If the universe is infinite, is there another you?

Somewhere, very far away, a person with your face, your name and your memories might be reading this sentence. Not a lookalike: a copy identical in every detail we could observe. Some cosmologists, scientists who study the universe as a whole, have argued that this follows from a few assumptions that sound simple [Reference 1,Reference 2].

Here is the catch, up front. The argument only works if several things are true, and none of them can be checked beyond the part of the universe we can see: that space really goes on forever, that the matter in it was scattered at random in a particular statistical way, and that any region of space can only be arranged in a finite number of distinguishable ways. Grant all of that and the copies follow from simple counting. Doubt any of it, and they may not exist at all.

The argument, in six assumptions

The reasoning is old. In 1979 George Ellis and G. B. Brundrit argued that an infinite, roughly uniform universe should contain regions with histories very similar to ours [Reference 3]. In 2001 Jaume Garriga and Alexander Vilenkin sharpened it into a claim about identical histories [Reference 2], and in 2003 Max Tegmark popularized it, calling each faraway region the size of our observable universe (the bubble of space around us from which signals have had time to reach us) a “parallel universe”, even though it is really just a distant part of our own space [Reference 1].

It rests on four big assumptions and two hidden ones. Each deserves a label.

1. Space is infinite, or at least unimaginably larger than the distances below. Status: unknown. Space is measured to be very nearly flat, which means the geometry you learned in school works on huge scales: parallel lines stay parallel. But a flat universe can still be finite, wrapped around like a video-game screen where leaving on the right brings you back on the left, and a curved one could be enormous but finite (Part 1) [Reference 6,Reference 7].
2. Matter is spread roughly evenly, with the same laws of physics everywhere. Status: well supported within the part we can see. The density of the early universe, seen in its afterglow (faint leftover light from the hot early universe that we still detect today), varied by only about one part in 100,000 from place to place [Reference 1,Reference 8]. Whether it stays that way far beyond is an extrapolation.
3. The starting conditions were random, in a specific sense. The tiny ripples that grew into galaxies are thought to have come from quantum jitters, the tiny random flickers quantum physics says happen everywhere, in the very early universe; Tegmark calls the result random “for all practical purposes” [Reference 8]. His version assumes the ripples form an ergodic random pattern: watching one coin flipped a million times tells you the same as flipping a million coins once. In space terms: looking at many far-apart regions of our one universe tells you the same as looking at many possible versions of our region. Tegmark’s summary is that “everything that could in principle have happened here did in fact happen somewhere else”; his own definition of this randomness assumes space is infinite, so it can’t be used to prove that it is [Reference 8]. Status: a theoretical assumption, consistent with what we see but untestable beyond our horizon, the distance beyond which light has not had time to reach us.
4. Any finite region with a limited amount of energy can be arranged in only finitely many distinguishable ways. Status: standard quantum physics for ordinary matter whose energy is limited. But once gravity is included, energy no longer sets a limit, and the finiteness has to be assumed; Garriga and Vilenkin do exactly that [Reference 2]. Untested in that regime (see below).
5. (Hidden) Distant regions develop independently of each other [Reference 3]. Status: plausible; untestable for regions beyond our horizon.
6. (Hidden) Probabilities about an imagined collection of possible universes apply to the one universe we have. “Probability 1” is a statement about that imagined collection. A forecast of “90 percent chance of rain” means something because there are many days like today; we get only one universe. Ellis points out that probability arguments are weakest “when there is only one case to consider” [Reference 9]. Status: a philosophical step, rarely stated.

If all of that holds, the conclusion is simple. An infinite space contains infinitely many regions the size of our observable universe. Each can turn out in only finitely many ways. So every possible way, including the exact way our region turned out, keeps happening, again and again, with probability 1 [Reference 2,Reference 3]. A copy of you does not need a copy of everything around you. Matching a person-sized region is far likelier than matching a whole observable universe, which is why Tegmark’s distance for “you” (10¹⁰²⁸ m) is so much smaller than his distance for a whole region (10¹⁰¹¹⁸ m) [Reference 1]. A copy whose whole life story matched yours, though, would need matching surroundings too.

