[mpweiher]

As I wrote: the tradeoff is the complexity of the device.

I once listened to a fairly long podcast on the history of Wendelstein-7X, and apparently the mathematical/computational modelling required to figure out a decently optimized stellarator simply couldn't be done when the first ones were built, so they were very inefficient and easily beaten by the Tokamaks, and so interest and funding understandably shifted.

That changed when compute became much cheaper, and apparently some researchers created software models that they could run on their PCs and the late 80s, maybe early 90s. This was largely ignored for quite a while, as all the focus was on Tokamaks, for example JET and later ITER.

Having cracked what to build, and somehow gotten funding, the problem was then actually building it. Very, very challenging, and it did take a lot longer to build than even initially planned, and it wasn't clear that they would succeed in building it. But they did.

And ever since then, it's been humming along nicely.