Space habitats more expensive than terraforming Mars? You're having to import large quantities of volatiles in either case, Mars needs more imported per person supported (since you'll be using lots of them in the upper atmosphere where they only provide radiation shielding and pressure, instead of smaller quantities to directly support biologically active processes), Mars outgasses the hydrogen you import thanks to photodisassociation of water molecules, and Mars has half the available solar power of a 1 AU orbit (and while you can possibly power a habitat with nuclear power, good luck using anything but solar for a terraforming project).

Now, if you're dealing with habitats, yes, Mars gives you the advantage that by reasonably good placement you can have all the CHONSP you need for your habitat right at your doorstep, and an atmosphere that unmodified already shields reasonable amounts of radiation on its own, et cetera. Mars habitats will be cheaper than orbitals (or sealed habitats anywhere but Earth). But you were talking terraforming. Orbitals vastly outperform terraforming Mars in cost effectiveness per person supported (and Mars outperforms any other solar system body other than Earth).

Now, fragility. Yes, orbitals are relatively fragile. But they are also more numerous. If an orbital faces an independent one-in-a-hundred chance of a local catastrophic total failure leading to death of all inhabitants in a decade, and Earth faces a one-in-a-trillion chance of a local catastrophic total failure leading to death of all inhabitants in a decade, then ten orbitals are a hundred million times less fragile collectively than Earth despite being ten billion times more fragile individually. When you consider the raw number of orbitals you can make for the cost of terraforming Mars, orbitals as a group vastly outclass terraforming Mars in their ability to avoid existential risk, over all classes of risk where a terraformed Mars would actually reduce existential risk.