Why the number of possibilities is finite

Think of a phone screen. It has a fixed number of pixels and colors, so it can only show a finite number of different images: a huge number, but finite. A deck of 52 cards can be shuffled into “only” about 8 × 10⁶⁷ orders. Shuffle infinitely many decks and every order turns up again and again.

In everyday physics, it seems a box of gas could be arranged in infinitely many ways: shift one atom by a billionth of a millimeter, and you have a new arrangement. Quantum mechanics says nature has a resolution limit, like pixels: below a certain level of detail, two arrangements of a region with limited energy stop being distinguishable, so there is only a finite number of genuinely different states [Reference 2,Reference 3]. Gravity complicates this, because gravitational energy can be negative, so energy no longer caps what is possible; Garriga and Vilenkin, who make the argument most carefully, end up simply assuming that the amount of information in a region cannot be arbitrarily large [Reference 2].

Physicists have precise versions of such limits. The Bekenstein bound caps the information that can fit inside a region of a given size and energy (mass counts as energy, through E = mc²) [Reference 10]. For a 70-kilogram person inside a sphere 2 meters across, it allows at most about 2 × 10⁴⁵ bits, or yes/no answers. Since N yes/no answers can describe 2^N different things, such a region can be in at most about 10¹⁰⁴⁵ distinguishable states: mind-bogglingly many, but finite. Garriga and Vilenkin estimated that the number of distinguishable histories (possible “movies” rather than “photos”) of a region the size of our observable universe, from when the universe first became transparent, about 380,000 years after the Big Bang [Reference 11], until today, is finite too [Reference 2].

How far away? Tegmark’s numbers, and why they disagree

Tegmark made the argument concrete with a back-of-the-envelope count [Reference 1]. Take a region the size of our observable universe, which he calls a Hubble volume and gives a radius of about 4 × 10²⁶ meters (about 42 billion light-years in his 2003 figures, an older, rounder estimate). Ask how many protons could be packed into it if it were no hotter than 100 million kelvin, a temperature scale starting at absolute zero. His answer: about 10¹¹⁸ [Reference 1]. In the longer version of the essay, Tegmark says the count rests on the Pauli exclusion principle, which forbids two protons from sharing a quantum state [Reference 8]. In our reading, that is why temperature matters: quantum mechanics gives each particle a “fuzziness” that shrinks as it gets hotter, so the fuzziness sets how small each slot can be, and capping the temperature caps the number of slots. Most slots are empty: the observable universe actually contains only about 10⁸⁰ protons [Reference 8]. What matters is how many places they could be.

Treat each slot as filled or empty. Then the number of possible arrangements is 2 multiplied by itself 10¹¹⁸ times, roughly 10¹⁰¹¹⁸ [Reference 1]. That needs a moment. 10¹¹⁸ is a 1 followed by 118 zeros. But 10¹⁰¹¹⁸ is a 1 followed by 10¹¹⁸ zeros: more zeros than there are protons in the observable universe, by a factor of about 10³⁸. In this notation, changing the 118 at the top to 119 would multiply the number of zeros by ten.

If every arrangement were equally likely and regions independent, you would have to look through about that many regions before expecting a repeat of ours. Pack them into a cube (a cube with 10 regions along each side holds 1,000): its side, in meters, is the cube root of that count times the size of one region. Strictly, 2 multiplied by itself 10¹¹⁸ times has about 3 × 10¹¹⁷ digits, not 10¹¹⁸. Taking a cube root divides that by three, to about 10¹¹⁷ digits, and multiplying by 10²⁷ meters adds only 27 more. So the cube is about 10¹⁰¹¹⁷ meters across, which Tegmark rounds to 10¹⁰¹¹⁸. Beyond that, Tegmark wrote, universes, “including ours”, must repeat [Reference 1]. Strictly, that distance only makes a copy of our region somewhat better than even (63 percent, like the two-in-three chance of rolling at least one six in six throws of a die); no finite distance guarantees one.

Tegmark then shrank the target to a sphere 100 light-years in radius, so that “all perceptions that we have during the next century will be identical”, since nothing from farther away can reach us within a century: about 10¹⁰⁹² meters. A copy of you: about 10¹⁰²⁸ meters [Reference 1].

