The simplest answer is: we don't know. But we can answer a different question—and potentially demonstrate something remarkable: there may be no second Earth in the Milky Way.The usual formulation of the Fermi paradox goes something like this:
“The universe is enormous, so there must be enormous numbers of opportunities for life to arise. Where is everybody?”
But there is a fundamental problem with that reasoning. We know that the universe contains an enormous number of stars and planets, and we can identify planets that appear potentially habitable. What we do not know is the probability that life actually arises under those conditions.
We have exactly one confirmed example of life: Earth.
From one example, we cannot determine whether abiogenesis is almost inevitable, extraordinarily rare, or somewhere in between. And the uncertainty becomes even greater when we move from simple life to complex life, intelligence and technological civilization.
So instead of trying to calculate the probability of life, let's ask a different question:
How frequently should we expect a planetary system to reproduce the particular sequence of physical events that produced Earth?
This is a question we can at least begin to attack quantitatively.
We can start with the roughly 100 billion stars in the Milky Way.
Milky Way → Sun
First, let's require a star broadly like the Sun rather than a red dwarf.
The Sun is a G2V star. G-type stars make up about 6% of the Milky Way's stars, according to NASA. Red dwarfs, by comparison, account for about 73%.
That gives approximately:
100 billion × 6% = 6 billion Sun-like stars.
This is not yet a statement about habitability. We are simply selecting the stellar environment we want to reproduce.
Sun → Jupiter
Next, we require something much more specific than merely “a planet.”
We want a Jupiter-like giant planet, remaining beyond the ice line at roughly the right distance and on a relatively stable orbit—but not a hot Jupiter, whose inward migration could disrupt or even prevent the formation of terrestrial planets.
This is one of the places where we have an actual observational constraint. A long-term radial-velocity survey of solar-type stars estimated that 6.2% have a Jupiter-like giant planet between approximately 3 and 7 AU.
So:
6 billion × 6.2% ≈ 372 million systems.
This factor already excludes most stars.
Jupiter → Saturn
Now the really restrictive part begins.
We don't merely require another giant planet. We require a Saturn-like planet with sufficient mass, at an appropriate distance, and with a Jupiter–Saturn dynamical architecture capable of producing the kind of stable terrestrial region found in our Solar System.
The resonance and subsequent migration are important here. Jupiter and Saturn's gravitational interactions affect the secular resonances that shape the terrestrial region, and N-body simulations demonstrate that changing the masses and architecture of these outer giants changes the resulting terrestrial system.
There is currently no observationally measured frequency for this exact Jupiter–Saturn configuration.
So here we have to introduce a model parameter.
If we deliberately require an unusually restrictive combination—sufficient Saturn mass, appropriate orbital separation, and a favorable resonant/dynamical history—we can use 0.1% as a conservative working value.
That gives:
372 million × 0.1% ≈ 372,000 systems.
This is the first major uncertainty in our calculation, but importantly it is a quantity that better exoplanet surveys and population-synthesis simulations could eventually constrain.
Saturn → Gaia
Now we require the formation of Gaia, the proto-Earth: a sufficiently massive rocky planet forming in the habitable zone within this particular giant-planet architecture.
We know from Kepler that Earth-sized planets in habitable zones are not exceptionally rare in the broad sense. Published estimates vary substantially depending on the definition of “Earth-sized” and “habitable zone”; some analyses have obtained values around 20% or higher for Sun-like stars.
But that is not the probability we need.
We are conditioning on an already highly unusual Jupiter–Saturn architecture and asking for a planet sufficiently close to Earth's mass and orbit to become Gaia.
For our deliberately restrictive es/timate, let's use 10%.
That gives:
372,000 × 10% ≈ 37,000 Gaia systems.
And this is exactly why conditional probabilities matter. We are not claiming that only 10% of all stars have Earth-like planets. We are asking what fraction of the systems that have already passed all the previous filters also produce Gaia.
Gaia → Theia
Now we require something much rarer.
A sufficiently massive planetary embryo—Theia—must form and ultimately collide with Gaia.
This is not the same as saying that giant impacts are uncommon. Giant impacts are a normal part of terrestrial-planet formation. The requirement here is much more specific: a sufficiently massive embryo must survive the chaotic early evolution and end up making the particular late giant impact that Gaia experienced.
Recent dynamical simulations of terrestrial-planet formation can reproduce a Theia-like last giant impact and constrain the required embryo population and timing.
