The calculation, in both directions. A call gives the right to buy at the strike, so it is worth exercising when the strike is below the market price, and its intrinsic value is the stock price minus the strike. A put gives the right to sell at the strike, so it is worth exercising when the strike is above the market price, and its intrinsic value is the strike minus the stock price. In both cases the figure is per share, and a standard equity option contract represents 100 shares, so the per-contract amount is 100 times the per-share figure.
The floor at zero is a definition, not a convention. If the subtraction produces a negative number, the option is out of the money and its intrinsic value is zero. It is never negative, because nobody is obliged to exercise an option that would lose money to exercise, and the right not to act has no negative worth. FINRA's own entry for out-of-the-money says such a contract has "no intrinsic value". A reader who computes a negative figure has discovered how far out of the money the contract is, which is a real and useful number, but it is not the contract's intrinsic value.
Where the rest of the premium comes from. An option's market price is almost always more than its intrinsic value, and the difference is the second component, called time value in FINRA's materials and extrinsic value on this site. The same net capital rule that defines intrinsic value defines it directly: "The term time value shall mean the current market value of an option contract that is in excess of its intrinsic value." So the whole premium decomposes into exactly two parts, and the split explains the behavior readers find puzzling. An out-of-the-money option has zero intrinsic value, so its entire price is the second component, which decays to nothing by expiration. A deep in-the-money option is mostly intrinsic value, so it moves nearly dollar for dollar with the stock.
Why the SEC's list of premium determinants matters here. The SEC's own investor bulletin says a premium's amount depends on "the underlying stock price in relation to the strike price", "the length of time until the option contract expires", and the price volatility of the underlying stock. Only the first of those three produces intrinsic value. The other two produce the second component, which is why an option can be worth a great deal while having no intrinsic value at all.
Intrinsic value is not profit, and conflating them is the commonest error. The figure says what exercising would be worth today; it says nothing about what the holder paid to be in the position. A call bought for $6.00 a share that now trades at $4.60, of which $4.00 is intrinsic value, is down $1.40 a share even though the contract is in the money and worth exercising rather than abandoning. Note which two numbers that subtraction uses: the price paid and the price now, not the intrinsic value. The two questions, "is this worth exercising?" and "did this trade make money?", have different answers and need different arithmetic.
A small consequence for holders near expiration. Because the second component shrinks toward zero as expiration approaches, an option's price converges on its intrinsic value, and at expiration the two are the same thing. That is the sense in which intrinsic value is the floor an option's price approaches rather than the price it currently trades at.