How rough are these numbers? Rougher than they look. Changing the 28 at the top to 29 multiplies the number of zeros by ten, so these are guesses at the size of the size of the distance. Treat any precise-sounding figure you see quoted as a guess.

• Tegmark published two versions of this essay. The arXiv preprint (arXiv is a free website where scientists share papers), later a book chapter, gives 10¹⁰²⁹ m for you, 10¹⁰⁹¹ m for the 100-light-year sphere and 10¹⁰¹¹⁵ m for a whole Hubble volume, based on 10¹¹⁵ slots [Reference 8]. That is three units below the magazine’s 118.
• Neither version shows how the figure for you was obtained [Reference 1,Reference 8]. The same recipe applied to a sphere big enough to hold a person gives about 10¹⁰³⁸ m.
• Other ways of counting states give bigger numbers: a conjectured limit on information based on a region’s surface area (the holographic bound) allows up to about 10¹⁰¹²³ states for a Hubble volume, against Tegmark’s 10¹⁰¹¹⁸ arrangements.
• Rounding choices alone (counting in powers of 2 or of 10, the cube root, the region’s size) shift the top number by about one.

Tegmark himself called the estimates “extremely conservative”, since they simply count every possible quantum state, and added that your nearest copy is probably much closer, because planet formation and evolution “tip the odds in your favor” [Reference 1]. The honest message: even the top exponent is uncertain by many units, and the true distance is probably much smaller.

How big is that number?

Try to picture 10¹⁰²⁸. It has 10²⁸ + 1 digits. Write it out at one digit per millimeter, and the string of digits would stretch about a billion light-years (1.06 billion), about a hundredth of the way across the observable universe. Write one digit every tenth of a nanometer, and it still runs to about 106 light-years. Writing one digit per second would take about 23 billion times the current age of the universe, 13.8 billion years [Reference 11].

And that is just the number of digits in the distance. The distance itself, in meters, is 1 followed by all those zeros. For comparison, the whole observable universe is about 10²⁷ meters across. In this notation, our entire visible cosmos is 10^101.43 meters, and a copy of you is at 10¹⁰²⁸.

Now try your own version of the calculation. Choose what has to match, how finely it has to match, and see how far away the nearest copy would be expected.

Probability 1 is not certainty

Suppose each region has some tiny chance of being an exact copy of ours. Search more regions and the chance of finding at least one copy rises. If every pattern is equally likely, searching as many regions as there are possible patterns gives a chance of about 63 percent. Search ten times as many, and it is 99.995 percent. Only in the limit of infinitely many regions does it reach 1.

Joshua Knobe, Ken Olum and Alexander Vilenkin put the conclusion carefully: in such a universe, every possible history occurs in infinitely many regions “with probability 1” [Reference 3]. They add a crucial caveat: it is still possible for an infinite universe to contain only finitely many copies; that possibility simply has probability zero [Reference 3]. Mathematicians call this kind of statement almost sure. Think of flipping a coin forever: getting tails every single time has probability zero, yet nothing in the rules forbids it. As Ellis puts it, probability arguments can show that something is very improbable, but that “does not prove it is impossible” [Reference 9].

There is also the sixth assumption hiding here. “Probability 1” is computed over an imagined collection of possible universes, each with its own random starting conditions. We have exactly one universe. Saying that what is almost certain in the collection is true of ours is a further step, and Ellis argues that probability arguments carry least weight precisely when there is only one case to examine [Reference 9]. Knobe, Olum and Vilenkin, for their part, argue that an infinite universe supplies its own collection: it contains infinitely many systems just like any given one [Reference 3].

Counting infinities is harder than it looks

A separate problem appears when we try to test theories by their predictions, such as “what should a typical copy of me expect to see next?”. Those are fractions: what share of all observers see something. With infinitely many observers, fractions become slippery. What fraction of whole numbers is even? Counting 1, 2, 3, 4, …, the answer is one half. But list every whole number in a different order, two odd numbers then one even (1, 3, 2, 5, 7, 4, 9, 11, 6, …), and only a third are even, even though every number still appears exactly once. Tegmark makes the same point with a different ordering: the answer depends on how you count [Reference 8].

