The attraction of fusion is "free energy" in that hydrogen is plentiful (even deuterium is plentiful, tritium less so). Helium is not unlimited at least here on Earth and we're busy pissing away our supply on party balloons thanks to a US government decision to sell its strategic reserve in the 90s.
But a fusion plant is expensive. It costs money for the raw materials and the labour to build and maintain it such that the energy it produces--if it ever becomes economically viable, which is far from certain--will not be "free".
Computing power might be cheap but it's not free. An AI/robot won't be "free" in the same sense either. They'll cost money to build. Those resources will cost money.
Also, not all energy is the same. Plants that power the electrical grid are one thing. The energy required to hurtle a large metal object into space is something else. Fusion might work quite well for infrastructure but will it be made to work where we currently use kerosene and oxygen? That vehicle too costs money.
The cost of mining raw materials goes to zero if we can build a machine that can build other machines that can autonomously space mine.
The cost of producing energy goes to zero if we can build machines that can build and maintain energy plants.
I would be surprised if much of this hadn't occurred within 200 years, let alone 500.
Add in strong AI, and you remove the need for information workers; we'd all be servants in the employ of a single CEO at the apex; or servants to other servants; or prostitutes, or some other power relation not reproducible from machines because of wanting the authenticity of a person. I don't see such a state as stable (certainly not in a democracy), and if it existed tomorrow, there would be a revolution. But of course the transition will be gradual, so there won't be a revolution; but something post-capitalist, post-scarcity, will need to emerge to stop blood flowing.
The cost of the raw material used to make fuel is absolutely trivial in all nuclear power plants. The market price of uranium would have to increase by more than three orders of magnitude to increase the cost of produced electricity by 10%. At that point, we have unlimited supplies in seawater.
The capital costs of power plants have trended up as safety and monitoring requirements have risen. Given sufficient automation, AI, and simpler and fundamentally more safe designs (which do not need as much active safety), these costs can probably be brought down.
I'm not convinced that 'economically tricky' will mean anything in 500 years. Assuming that we have essentially unlimited free energy, and AI capable of autonomously handling things like building, manufacturing and farming (which doesn't necessarily require strong AI), the cost of everything will ultimately head towards zero, surely?