In the time I was at VLSI, we made a couple of strategic errors relating to EDA. It is perhaps unfair to characterize them this way since it is only with hindsight that the view is clear.
First a bit of history. VLSI was created in the early 1980s to do what came to be called ASIC designs. To do that we had internal tools and they made VLSI pretty successful, first in ASIC and later standard product lines for PC chipsets and GSM phones. VLSI was a pre-IPO startup when I joined and it grew to a $600M company that was eventually acquired by Philips Semiconductors (now called NXP) in a $1B hostile takeover. In 1991 VLSI spun out its internal software as a new company, Compass Design Automation, which never really achieved success. It grew to nearly $60M and eventually (by then I was CEO of Compass) was sold to Avant! for about $90M depending on how you count in 1997.
But let’s go back a bit. In the mid 1980s, VLSI had a major problem. It didn’t have enough money. It didn’t have enough money to build a 1um fab, and it didn’t have enough money to fund TD (technology development, meaning development of the semiconductor process itself) for a state-of-the-art 1um process. So they did major strategic deals with Hitachi and Philips Semiconductors that brought in process technology, patent licenses and money. This meant that VLSI was in business in 1um as a semiconductor company with a new fab in San Antonio and the Hitachi 1um process up and running there.
The really clever decision would have been to foresee that the profitable part of the business was going to be EDA, and VLSI was one of the leaders, if not the leader, at that time. If they forgot about all this fab stuff, they wouldn’t need the money, they wouldn’t need the process, and they could be a very profitable software company. They could have been Cadence, which was just starting to get going at the time. The trouble was that they had semiconductor management whose deep operational experience of running fabs would have been pretty useless for running a software company.
In effect, this would have been spinning out the Design Technology group from VLSI to become Compass, and leaving VLSI as a semiconductor company to die or become an early version of an eSilicon type of fabless ASIC company (since it would have limited money, no process and no fab). But, in any case, eventually it became really obvious that combining a semiconductor company and an EDA company was not a good idea and it was time to split into two viable companies.
The second strategic error was waiting too late to do this. By 1991 when VLSI did it, their technology was no longer way out ahead of the competition, and the industry was not yet looking for the integrated solutions that Compass had (since it had not grown by acquisition). This meant that Compass always struggled to both acquire customers and to acquire library support from other semiconductor companies. This would have been helped if VLSI had sold part of Compass to a VC or someone independent, since there would have been at least part of the ownership of the company that didn’t care about VLSI and only cared about the value of the Compass stock. That would act as a guarantee of independent arm-length behavior by VLSI, which wasn’t there when VLSI owned 100% of Compass (which it did until it was sold to Avant! in 1997). When LSI wanted to license our datapath technology, it was apparently vetoed by Wilf Corrigan because, if Compass’s resources got tight, VLSI would get them and LSI would not.
Interestingly, a couple of years earlier in 1988 Daisy had made a hostile bid for Cadnetix, and had merged the companies to create Daisix. For a number of reasons, this never worked and eventually in 1990 they filed for bankruptcy. VLSI turned out to be owed a lot of money by Daisix (or one of its parents, I don’t remember the details). Daisix was offered in settlement of the debt, but VLSI wasn’t interested and it went to Intergraph instead. If they had taken Daisix, up and running as an EDA company with huge breadth of 3rd party library support, and merged it with Compass’s technology then there was certainly the possibility for Compass hitting the ground running, rather than struggling to earn library party support from other vendors which eventually limited their market largely to people licensing Compass’s libraries.