How Many Image Generator Vendors Will There Be in Ten Years?
by Roy Latham

The simulator business continues under the curse of "may you live in interesting times." This month we have three news items to ponder: Evans & Sutherland acquires Terabit and makes an alliance with SEOS, AAI Imaging (nee Graphicon) bites the dust, and Silicon Graphics unveils a new image generator that will make a very serious run at aerospace simulation markets. The latest news can be pondered in the light of past events, such as the Lockheed Martin acquisition of Loral's simulator operations, and last year's Hughes' acquisition of Link.

Let's start with a tabloid-style question: In the long run, will anyone be able to compete with Silicon Graphics in the high-end image generator market? The argument that no one will be able to compete is succinct. Making a new high-performance image generator is enormously capital intensive. Silicon Graphics designed thirteen new ASICs from scratch for their new product, including a number of huge, expensive designs. A rough guess is that the development investment would be $30 million, provided one already had the basic infrastructure of technology available from a previous generation of image generator product. Otherwise it would not be so cheap. To match that kind of product development, a company must have both the money to spend and the will to spend it.

There are a host of companies that unquestionably have the financial ability to mount the effort. Lockheed Martin surely must be in the forefront of such speculation. With Loral merged, it will be a $30 billion dollar company. Moreover, they unquestionably have the technical talent and they are (with Hughes) one of the two largest simulator companies, so they have market opportunities. The problem is that aerospace companies have, with exceptions, generally lost the habit of investing in company-sponsored research and development. The mentality from the shrinking defense industry is now in consolidation, downsizing, and pay-as-you go projects. There is reason to suppose that Lockheed Martin is actually more visionary than many of their aerospace peers, but let us just say they are living in tough times in a tough neighborhood.

French aerospace giant Thomson is an example, they seem to be ready to abandon their previous investment in image generator technology and to jump on the SGI bandwagon. The AAI Imaging case seems even more straightforward: the IG market outlook looks grim for minor players. Hughes started and abandoned four image generator development projects in the past decade; it's doubtful they would try again.

It is even more difficult to make the case for other able companies that they might be willing to make the big investment. Computer companies like Sun or Hewlett-Packard are likely to be uncomfortable with the long term heavy commitment and the major differences between graphics products and computers. Sun made a run at graphics in the late 1980's and came away snake bitten; they now make only mid-range CAD-oriented products.

Large companies further afield, for example communications companies, are unlikely to think of graphics as within their domain of expertise.

E&S is an interesting case. Their acquisition of Terabit and alliance with SEOS shows they are mustering the will to do what it takes to be competitive in the simulator business, after much wandering in the desert in search of enlightenment. They are a technically strong organization, though traditionally a bit too subject to falling in love with their own ideas. As one of the few major independent companies in the simulator business, however, they do not have an inside track on applications and they do not have a wealthy parent to fund product development - or to refuse to fund product development. They have the will, but will they find the way? Probably.

Insofar as SGI makes a good business out of high-end image generators, they inadvertently encourage others to find both the will and the means to compete. There is nothing like the smell of huge profits to shake capitalists from their lethargy. Particularly profits in a proven market by someone else.

Can anyone compete with SGI in making high end image generators? Yes. Because of the high costs of new designs, there will be fewer vendors in the future than today. Ten years hence there may be three or four. The market can and will support several competitors that prosper, so the will and the way will be found.

Away from the high-end of image generators, the potential markets are huge and the costs are relatively modest to develop one or two custom chips to enter the market. Safe to say there will be many competitors. It is possible that low-end and mid-range products might rise up to challenge the high-end products. For this to happen, much of the high-end applications must be satisfied by widely available products. If virtually all users needs are met by, say, 10 million polygons per second and 500 million pixels per second, then one or two chip products will, with the advance of technology, eventual satisfy virtually everyone's needs. Automobiles today are no bigger and no faster than they were thirty years ago. No, in the graphics world it's not likely.