If you'd made sixteen sales last year, four of them used equipment and three more a new product replacing an old one at half the price--you might not be pleased.  But let's say half price is four million, the top of your line does a billion things a second day in and day out, and among your customers are NASA, Los Alamos, Aramco, General Motors, the German ministry of science and industry and the French Institute of aeronautical research.  It was a good year.

Cray makes the world's most powerful supercomputer (no qualifications, the highest tech of all), introduced in 1983 with a "for my next trick," since Cray had developed the fastest computer so far several years earlier.  The new one ups throughput by five times and burst rate by eight, and that's saying something, but it's an interim state-of-the-art produced by a second group while Seymour Cray, who designed the first one, works on a third, which will establish itself as, you guessed it, the world's most powerful supercomputer.

These machines simulate complex physical systems and intricate structures.  They do weather modeling, oil field analysis and a host of other things--many still unimagined.  Cray is beyond specifications doing basic research, focusing on pure capacity and assuming applications will follow.

This puts the company in perpetual anticipation of its market, a somewhat awkward position.  But halving the price of the bottom of its line (new technology retains capacity at the reduced cost) has put supercomputing within reach of a broader range of institutions, and there must be a great deal of satisfaction in being at the growing tip of the computer revolution.