The Ceiling: What Science Now Says About the Maximum Human Lifespan

A number, and the number nobody printed

In June 2026, a group at the Skolkovo Institute of Science and Technology published a paper in npj Aging with an unusually disciplined ambition. Rather than asking how long people do live, Evgeniy Efimov, Dmitrii Kriukov and their colleagues built a survival model they could switch on one mechanism at a time, and then asked a hypothetical worth taking seriously: suppose every reversible hallmark of aging were solved. Suppose the senescent cells were cleared, the epigenome restored, the mitochondria renewed. One thing would remain, because it is not a process but an accumulation — the random mutations that gather in the DNA of your cells simply as a consequence of existing. How long would you then live?^[1]

Their answer, integrated across organs, was a median of 146 to 194 years. That is roughly twice the current human span, and it travelled the world in a fortnight as good news.

The more interesting figure did not travel at all. In the same model, a body with no mutation accumulation whatsoever — the theoretical non-aging baseline — reaches a median of 1,759 years.^[1] Introduce mutations into post-mitotic tissue and that collapses to 156 years. One mechanism, acting alone, removes something on the order of 91% of the theoretical span.

Read in that direction, the headline inverts. The paper is not a promise of 194 years. It is a measurement of how much a single, currently unfixable process costs us — and, by subtraction, an argument that the remaining distance from 156 down to the eighty-odd years most of us actually get must be paid for by everything else. The authors say so plainly: mutations drive aging significantly but cannot alone account for observed mortality, which implies comparable contributions from the other hallmarks.^[1] There is no single lever here. There is a queue of them.

Why the heart and the brain set the ceiling

The most consequential finding in the paper is not a number at all but an asymmetry. Not all tissue ages on the same terms.

Your liver is a fundamentally different proposition from your brain. Hepatocytes divide; damaged ones are replaced; the lineage is continually refreshed. In the model, proliferating tissue of this kind maintained function for thousands of years, because cellular replacement effectively neutralises mutation-driven decline.^[1] Damage in a renewing tissue is a flow problem, and flow problems can be managed.

Neurons and cardiomyocytes are not renewing tissue in any meaningful sense. The neuron firing in your prefrontal cortex as you read this sentence is, in all probability, the same cell that was there when you were three. It will not be replaced. Every mutation it acquires it keeps, permanently, and the errors compound in a cell that has no exit strategy. Damage in a post-mitotic tissue is not a flow problem. It is a ledger, and the ledger only runs one way.

This is why the paper's title speaks of an entropic bound. The limit is not a designed expiry or a programmed clock. It is the statistical certainty that a cell which cannot be replaced will, given enough time, accumulate enough disorder to fail.

The clock runs at a species-specific speed

What makes the modelling result credible rather than merely elegant is that the empirical biology had already pointed the same way.

In 2022, researchers at the Wellcome Sanger Institute sequenced 208 intestinal crypts from 56 individuals across 16 mammalian species — mice, humans, giraffes, naked mole-rats, tigers.^[2] The species differed in lifespan by roughly 30-fold and in body mass by around 40,000-fold. Somatic mutation rate per year varied enormously between them, and it varied in a specific way: inversely with lifespan. Short-lived animals accumulate mutations quickly; long-lived animals accumulate them slowly.

Then comes the detail that should give any longevity investor pause. Despite that enormous variation in rate, the mutation burden each species had accumulated by the end of its natural life varied by only about threefold.^[2] A mouse and a human arrive at death carrying broadly comparable amounts of genomic damage. They simply take very different amounts of time to get there.

That is what a real biological constraint looks like: not a number that varies with circumstance, but one that holds across four orders of magnitude of body size. It is also the strongest available evidence that the 2026 model is measuring something structural rather than an artefact of its own assumptions.

The argument that has never been settled

None of this occurs in a vacuum, and honesty requires acknowledging that demographers have been fighting over the ceiling for a decade without resolution.

In 2016, Xiao Dong, Brandon Milholland and Jan Vijg published an analysis in Nature arguing that the limit is already visible in the data. Improvements in survival, they showed, tend to decline after age 100, and the age at death of the world's oldest person has not increased since the 1990s — despite every advance in medicine since.^[3] Their conclusion was that maximum human lifespan is fixed and subject to natural constraint, with a practical ceiling near 115 years. Jeanne Calment, who died in 1997 at 122, remains the outlier nobody has approached in nearly three decades.

