Your scare quote suggests you may already know this, but for the benefit of others, the answer is that there isn't any. Fractals may look visually complicated, but their information content is fully captured by the routines used to generate them, which include the formula and the coloring system being used.
This is one of the ways in which "information" is a highly counter-intuitive quantity for people. Very small numbers of bits in a given encoding scheme can produce incredibly complicated pictures, but there's still no more information that what was put in to start with. Simply looking at something and going "Yup, that's complicated" does not mean it has a lot of information in it.
This is not addressable with the contrived example of "you can create an encoding scheme that reduces an arbitrarily complex description to an arbitrarily short identifier". After all, this is not a constructed compression scheme. Why does nature expand simple expressions to these particular complicated forms?
Another example might be the primes, which have a very simple description and very subtle structure in their distribution.
I think because that simple description describes a non-linear iterative process that feeds back on itself.
If not these particular forms then it would be some others. That fractals exist is pretty-much the fact that pseudo-random number generators exist: you plug in a small amount of input and it generates noise. The fact that it looks like something is more a sign of the overactive human facility for pattern recognition than anything else. (Also I suspect it would look a lot less like a burning ship if we hadn't been told it was a burning ship)
My point is basically, the system includes the interpreter all the way up to the eye and the mind, and it seems tricky to completely detach such context from a description of how 'complex' things are.
Real numbers are weird. Totally.
Also the fact that you are defining something depending on convergence makes it even weirder, as limits do not tend to commute with anything unless smooth conditions take place. And for these fractals, there is no smoothness near the "border".
In my view, the real source of complexity is the fact that the algorithm can use arbitrary iterations to magnify small differences. It's the unpredictable nature of computing machines, not real numbers. The nature of real numbers is only a guide to developing and explaining that underlying complexity.
Perhaps the interesting thing is that the set of n-bit programs expands to only 2^n possible outputs. Why are some finite number of infinite-length outputs accessible, but not others? Why does nature favor those sequences?
Are they though? A 5-state Turing machine might seem trivial, but there exist programs in it, that we don't know if they ever halt or not [1]. Their behavior is totally non-deterministic for our understanding. There are also minimalistic cellular automata that produce completely unpredictable patterns. [2]
[1] https://en.wikipedia.org/wiki/Busy_beaver#Known_values_for_....
The deep “weird” “inherently chaotic” structure you’re talking about is the structure of the rational numbers.
People arguing about the complexity of the human brain (especially those that know a lot about the way it works) fall into this trap constantly, and it drives me nuts: there's a ton of structure, but proportionally negligible information in the brain's programming.