TL;DR; Optionals in swift are translated by the compiler to Optional enum instances, and since Any can map to any value, it can be used to store optionals.
How does Swift represent optionals? It does it by mapping SomeType? to a concrete implementation of the Optional enum:
Int? => Optional<Int>
String? => Optional<String>
A simplified declaration of Optional looks like this:
enum Optional<T> {
case none // nil
case some(T) // non-nil
}
Now, a variable of type Any is able to hold an enum value (or any other kind of value, or even metatype information), so it should be able to hold for example a nil String, aka String?.none, aka Optional<String>.none.
Let's see what happens, though. As we see by the Optional declaration, nil corresponds to the .none enum case for all types:
nil == Optional<String>.none // true
nil == Optional<Int>.none // true
[Double]?.none == nil // also true
So theoretically, you should be able to assign nil to a variable declared as Any. Still, the compiler doesn't allow this.
But why doesn't the compiler let you assign nil to an Any variable? It's because it can't infer to which type to map the .none enum case. Optional is a generic enum, thus it needs something to fill the T generic parameter, and plain nil is too broad. Which .none value should it use? The one from Int, the one from String, another one?
This gives an error message supporting the above paragraph:
let nilAny: Any = nil // error: nil cannot initialize specified type 'Any' (aka 'protocol<>')
The following code works, and is equivalent to assigning a nil:
let nilAny: Any = Optional<Int>.none
, as the above Any variable is actually holding a valid value of the Optional enum.
Indirect assignments work too, as behind the scenes nil is converted to Optional<Type>.none.
var nilableBool: Bool? // nilableBool has the Optional<Bool>.none value
var nilBoolAsAny: Any = nilableBool // the compiler has all the needed type information from nilableBool
Unlike other languages, in Swift nil corresponds to a concrete value. But it needs a type to work with, for the compiler to know which Optional<T>.none it should allocate. We can think of the keyword as providing sugar syntax.