One need only read through the 1979 Proceedings of the Caltech Conference on Very Large Scale Integration to realize that the VLSI-research community has succeeded beyond our wildest dreams of twenty years ago. At least to the technological reactionaries of those days, the dreams of the visionaries presenting papers at this 1979 conference seemed quite wild, indeed. Million-transistor chips? Extrapolations that thousands of person-months might be required to define and design of a single chip? Channel lengths of 0.2?m, and gate-oxide thicknesses of 70 Angstroms, or even less? Diffusive delays in aluminum wires? Computer-aided design that is truly an aid? Asynchronous and self-timed logic? Computers with thousands of processors?It is, perhaps, only moderately surprising that the future of the technology -- fabrication, design, architectures, and applications -- could be predicted with some accuracy from that 1979 vantage point. What is astounding to me is that the restructuring of the "semi" industry from the vertically integrated form accepted in the 1970s to today's efficiently segmented niches could be foreseen as necessary, desirable, and profitable.In addition to a few reminiscences, I shall describe in this invited talk some of the technical and business adventures that a small group of VLSI researchers from Caltech and USC/ISI has had over the past five years in founding Myricom. Our company is in the commodity interconnect business, and our competitive position depends largely on the chips we design. In way of background, our initial, technical base in multicomputers and routing chips is described in the first and last papers in the Proceedings of the 1992 conference in this ARVLSI series. Except for the design effort, which rarely exceeds ten person-months per chip, Myricom's custom and semi-custom VLSI chips can easily be associated with all of the question-mark phrases above, along with a few other radical ideas.As a business, Myricom could not have started as a commodity-chip company. It takes too long from the substantial costs of designing, fabricating, and characterizing a family of chips, to persuading OEM companies to design chips produced by a startup company into their products, to seeing a profitable volume of sales. Instead, Myricom started by designing and building chips that we used in our own board-level products -- interfaces and switches --, which, when bundled with "open-source" software, gave our early, research customers a turn-key capability of building high-performance clusters of workstations and PCs. From that base in the research community, Myricom "crossed the chasm" to a business in which more than 80% of the company's revenue today derives from sales to OEM and other commercial customers.Our success in this demanding market has required a great deal of attention to and investment in the production and on-time delivery of our products, but has also benefited from our engineering passions for reliability, operating margins, and low-power operation. Of course, we out-source most of our manufacturing -- wafer fabrication, packaging, and circuit-board assembly --, but we retain material-control, production-testing, and quality-control functions in our internal production department. Although Myricom is a somewhat unconventional company, it has been highly successful by conventional, financial measures.Will your technical ideas be next to become a company?