OVER RECENT YEARS, FRESH GENerations of microcomputers and their concomitant acronyms have been appearing with bewildering frequency, doubling and redoubling computing power. Now, just when MIS has made peace with RAM and DOS, the biggest wave yet is poised to strike. The "engine" (as techies are calling little electronic things tht regulate bigger electronic things) of the next units will be a tiny semiconductor that, almost unimaginably, has the data-processing power of a sprawling mainframe of 10 years ago. Ruled by this perky microprocessor, the system on desktops next year -- when the new 32-bit micros are expected to be more plentiful -- will be able to address four gigabytes (1 giga = 1,000 megas = 1 million kilos) of random access memory. That's enough to hold 2 million single-page memorandums all at once, up some 1,992,000 from current levels of the most advanced micros. The market for such "wunderchips" is being contested by, among others, Motorola, AT&T, National Semiconductor, and Intel. Guess which of these IBM Corp. owns nearly 20% of, and you've probably determined the eventual winner.

As for what the future holds, there is about $6 billion worth of IBM-compatible software out there right now, according to one glutton for marginally useful statistics. Doubtlessly, 32-bit machines will continue to run most of it. But not one program at a time, as in the days of '86. The new processors have opened up so much memory and manage data with such alacrity (they are said to be four times faster than some of today's super-minicomputers) that any number of programs, as well as a large number of desk-top terminals, can be worked simultaneously.

The problem is that in comparison to what the new chips can achieve, conventional software is hopelessly insipid, merely condensing chores that would take a human a million times longer to execute. So chip-makers are looking to artificial intelligence -- memory-intensive programs that with the new chip can fit inside a microcomputer, where they will perform logical deductions and reach errorless conclusions on their own -- to revive the sluggish microcomputer market. Personal-productivity software with functional AI, which requires hundreds of thosuands of kilobytes of memory, has yet to be programmed and could be a few years away for desktop use. But it seems clear that the 60-nanosecond manager is on its way -- a least so far as we can figure.