AS THE cost estimates keep rising, the case for building the Superconducting Supercollider gets steadily weaker. The issue isn't whether the federal government ought to support science generously -- for it certainly should. Nor is there any doubt that the Supercollider is a brilliant concept that, if all goes well, may tell scientists much about the basic structure of matter.

The danger is that this gigantically expensive machine will suck money away from other essential research in basic science to support one line of investigation in the interesting but narrow field of particle physics. You can already see a familiar and disquieting pattern. To sell the idea to the politicians, its scientific promoters have presented it as crucial to American prestige in the world. That appeal is now reinforced by the purely pork aspects of the thing -- the prospect of all that money raining down on Texas. The Supercollider is a circular tube to be built in a 54-mile tunnel at Waxahachie, south of Dallas, to spin subatomic particles into collisions at far higher energies than smaller devices have ever achieved.

Once construction has irrevocably begun, the machine will be completed regardless of price. And if the costs overrun a too-optimistic budget, the money will be found from other budgets -- starting, despite all those promises to the contrary, with the budgets for other scientific research. The principal losers would be the university laboratories that do most of the country's basic research and train its young scientists.

The machine was originally, in 1987, to have cost $5.3 billion. By last spring the Energy Department thought that the figure would be about $8 billion. To get a more precise fix on it, the department has had several committees working on the numbers. One of them has come to a total of $8.9 billion. There are now reports that the department's internal cost-estimating group is working toward a number that is several billion dollars higher.

The Energy Department hopes to have all these figures reconciled into a firm estimate by early September. Congress, pushed hard by the Texans, is already on the verge of appropriating the full $318 million that the administration requested for 1991. But to complete the Supercollider on schedule in 1998 would require scaling those appropriations up rapidly to well over $1 billion a year. Now's the time for Congress to slow down, cool off, and figure out where this money is coming from. If it means squeezing down a broad range of American scientific research to give a priority to this one huge project, the Supercollider's not worth building.