Why do computers use binary?
Because it is far easier to build a component that reliably tells apart two states than ten, and reliability is what matters when you are switching billions of times a second.
Simple intuition
The plain reason, in everyday words
Inside a chip, a number is stored as a voltage on a wire. Suppose you wanted to store a decimal digit directly: you would have to divide the available voltage range into ten bands and reliably tell which band you were in. Every wire picks up a little electrical noise from its neighbours, temperature drifts, and the supply voltage wobbles — so those ten bands would each be narrow and easy to confuse. With two states you get the widest possible bands: anything low is zero, anything high is one, and there is an enormous gap in the middle where nothing is allowed to sit. A signal can be badly degraded and still be read correctly, then cleanly regenerated. Binary is not deep mathematics. It is the choice that makes each individual decision almost impossible to get wrong, repeated billions of times a second.
Computers use binary because computers only understand ones and zeros in some fundamental way.
It is an engineering decision about reliability. Decimal and ternary computers were built and worked; they lost on cost and noise margin, not on principle.
Binary is the mathematically optimal base.
Under a simple cost model the optimum is around e, closer to three than two. Binary wins on physical implementation — discrimination margin and gate simplicity — which the mathematical model does not capture.
Modern hardware has moved beyond two states.
Logic remains binary. Multi-level encoding appears in flash memory and high-speed links, where density or bandwidth is the constraint, and always with substantial error correction attached.
Quantum computers replace binary with many states.
A qubit is a two-level system too. The difference is superposition and entanglement, not a larger number of distinguishable levels.
It shows that a foundational-looking design choice is actually a trade-off that was tested against alternatives and won on practical grounds. Understanding the noise-margin argument also explains a lot of otherwise unrelated engineering: why digital audio survives copying and analogue tape does not, why fibre links use error correction, and why storage density and reliability pull in opposite directions in every flash drive you own.
Who worked it out
Binary arithmetic was described by Gottfried Leibniz in 1703, and George Boole's 1847 algebra of logic supplied the mathematics of two-valued reasoning long before there was anything to implement it on.
What problem forced it
Claude Shannon's 1937 master's thesis connected the two, showing that relay switching circuits could implement Boolean algebra directly — arguably the single most consequential master's thesis ever written.
How it changed since
Early machines were not uniformly binary: ENIAC was decimal and the Setun was ternary. As transistor counts grew, the reliability and simplicity of two-state logic settled the question, and the alternatives survive only where density outweighs margin.
How flash memory stores several bits per cell
The one place multi-level storage won, and the price it pays in endurance and error correction.
Why digital copies do not degrade
Regeneration is the reason a file copied a thousand times is identical while a tape copied ten times is not.
Written for Curio rather than collected from a forum — it is part of the curated corpus that ships with the platform. The references it draws on are listed under Sources.