Scuds themselves were notoriously inaccurate. Did Saddam’s weapons developers think they could target this thing more precisely than, say, half a hemisphere?
Just how accurate do you have to be with a nuclear tipped missile?
There are Scuds and there is the Al-Hussein. The missiles that Saddam launched during Gulf War I were Al-Husseins. The Iraqis chopped up three Russian Scuds to make two Al-Hussein, which had longer range and a smaller payload.
The Scud had pretty good accuracy, a coupla hundred meters CEP. The Al-Hussein, like the Scud, was non-separating by design. The longer range by itself would lead to larger errors, obviously. The problems of the Al-Hussein were multiplied by the fact that it was aerodynamically unstable on reentry and would oscillate violently in angle of attack. (Not a prescription for accuracy in a ballistic missile.) The oscillations would cause the jury rigged airframe (bits of one and a half Scuds welded together by crack Iraqi craftsmen) to break up which did not enhance its accuracy very much.
It still had like a kilometer size CEP and could hit cities and posed a threat to large targets like airbases.
There is still the problem of “carrying a nuclear warhead”? During the 1950’s the question was, “Would the AEC make ballistic missiles practical?” Early nuclear weapons needed something the size of a B-29 to carry them. It took the U.S. years of effort and billions of dollars to make nuclear weapons light and rugged enough to be carried by ballistic missiles.
Hussein spent about ten billion ( 1e10) dollars on his nuclear weapons program prior to Gulf War I; he had 100,000 people working full time for ten years on nuclear weapons. He may have been about a year away from making a nuclear weapon in 1990. Had he waited, the invasion of Kuwait may have been completely unnecessary.