A little bit on what makes the sample return mission so hard: The crux of the matter is getting a fueled rocket onto Mars that can launch back into low Mars or

A little bit on what makes the sample return mission so hard: The crux of the matter is getting a fueled rocket onto Mars that can launch back into low Mars orbit and rendezvous there with a waiting satellite. The size and weight of this return rocket is itself constrained by its ability to fit into an aeroshell and survive entry and landing without breaking. You quickly get into a regime where you're trying to land the heaviest thing ever attempted on Mars, which makes the landing conditions harder so the rocket has to be sturdier, which makes the landing vehicle heavier, and so on. The Martian atmosphere gets you both coming and going. On the descent, it is thick enough to make supersonic retropropulsion very complicated and potentially unstable, but too thin to slow a really massive re-entry vehicle down sufficiently with parachutes. You end up having to fire rockets that kick up a lot of rocks and dust during the very last seconds of the landing, while at the same time relying on your onboard computer to find a safe landing spot. On the ascent, the atmosphere is thick enough that you have to deviate from the most efficient launch trajectory and worry about aerodynamic forces on the ascent rocket. Because of the atmosphere, the launch rocket has to be staged, which makes it bigger and heavier than you'd ideally want. And you can't use the most efficient cryogenic fuel, because there's no way to keep it from boiling off on the long trip over to Mars and down. There are also non-obvious couplings between the lander and the ascent rocket, like center of gravity issues when you pack the rocket into the aeroshell, the question of how to get the samples into the rocket and point it up, and constraints on the design imposed by the hard conditions of atmospheric entry (15G acceleration) and the low surface temperatures. None of these problems are insurmountable, but they are thorny enough and interwoven enough that it's very easy to see how the price tag climbs into the billions, even before you factor in organizational dysfunction and inefficiencies at NASA. The fact that it has to work on the first try (since political factors rule out trying it piecemeal) is also a huge cost driver. There's a great paper that goes into some of these challenges in detail: "Drivers, developments and options under consideration for a Mars ascent vehicle", DOI 10.1109/AERO.2016.7500822