A cyclotron for 500MeV protons isn't that big: D=6.6m H=2.9m on the appended sketch - unless beam focussing brings other constraints.
115kA per small lukewarm copper coil consume only 59kW together, while accelerating the protons for 1mol/year neutrons draw some 1.2MW at the plug. Bigger coils would reduce the 59kW or let increase the gap.
730t Fe cost ~1M€, as little as 100kW over 5 years.
A smaller cyclotron needs more than 1.8T, hence without iron, and the consumption jumps. 4T at the centre need 2×2.2MA. Pure Al at 20K and 60kOe resists 76pΩ×m, so each coil dissipates 52kW at 20K, and the cooler evacuates 5.6MW: as expensive in a month as iron costs once. The power is only proportional to the induction since the dimensions vary too. 6T at the rim limit the coils' size, which I didn't check as it won't improve much the consumption.
Maybe superconductors do that better, I can't tell. The unshielded field remains badly dangerous.
Marc Schaefer, aka Enthalpy