This is the measure problem. Tegmark argued that it is mild and treatable for a single infinite space, but severe for theories in which cosmic inflation, a proposed split-second burst of enormously fast expansion at the start, never ends (called eternal inflation), endlessly creating new regions, each itself infinite [Reference 8]. In a single, evenly mixed infinite space there is a natural way to count: in bigger and bigger spheres, where the fractions settle down to the same answer wherever you center them [Reference 3]. In eternal inflation there is no agreed way to tame the infinities, and different ways give different predictions [Reference 13]. Assuming we are typical observers (not special: one of the crowd), several proposed ways of counting have already been ruled out, because they predict we should see something very different from what we do [Reference 13].

The case against

Not everyone is persuaded. George Ellis, a leading cosmologist and a co-author of the original 1979 argument, has become a prominent critic of where it leads. His central point: you can never count to infinity, so no observation can ever confirm that space is infinite, and claims about regions beyond our horizon are, in his words, “an extrapolation of an extraordinary kind” [Reference 14,Reference 9]. He doubts that real, physical infinities exist at all: infinity “is quite different from a very large number” [Reference 9]. “Maybe the universe closes up on a very large scale,” he writes, “and there is no infinity out there” [Reference 14].

That possibility matters. In a finite universe the infinitely many copies disappear. Copies remain likely only if the universe is bigger than these enormous distances; if it is smaller, there may be none. And in a small universe that wraps around on itself, like a three-torus (the 3D version of the wrap-around screen), looking far enough you would see the back of our own galaxy, not a twin: the same region seen again from around the universe [Reference 15]. A universe that small along loops through us has already been ruled out: such a loop must be longer than 98.5 percent of the diameter of the sphere from which the oldest light reaches us [Reference 7].

Defenders argue that infinite space isn’t an arbitrary add-on. Garriga and Vilenkin argue that, in eternal inflation, the region we live in is naturally infinite [Reference 2]. But that route has a price. In relativity, “now” is not the same for observers far apart; in their picture, whether our region is infinite, and whether a distant copy exists “now”, depends on how you define “now” across the universe [Reference 2]. And eternal inflation is exactly where the measure problem is at its worst [Reference 13]. On one point there is no dispute: today, no information can reach us from these other regions [Reference 2,Reference 14].

What about life out there?

A copy of you would, of course, be life beyond Earth. The same logic gives a striking conditional: if the assumptions hold, and life can arise somewhere with non-zero probability (Earth shows it isn’t forbidden), an infinite universe contains infinitely many inhabited planets, with probability 1 [Reference 1,Reference 3]. That is a statement about a model, not evidence.

The evidence, so far, is this: no life beyond Earth has ever been found [Reference 4], and in 2021 NASA said there was no scientifically supported evidence for extraterrestrial life [Reference 16]. The most intriguing recent result is a rock sample collected by the Perseverance rover on Mars, which NASA announced in September 2025 as a “potential biosignature”: a feature that might have a biological origin but needs more study before anyone can say whether life was involved [Reference 5]. Non-biological explanations have not been ruled out [Reference 5].

So the honest summary of this whole series is a nested set of “ifs”. If space is infinite, which we can’t establish (Part 1); if matter is scattered at random in the right way; if the number of possible arrangements is finite; and if probabilities about imagined universes apply to ours, then there are copies of you out there, almost surely. We will never meet them.

Open questions

• Is space infinite? Unknown, and possibly unknowable; curvature measurements and topology searches can tighten the limits but cannot prove infinity [Reference 6,Reference 7,Reference 9].
• What is the right way to count infinitely many observers? The measure problem is unsolved in general, and different proposals give different predictions [Reference 13,Reference 8].
• Are distant regions really independent and “typical”? Inflation makes this plausible, but it cannot be tested beyond our horizon [Reference 8,Reference 2].
• What counts as “you”? An exact quantum copy and a person with the same memories are very different targets, and the estimated distance depends enormously on which one you mean [Reference 2,Reference 3].
• Is there life elsewhere? NASA says the Perseverance sample needs further study to confirm or refute its biological potential [Reference 5]; beyond that, only future searches can tell.

Five Sigma, “If the universe is infinite, is there another you?,” Five Sigma (12 October 2026), https://thefivesigma.com/articles/another-you. Version of 12 October 2026.