Is it due to precision requirements being difficult to achieve? Supply chain delays for custom parts? Lots of experimental runs that require dis-assembly and tuning?
On top of all the custom hardware you have to be able to set up the infrastructure to use your shiny new hardware, places for staff to work and then land to put it all on (And this is on top of the cost of actually building the main experiment).
In short, everything involved in experimental science is expensive.
On the other hand, we DO build incrementally larger versions. The machine is built and tested in stages, it would be impossible to build such complex machines in just one go.
You're referring to JET? Does ITER benefit from production facilities that were used for it?
Regarding JET and ITER, according to what I heard from my coworkers, all the knowledge that was gained at JET is being used for ITER. I guess many of the productions facilities are being reused too, but I would need to ask them.
Regarding our project (IFMIF), we are building a linear accelerator to simulate the neutron flux inside a fusion reactor and study the behavior of different materials. The accelerator is composed of an injector, several acceleration stages, and a beam dump. First the injector and the beam dump were installed and tested, and then all the other stages are being installed and tested incrementally. But a linear accelerator is a very different machine from a tokamak.