The proposed acquisition by On Semiconductor Corporation of AMIS Holdings is a strong statement by On Semi about the state and prospects of the analog/mixed-signal semiconductor market. With the acquisition On Semiconductor will gain both design expertise and manufacturing capacity worldwide.
In 1970 a startup company, then called Computer Design Corporation, became a customer of AMI a semiconductor company based in Santa Clara, CA. The goal was to produce a multi-chip microprocessor that would power a programmable handheld calculator. I had developed the Instruction Set Simulator of the microprocessor, first as a consultant and later as an employee of Computer Design Corporation. That is how I got to know AMI. During the next couple of years I would continue to work with AMI to implement and improve the testing of the chips in manufacturing. Eventually Computer Design, which had gone public under the name Compucorp, declared bankruptcy, and AMI, that had moved its headquarters to Pocatello, Idaho, left the digital CMOS business to concentrate on mixed signals products. I briefly worked with AMI again about ten years ago when I was VP of Engineering at VeriBest, but this time the relationship was brief.
As the article by Peter Clarke shows (see On Semi to Buy AMI) the resulting company will have a presence in complementary market segments and will have a greater breath of design expertise which could result in new products covering multiple markets. Thus, at least on paper, this is an acquisition that is not only good for both parties involved but also for companies designing analog/mixed-signal systems. But what about EDA companies? You do not read much about relationships between an EDA company and a foundry. We are consistently advised of deals between foundries and EDA companies on methods to support the latest digital CMOS process, to the point that a couple of years ago some even speculated that a semiconductor company would acquire an EDA company. But why do we not hear about the same kind of partnerships in the manufacture of analog parts?
Are we so stuck on SPICE based verification that we think there is no way to optimize the design and development methods? Is it not possible to simplify design and verification by narrowing the band of physical behavior engineers need to be concerned about? Could we build macros that can be used as black box components in a larger circuit design verification task?
What do you think? let me know. Go to the Forum section and find the question.