NO one can predict what the United States nuclear arsenal will look like in 20 years. But no one disputes that the recent Strategic Arms Reduction Treaty and the end of the cold war make it very likely that significant reductions will occur in the next two decades.For the giant US nuclear-weapons complex, these events are the equivalent of a mid-life crisis. Built to produce America's huge nuclear arsenal, its intermediate and long-term task is to plan for a smaller, more modern nuclear arms industry while simultaneously solving the environmental, safety, and health problems it created over the last 45 years. The American nuclear-weapons complex is made up of 15 major sites in 13 states encompassing an area the size of Connecticut. Employing about 100,000 people, it is run by the US Department of Energy (DOE). Since its inception in 1945, it has produced an estimated 60,000 nuclear weapons, some 20,000 of which are still active, (12,000 strategic, 8,000 tactical) at a cost of $300 billion in 1991 dollars. "The amount of money DOE spends actually producing things will decline," while the budget allocated for compliance with environmental regulations and cleanup of contaminated nuclear production sites "will easily offset any savings," says John Medalia, a specialist in national defense for the Congressional Research Service (CRS). The nearest dollar figure for hazardous-waste cleanup through the year 2019 is $100 billion to $300 billion, he says. DOE's current operating budget for defense, $11.768 billion, is projected to grow to $14.286 billion by fiscal year 1996. Changing world events have created policy indecisiveness as regards nuclear weapons, says one congressional staff member with extensive knowledge of the issue. The result is that DOE must paint its out-year weapons budgets with a broad brush, he says. Not only should DOE assume that the nuclear arsenal will shrink, its approach must be to base its plans on weapons production some 30 to 85 percent smaller than it is today - in the range of 3,000 to 12,000 warheads, says Matthew Bunn, editor of Arms Control Today magazine. The best estimate upon which to base long-term planning is a gradual "draw-down to roughly 3,000 weapons within 20 years," says Mike Nazarr, of the Center for Strategic and International Studies. But "in order to go that low you must have very stringent verification standards - something on the order of eliminating a US and Soviet nuclear weapons manufacturing complex," he says. Nuclear arms experts note that none of these decisions need be made in any hurry, since there are ample nuclear weapons to provide a US nuclear deterrent for years to come. Major decisions by DOE are not likely to be made until the 1993 fiscal year, which begins July 1, 1992, and after the November 1992 presidential election, say these experts. The multibillion-dollar nuclear-weapons complex can best be understood by its three primary functions: production of nuclear materials, the most critical being plutonium, highly enriched uranium, and tritium; manufacture and assembly of weapons components into a bomb; and research, development, and testing. Of these, the research, development, and testing role is most critical, say DOE officials. The reason, in the jargon of DOE, is the need to maintain "nuclear competence," with a strong cadre of nuclear-weapons designers and engineers supported by specialized laboratories and equipment. Such competence resides primarily in three well-known nuclear arms laboratories: Los Alamos, in Los Alamos, N.M.; Lawrence-Livermore, in Livermore, Calif.; and Sandia, in Albuquerque, N.M. A report in progress by Mr. Medalia and others at CRS confirms that the maintenance of competence is likely to be the single most important task of the weapons labs. DOE Secretary James Watkins told Congress this past summer, amid concern that Iraq's nuclear weapons development was much further along than expected, that the labs are strategically qualified to keep track of nuclear weapons proliferation in developing countries. The three laboratories are about equal in size. And though each expects budget cuts, their concern is that the cuts not be so deep as to scatter, and eventually lose, the pool of experts in place. Livermore has a current budget of $1.175 billion and a payroll of 8,000 employees, some 3,000 of whom are nuclear-weapons experts. Los Alamos has about 7,540 employees and a budget of $928 million. Sandia has 8,500 with a budget of about $1.2 billion. According to officials at Los Alamos and Livermore, one-third of their budgets goes for research, development, and testing of nuclear weapons. Add to that the related fields of research in arms-reduction and nonproliferation verification, intelligence, safety and security, and nuclear weapons total a little more than half the two laboratories' budgets. One change in research goals for the nation's nuclear-weapons complex is already evident. DOE will spend $4 billion on cleaning up buildings, land, and waterways that have been contaminated by the dangerous byproducts of nuclear-weapons manufacture.