The IEEE is organising a panel session on this very important topic at the International Solid State Circuits Conference (ISSCC) February 11th in SF. The panel includes Venture Partners from Tallwood (Dado Banatao), KPCB (John Doerr) and August (Andy Rappaport). We also have CEOs from Broadcom (Scott McGregor), Cypress Semi (TJ Rogers), ARM (Simon Segars) and others. See www.issccpanel.org for information on this panel, the topic and the panelists. Attendance is free for IEEE members.
"Rising design complexity is a valid issue in this context, except that one easily finds a huge number of design opportunities which are not billion-transistor scale. Has he heard of wireless sensors, mobile medical, embedded computing, etc. etc. (you see I didn't even use the IoT buzzword...)?"
Excellent point Jayna. All too often, people overlook the fact that the semiconductor market is far more than just the latest multi-billion transistor processor. The world outside of the processor is still an analog one, and those applications you mentioned, and others, will continue to require new designs and new silicon to sense, to control and to communicate beyond the boundaries of the big digital processor.
Despite all the arguments, the semiconductor industry is maturing. The VC are looking for returns on their investments that the semiconductor industry probably can no longer supply. The investments are too large and the competition reduces the odds of success.
I'm not arguing that investment in (expensive) fab capacity is not needed, but it hasn't come from venture capitalists for a long time, and it makes no sense to conflate the issues. Excess fab capacity goes up and down, but as a matter of current reality there is plenty of money in Asia to build more if the business is there to use it. I don't see the evidence that semiconductor startups are unable to get their chips made at a market-tolerable cost -quite the contrary.
And as for huge wins, 1x10^9 is not larger than 1000x10^6. This looking for "killer apps" and "we're not interested unless we're handling a billion $" is just laziness. Of course every VC would like to have taken those quick profits on Facebook (??); not every mineral deposit is a mother lode. (And the entire output of the California gold rush is estimated to be slightly more than one year's output from the rather mundane methods of mining gold in the U.S. today. It is only an analogy but it is entirely apt.) The Internet of Things silicon volume, taking the average of the estimates of many analysts, will be comparable to the PC if not greater.
@Jayna Sheats: The mental processes of professional investors never cease to amaze. Mr. Niles says "More importantly" ultrasmall geometries and cost of fab equipment, etc. are responsible for the decision not to invest in fabless semiconductor companies
You need to read his comments again. While the chart lists fabless as well as fabed suppliers, he wasn't speaking of just the fabless folks.
But the comment is still applicable, even to fabless suppliers. As process geometries shrink and costs skyrocket, problems in getting investment will affect the fabless folks, too. After all, fabless simply means "Someone else will actually produce the silicon." If the someone else runs into problems because of issues getting the financing to keep current on technology, what happens to the fabless shops using them? (Gee, you have this killer new design using bleeding edge technology you invested an enormous amount in, but you can't get design wins because the only fabs that can manufacture your design don't have the capacity to generate the volume you need, and the fabs you were hoping to use can't get the financing to implement the needed technology? You might not be long for this world...)
"Rising design complexity" is a valid issue in this context, except that one easily finds a huge number of design opportunities which are not billion-transistor scale.
There certainly are. But if you are in his part of the financial world, you are looking for huge wins, not bread and butter investments. The stuff you metion may be profitable, but it probably won't be profitable enough.
The mental processes of professional investors never cease to amaze. Mr. Niles says "More importantly" ultrasmall geometries and cost of fab equipment, etc. are responsible for the decision not to invest in fabless semiconductor companies (your historical chart shows that investment in equipment and process disappeared long ago). The disconnect here is a yawning chasm that not even Geoffrey Moore can cross.
"Rising design complexity" is a valid issue in this context, except that one easily finds a huge number of design opportunities which are not billion-transistor scale. Has he heard of wireless sensors, mobile medical, embedded computing, etc. etc. (you see I didn't even use the IoT buzzword...)?
I'm more on the systems / software side of the industry, so I don't pay as much attention to the semiconductor industry, but this is surprising. Just the other day on CNBC they were analyzing a memory chip company and they made the point that there was not the capacity glut that has happened in the past. Has the industry matured to the point where semiconductors are purely a price-driven commodity?
A Book For All Reasons Bernard Cole1 Comment Robert Oshana's recent book "Software Engineering for Embedded Systems (Newnes/Elsevier)," written and edited with Mark Kraeling, is a 'book for all reasons.' At almost 1,200 pages, it ...