What short interest measures
Short interest is the number of shares that have been sold short and not yet covered. It is a position: the outstanding stock of bearish bets in a security, not the activity in them.
Broker-dealers report their short positions to FINRA on a fixed schedule, twice a month, as of the settlement dates around the middle and the end of each month. The exchanges publish the aggregated figures several business days after each reporting date.
That publication lag is the first thing to internalise. When you read a short interest figure, you are reading a snapshot that is typically a week or more old, and the position may have changed substantially since.
The two ratios
The raw share count is rarely used directly. Two normalisations are standard.
Short interest as a percentage of float divides shorted shares by the shares actually available for public trading, excluding insider and other restricted holdings. This is the more meaningful denominator, and it is the one that produces the eye-catching numbers, a stock can show short interest at a modest percentage of shares outstanding while being a much larger percentage of its genuinely tradeable float.
Days to cover, sometimes called the short interest ratio, divides shorted shares by average daily volume. It is often described as how long it would take shorts to exit.
That description is worth resisting. Days to cover is a ratio of a position to a recent average, and its denominator is exactly what changes when covering actually starts. In a genuine squeeze, volume multiplies, and the true unwind takes a fraction of the days the pre-event ratio implied. As a static gauge of crowding it is useful. As a forecast of duration it is unreliable in precisely the scenario people invoke it for.
Short interest versus short volume
These are constantly conflated, and they are not the same measurement.
Short interest is a position, reported twice a month, showing what remains open.
Short volume is a flow, published daily, showing what fraction of the day's trading was marked as short sales. Much of it is market-maker inventory management, a firm shorts to fill an incoming buy order and is flat again within minutes. That activity is marked short and never becomes a short position.
The consequence is that a high daily short volume ratio routinely means nothing about bearish positioning. See short volume ratio for why the two diverge so far.
Failures to deliver are a third, unrelated measure, a settlement statistic about shares not delivered on time, which is neither a position nor a flow.
What high short interest implies
Crowding, and asymmetric mechanics.
A short seller's loss is unbounded, which means a rising price forces buying rather than merely encouraging it, margin calls and risk limits produce demand that is not discretionary. That is the entire basis of a short squeeze: the position itself creates the buying that hurts it.
What high short interest does not imply is that a squeeze is likely. Most heavily shorted stocks are heavily shorted for reasons that turn out to be correct. The distribution of outcomes has a dramatic tail, and a much larger body in which the short thesis simply plays out.
The interaction with options positioning is where this gets interesting: a squeeze that begins in the stock can be amplified by dealer hedging of call options into the move. See gamma squeeze versus short squeeze for how the two mechanisms differ and where they compound.
Reading it well
Compare a security against its own history rather than against other securities, a level that is extreme for a large-cap index constituent is ordinary for a small-cap biotech.
Watch the change between reports rather than the level. A position being built or unwound is a decision someone is making now; a stable level is a state that may have persisted for months.
And hold the lag in mind throughout. By the time an unusual figure is published, the market has had over a week to react to whatever prompted it.