But there is still no observationally calibrated percentage for the probability of exactly this event.
For a deliberately restrictive working estimate, we use 1%.
Thus:
37,000 × 1% ≈ 370 systems.
Notice that at this stage we have not yet said that the collision produces the Moon. That is the next filter.
Theia → Moon
Now we ask:
Given that Gaia is hit by a sufficiently massive Theia-like body, how often does the impact produce a large Moon like ours?
This is a very restrictive condition. The impact velocity, impact angle, masses, spins and angular momentum all matter.
Numerical studies have found that satellite-forming impacts can occur in a non-negligible fraction of terrestrial-planet simulations, but the probability depends strongly on exactly what is meant by a “massive Moon” and on the assumed planet-formation model. One detailed study estimated roughly one in twelve Earth-like terrestrial planets could acquire a satellite more massive than the Moon, with a very wide uncertainty range from about 1 in 45 to 1 in 4.
However, that is not the same as our extremely specific Earth–Moon outcome.
Indeed, a recent study found that in a particular pebble-accretion scenario, the required proto-Earth–Theia collision at the correct timing and configuration occurs with a probability below 0.1%, illustrating just how dependent the answer is on the assumed formation mechanism.
For our conservative Solar-System-reproduction calculation, we therefore use 1%.
370 × 1% ≈ 3.7 systems.
Moon → Earth
Finally, we require the post-impact planet to remain a stable, long-lived Earth-like planet.
Although the total angular momentum around the Sun is conserved, the heliocentric orbit of the resulting Earth–Moon couple can differ substantially from Gaia's original orbit. If Theia arrived on a significantly eccentric or otherwise unfavorable orbit, the collision could leave the Earth–Moon system on a much more eccentric or dynamically unstable solar orbit. It could subsequently approach the Sun, be ejected from the system, or undergo another catastrophic planetary encounter. Thus, even a successful Moon-forming collision does not guarantee that the resulting Earth–Moon system remains on a stable Earth-like orbit.
Again, there is no observationally measured probability for this exact sequence.
We therefore use a provisional 10% survival factor.
That gives:
3.7 × 10% ≈ 0.37 Earth–Moon-like systems per Milky Way.
So our deliberately restrictive calculation gives an expected number of roughly:
0.4 Earth-like systems per Milky Way
The chain is therefore:
100 billion stars
→ 6 billion Sun-like stars
→ 372 million with a Jupiter analogue
→ 372,000 with our very restrictive Saturn/Jupiter architecture
→ ~37,000 with a Gaia-like proto-Earth
→ ~370 with a Theia-like collision
→ ~3.7 with a large Moon
→ ~0.4 with a stable final Earth-like system
This does not mathematically prove that Earth is the only Earth in the Milky Way. An expected value of 0.4 means that zero, one, or several such systems are possible.
But it does demonstrate something much more interesting than the usual Fermi-paradox argument suggests:
It is entirely plausible that an Earth–Moon system with a Solar-System-like history is extraordinarily rare—potentially rarer than one per galaxy.
And we have not even finished applying possible filters.
We have not yet required Earth's particular atmospheric evolution, its geological history, long-term climate stability, oceans, continents, magnetic environment, plate tectonics, the precise delivery of volatiles, or the extraordinarily long sequence of events that eventually produced complex life.
Most importantly, we have not assigned any probability whatsoever to abiogenesis.
That probability remains unknown because we have only one confirmed example.
This changes the way we should think about the Fermi paradox.
The usual argument is:
billions of stars → billions of habitable planets → billions of opportunities for life → billions of civilizations.
But that chain quietly assumes that “habitable planet” is already a meaningful statistical proxy for “opportunity for life.”
It isn't.
A more scientifically rigorous approach is:
Milky Way → Sun → Jupiter → Saturn → Gaia → Theia → Moon → Earth → life → intelligence → civilization.
We can potentially calculate, or at least constrain, many of the physical and astronomical filters in the first part of that chain.
We currently cannot calculate the probability of the biological filters.
And if the astronomical chain alone already gives us an expected number below one, then the Fermi paradox may have its simplest possible explanation:
There may simply be no other Earth in the Milky Way.
Not because we have proved that Earth is unique, but because Earth-like planetary systems may be rare enough that uniqueness is a perfectly plausible outcome—and we haven't even included all the factors that could make Earth rarer still.
