price and application. About twenty computer companies offer multicomputers based on interconnected microprocessors. All are aimed at providing large amounts of computation for highly parallel applications, generally in the technical computing market.

In the spring of 1983, TMC started up in Cambridge, Massachusetts, to build a large, massively parallel SIMD computer. The company was initially financed by a number of private investors, including William Paley and Frank Stanton of CBS. By 1990, over $40 million was raised to finance the firm, which has a valuation of nearly $200 million.

TMC's first Connection Machine was delivered in 1985. Unlike nearly all its SIMD competitors, TMC sells high-end computers in the $1 million to $l0 million range, which it defines as supercomputers. As is the case with Cray and IBM products, the cost of TMC's machines includes support personnel who help with customer applications. As a by-product, these support personnel become experts in applications and help sell additional computers. In 1989, TMC's first profitable year, the company's product revenues reached almost $40 million.

TMC is backed by the Defense Advanced Research Projects Agency (DARPA) Strategic Computing Initiative, whose goal is to develop a teraflop computer by the mid 1990s. A substantial amount of the product development funding comes from DARPA. For example, in 1989, TMC was awarded a three-year, $12 million contract aimed at building a 1-teraflop computer. The SIMD and the multicomputer, unlike traditional supercomputers, can be built in a scalable fashion to provide the most peak computing power. Government agencies and universities either buy or are given DARPA-purchased machines in order to encourage, understand, and develop the use of such machines. By 1990, TMC had sold about fifty computers for large database retrieval, petroleum exploration, and computational fluid dynamics.

This chapter has presented the stories of eight high-tech ventures to give readers some insight into how it should-and shouldn't-be done. The first two examples illustrated two classic failure modes for a start-up: no product but excellent marketing (Ovation), and good product but wrong approach to the market (Analytica). The last six examples illustrated how to create healthy component, system, and software firms. The Bell-Mason Diagnostic (Chapter 10) was used as an overall framework for discussing and evaluating the companies.