In a familiar animation, Andromeda approaches, passes through the Milky Way, and two spirals become one galaxy. It is easy to treat that sequence as a recording of the future. It is actually one calculation’s outcome for chosen starting conditions.
If the Milky Way and Andromeda pass close together, gravity will rearrange stellar orbits, may draw out long tidal tails, and can enhance star formation when gas is available. Repeated passages can end in a merger. But a guaranteed date and one inevitable scenario are not established: new estimates depend on measured transverse motion, galaxy masses and their satellites’ influence. NASA: the physics of galaxy interactions, Wu and colleagues, 2026: forecast sensitivity.
Approach does not determine the entire path
Andromeda, or M31, is a large neighboring spiral galaxy approximately 2.5 million light-years away. It is approaching the Milky Way. Measuring a changing separation, however, does not establish the full three-dimensional motion. We must also determine how fast it moves sideways, across the direction between us. NASA: Andromeda’s distance, Wu and colleagues: introduction and motion measurements.
Imagine two boats on a lake: their separation can shrink even though they will pass some distance apart. The analogy concerns incomplete velocity information only. Galaxies, unlike this simple example, gravitationally alter their trajectories and do not follow straight lines forever.
M31’s tiny apparent transverse motion is particularly difficult to measure. A small revision can produce a substantially different closest approach in the future. A forecast also needs masses and matter distributions, including the dark halos surrounding the visible disks.
Why sources give both 50% and 90%
Status checked on October 30, 2026. A study by Sawala and colleagues, published in Nature Astronomy in 2025, included uncertainties in the starting data and four major participants: the Milky Way, Andromeda, the Triangulum Galaxy M33, and the Large Magellanic Cloud. Its model gave approximately a 50% chance of a Milky Way–M31 merger within the next 10 billion years. That neither rules out a merger nor gives the probability of two individual stars colliding. NASA: the published 2025 result, accessible author version.
In 2026, Wu and colleagues used the same general framework with updated M31 and M33 proper motions, corrected for systematic astrometric offsets. Their fiducial model raised the merger fraction within 10 billion years to 90%. Among realizations that merged, the median merger time was approximately 6.5 billion years. This study is already published in The Astrophysical Journal Letters, rather than being only a new preprint headline. Wu and colleagues: method, results and limits, publisher record and publication date.
These numbers should not be averaged into a supposedly correct 70%. They describe different sets of input estimates and assumptions. The newer study favors a merger more strongly under its conditions, while also demonstrating strong sensitivity to transverse motion. Its 90% is no guarantee, and 6.5 billion years is no scheduled calendar date.
Smaller galaxies change the outcome too
M33 and the Large Magellanic Cloud are not weightless passengers traveling alongside. Their mass shifts the motion of their large hosts. The direction of that change matters for the Milky Way–M31 relative orbit: it can make the approach more direct or add motion that takes the galaxies sideways.
Thus “the Large Magellanic Cloud always saves us from Andromeda” is incorrect. It reduced the merger fraction in Sawala’s model; with the updated M31 motion in Wu’s model, its contribution favored merging. The geometry of relative motion changed, rather than gravity’s law. Wu and colleagues: Section III.2, Figures 5–6.
Compare two possible outcomes below. The illustrative lines explain the scenario difference; they are not computed future Milky Way trajectories.
A wider flyby
The galaxies approach and alter each other’s motion, but pass sufficiently far apart. No merger may occur within the chosen modeled interval.
Close passages and a merger
In a close interaction, orbital energy transfers into internal motions. Separation, repeated passages and a single common remnant are possible.
Solid markers are the two starting centers; dashed lines only sketch a scenario’s course. Colors distinguish galaxies, rather than luminous dark matter.
Schematic comparison, not a numerical simulation, a map of the actual future or two equally probable tracks. No distance or time scale. Real forecasts need masses, velocities, halos, satellites and uncertainties; first passage is not final merger.
Stars and gas behave differently during an encounter
A galaxy is not a solid ball. Stellar separations are enormous compared with stellar sizes, so direct collisions between individual stars during a galaxy interaction are very rare. Gravity nevertheless acts over large distances: stars can acquire different orbits, and some material can stretch into long tails. The absence of widespread stellar impacts does not mean an absence of change. NASA HEASARC: rare stellar collisions and structural rearrangement.
Those tails are tidal: different parts of an extended galaxy feel different gravitational pulls from its companion. They are not smoke expelled by an explosion. In real interacting systems, astronomers observe tails and young star clusters within them. NASA/Hubble: tidal tails and star-cluster research.
Gas clouds, unlike widely separated stars, can interact strongly: compressing, heating in shocks and redistributing. If suitable cold gas remains, compression can trigger new star formation. Any future starburst in our pair also depends on the available gas at that time; it is not a frame automatically fixed by an animation. NASA: collisions and gas, ESA/ISO: shocks and molecular gas.
First passage and final merger are different events
After an initial close passage, the galaxies can separate again. A final merger requires their relative orbit to lose enough of its energy. An important process is dynamical friction: gravitational interaction transfers orbital energy into internal motions of matter in extended systems, including dark halos. This is neither air resistance nor energy disappearing. Sawala and colleagues: mechanism and orbital model.
With sufficiently close passages, the orbit can gradually shrink and produce a single large remnant with a different structure. In a wider flyby the effect may be weaker, leaving no merger during the studied interval. “No merger within 10 billion years” does not mean “proven never to merge.”
Branch · With numbersOptional · you can skip this
Why distance divided by speed is not a merger clock
Take a rounded distance of 2.5 million light-years and hypothetical straight-line motion at a constant 110 km/s. That speed is close to the present radial approach speed in the Galactocentric frame, rather than M31’s speed relative to the Sun. With c ≈ 300 000 km/s, distance divided by this constant speed gives about 6.8 billion years. Wu and colleagues: velocity frames in the introduction.
This checks simplified arithmetic only: 2.5 × (10)^(6) × (300 000/110) years. It does not predict the actual merger time. Gravity changes speed and direction; transverse motion, satellites, halos and possible repeated passages matter. Approaching centers also do not instantly produce a finished new galaxy.
A better forecast starts with better measurements
A useful forecast requires more than one beautiful track: it must explore many starting conditions within measured uncertainties. Model assumptions matter too: halo shapes and masses, energy transfer, satellites and the wider environment. An ensemble’s merger fraction has meaning together with those conditions and its chosen time interval.
Branch · Preprint of October 28, 2026Optional · you can skip this
Another check: cosmic expansion
In a preprint dated October 28, 2026, without confirmed journal acceptance at our check date, Zhou and colleagues compare idealized pair models with and without cosmic expansion. At small transverse velocities, the difference is often negligible; for some wider orbits, merger timing or outcome can change. The authors explicitly identify limits of the isolated model, while adding M33 and the Large Magellanic Cloud reduces the differences in their examples. Original: Sections IV–V.
This work tests one modeling assumption; it does not establish a new universal probability for our pair. Its result also does not imply that cosmic expansion makes local gravitational approach impossible.
Consider the claim “Andromeda is approaching, so all stars in both galaxies must collide.” It leaves out transverse motion and gravitational orbital evolution, then mistakenly equates a system merger with every star hitting another star.
A possible encounter would change large-scale structure and stellar orbits over time, rather than producing a single instantaneous explosion. The Sun and planets’ specific fate also requires a separate account of their own evolution. Continue with the Sun’s future or revisit how models predict motion.