototype of the Kilogram, a platinum-iridium cylinder. Other standards acquired their values through comparisons with it. The BIPM, the international organization responsible for coordinating measurement standards, describes that earlier system in its guide to the kilogram’s practical implementation.
The cylinder was not merely a very accurate example of a kilogram. Its mass defined the unit. There was no separate, more fundamental kilogram against which to check it.
The present definition instead fixes the numerical value of the Planck constant at exactly 6.62607015 × 10⁻³⁴ joule seconds. Together with the definitions of the metre and second, that establishes the kilogram.
This did not make physical standards unnecessary. It changed their status. A particular piece of metal could now be measured against a definition that did not depend on that piece of metal remaining unchanged.
A definition still needs an experiment
A Kibble balance is one way to make the connection. In one measurement, an electric current in a coil produces a force that balances the weight of a mass. In another, moving the coil through a magnetic field produces a voltage. Combining the measurements removes a shared factor involving the coil and magnetic field.
NIST’s explanation of the instrument shows why this is useful: electrical measurements can be tied to quantum effects involving the Planck constant. The experiment also needs measurements of motion and local gravity. An exact constant does not make those measurements exact.
I like the reversal involved. Before the redefinition, a known mass could help determine the Planck constant. With that constant’s value fixed, the experiment can determine a mass. Much of the apparatus remains recognizable; which quantity is known has changed.
Five micrograms, with a date attached
The practical complication is that independent experiments do not automatically agree closely enough. The BIPM’s July 2025 guide explains why laboratories use a shared “Consensus Value”: it combines experimental results to maintain consistency while the individual methods continue to improve.
The next update is no longer a forecast. The BIPM’s announcement of the third Consensus Value sets its implementation date at 1 March 2026. For a one-kilogram standard, the change is a reduction of five micrograms in the assigned mass relative to the previous consensus reference. The cumulative adjustment relative to the old prototype-based reference is twelve micrograms. The stated standard uncertainty remains twenty micrograms.
These are changes in the values assigned through calibration, not instructions to remove metal from everyone’s weights. They are not a new definition of the kilogram, either.
Five micrograms per kilogram is five parts in a billion. That arithmetic helps put the adjustment in proportion without making it unimportant to the laboratories responsible for such small differences.
The useful part is that the difference can be checked
My first instinct was to describe the 2019 change as getting rid of dependence on an object. That is true about the definition, but incomplete as an account of measurement. There are still instruments, physical standards, comparisons, and decisions about how to keep results consistent.
What changed is that none of those objects has to be correct simply because it is the defining object. Experiments can disagree, the disagreement can be measured, and the practical reference can be revised without rewriting the unit.
To me, that is the achievement worth noticing. The definition is exact. The work of finding out how closely an instrument realizes it remains experimental, specific, and open to correction.