Several panelists argued the maturing semiconductor industry will consolidate into a few larger companies building fewer, larger chips. McGregor of Broadcom agreed there is a change ahead, but said the picture is not dire:
We are a consumer of semiconductor startups. We buy on average one or two companies per quarter -- 50 in the last 10 years.
Development costs per chip have gone from a few million to $30 to $40 million for a 28nm chip. That's not a good budget for a VC, and God help you if you have to re-spin the chip.
That said, the number of chips in a box used to be higher. A set top today is almost all air inside because customers perceive value in the size of the box. Its pretty much one chip in a set top these days, and it usually has our logo on it.
So the number of chips is coming down, and therefore the number of semiconductor companies ought to come down. We don't want to stop seeing startups, but there's a lot of M&A between companies that also creates value and a broad IP set.
Etron's Nicky Lu said chips are a rising investment area for top high tech companies such as Apple building super-platforms.
T.J. Rodgers of Cypress noted his company has spawned internal startups such as solar company SunPower which had a $2.6 billion IPO and Cypress' Programmable SoC (PSoC) group which generates $400 million in revenue:
Internal startups help avoid hardening of the arteries in a big company. We started 14 companies and two have been big winners.
PSoC is our newest method for attracting smart people by offering them a clear path to silicon success. It's basically an ARM M0 controller with all the peripherals on board and a blank space that says, "Your IP here." So they don't have to worry about working with TSMC.
Today the cost of a chip can be more than a fab was when we got started. We spent $95 million on PSoC 3. That filters out a lot of VCs.
VCs are out of fabs, and that's OK... The world is better off with another Google or LinkedIn rather than another fab somewhere. It's the right thing to spend VC money more economically in the US.
Of 59 companies ahead of us when Cypress started, only 19 in the US are left. So it's going to be a smaller industry.
Next Page: Looking East and West