Two years later the counter-argument arrived, and it was formidable. Elisabetta Barbi and colleagues, including the demographer James Vaupel, examined every Italian resident aged 105 and over between 2009 and 2015 — 3,836 documented cases, a dataset clean enough to avoid the age-misreporting problems that had compromised earlier work. They found that mortality hazard curves level off beyond 105.^[4] Past that age, your risk of dying in the coming year stops climbing. It is roughly the same at 110 as at 106.

A mortality plateau is not compatible with a wall. If the annual hazard stops rising, then extreme ages are not blocked by biology so much as made improbable by arithmetic — you must simply win the same coin toss many times in succession. The 2026 model sits, quietly, on this side of the argument. A 156-year median is an entropic bound, not a barrier.

Where the longevity market is looking, and where the ceiling actually is

Here is the observation we would draw from this literature, and it is not one the coverage made.

Sort the longevity industry by where it spends its attention, and the pattern is striking. Skin. Gut. Metabolic markers. Hair. Immune function. Muscle. These are, almost without exception, tissues that renew themselves — precisely the tissues the model found could remain functional for thousands of years, because replacement handles the damage. They are also, not coincidentally, the tissues whose improvement is visible in a mirror or a blood panel within a quarter, which is why they attract money.

The ceiling, meanwhile, sits in the two organs almost nothing on that list touches. This does not make the visible work worthless; healthspan is not a consolation prize, and there is little dignity in reaching 120 with a mind that left decades earlier. But it does clarify what is being bought. A regimen that optimises renewable tissue is buying quality of years. It is not, on this evidence, buying the ceiling.

It also explains why the interventions with the most serious longevity data behind them tend to be the unglamorous, systemic ones — rapamycin acting on nutrient sensing, senolytics clearing cells that will not clear themselves, epigenetic reprogramming attempting to restore cellular identity rather than replace cells. Notably, of everything currently in the field, partial reprogramming is the only approach that even proposes to address damage inside a cell that will never divide. That is either a reason to watch it closely or a reason for humility about how far the rest of the field can go.

The honesty clause

This is a mathematical model. It is not an observation, and it should not be worn as one.

Its central figure is conditional on a premise that has never been met by anyone: that every hallmark of aging apart from somatic mutation has been eliminated. We cannot currently eliminate a single one of them completely. The 146–194 year range is therefore a statement about the structure of aging, not a forecast about people now living, and anybody citing it as a life expectancy has misread the paper.

The range itself — 146 to 194, with 156 for post-mitotic tissue in isolation — is wide, which is a candid signal of how much depends on parameter choices. Press coverage tended to select whichever endpoint suited its headline, which is how the same study appeared in one outlet as 156 years and another as 194. And the model necessarily assumes we know how mutation burden translates into cell death and organ failure; that relationship is estimated, not measured directly in a living human brain.

Set against all of it stands one stubborn fact: no verified human has lived past 122, and nobody has come close in twenty-nine years. Theory says the ceiling is high. Observation says we are nowhere near testing it, and that the interesting question for anyone alive today is not the ceiling but the distance between current lifespan and the tissue quality we bring to it.

The royal verdict

The most valuable thing in this paper is not the headline number. It is the map.

What the model provides is a ranking of where the real constraints lie, and it is not where the market has been looking. Renewable tissue is a managed problem. Post-mitotic tissue — the heart, the brain, the two organs on which a life most obviously depends — is the unmanaged one, and no amount of attention to the visible will change that arithmetic.

There is something faintly aristocratic in the finding, in the older sense of the word: the things that matter most are the things that cannot be replaced. A liver can be renewed cell by cell for a thousand years. The neuron that holds a memory of a particular afternoon has exactly one chance. Whatever one spends on longevity, it is worth knowing which of those two accounts one is actually paying into — and worth noting that the interventions most likely to move the second are also the ones the field understands least. That, rather than 194 years, is the finding worth carrying away.

Common questions