Electronics Weekly Electronics Design & Components Tech News
According to Arizona analysts IC Insights, the fabless sector grew 27% last year while the overall semiconductor industry grew 31%.
Were the fabless sector’s IC designs any less innovative last year? Were its sales people less diligent? Or was its relatively poor performance due to the fact that manufacturing capacity last year was the tightest it has ever been?
One inclines to think the latter. While foundry wafers were cheap and plentiful, fab-lite made sense. When access to foundries is problematic and prices go up, as they did last year, then the fab-lite companies are just that – light on fab, short on wafers and lacking in sales.
Having said that, there was a Magnificent Seven in the fabless community which grew faster than the market last year:
Company Growth 2009-10 %
1 Mstar 605 – 1,067 76%
2 Altera 1,196 – 1,954 63%
3 Broadcom 4,271 – 6,589 54%
4 Novatek 819 – 1,149 40%
5 Avago 858 – 1,187 38%
6 Xilinx 1,699 – 2,311 36%
7 Marvell 2,690 – 3,592 34%
The Magnificent Seven represented a 47% increase in IC sales last year, and accounted for 45% of the total increase in fabless IC sales in 2010 ($5.7 billion out of $12.6 bn.
On the other hand there were the Seven Duffers who, in total only increased their sales by 8% – Qualcomm, MediaTek, Nvidia, LSI, ST-Ericsson, Realtek, and Himax.
You have to ask: Was the under-performance of the fabless sector in 2010 a function of wafer capacity constraints?
Is in-house manufacturing looking a bit more attractive?
I get your drift, Anonymous, but it does rather support the case for having fab – so long as you don’t have too much fab!
David, I think you are spot on all the time – but on this one I think you wiffed. The numbers are quite different if you take memory out. Since DRAM grew 80% on shortage related pricing and Flash was up 40%. Since Samsung, Toshiba, Micron, Hynix and Elpida are in the top 10 they overweight the % growth on tha back of increased ASPs.
see for yourself
http://www.isuppli.com/Semiconductor-Value-Chain/News/Pages/Semiconductor-Revenue-Expands-by-Record-Margin-in-2010.aspx
Thanks for putting the record straight about Avago, David, as to your analysis of the rationale for medium-sized IDMs, I think you’re spot on
Clever article, but guess what? Avago is not fabless. The Wireless Semiconductor Division is the leading revenue generator for the company and the Fort Collins front-end fab produces 90% of the volume for the division. I share martijn sentiments above that IP is critical and you truly run the risk of losing when you contract out to a foundry. A unique/niche portfolio of products that other companies can’t or won’t make, seems to be a model of success for the medium size companies with fabs (Triquint, Skyworks).
That does sound like a good idea, but the manufacturers that believe process is important will want to use their own and not share it with others to keep their competitive edge, and manufacturers that concede they can’t compete on process generally don’t want to pour resources in the process development anymore. So the outcome of the Japanese experiment isn’t that surprising to me (although I was not aware of it, being still new in semiconductors, so thanks for the information!).
Well Martijn, I think the business model needs thinking out, but the general idea is what the Japanese tried to do five years back – set up a joint fab for the use of a region’s semiconductor manufacturers – with financial input from the companies and the regional government. That was the Japanese plan, and the Japan government was very keen on it, and a company headed up by an NEC guy was set up to pursue it. But it foundered because the various Japanese chip companies couldn’t agree on what processes to run. Obviously hardly anyone can affoed their own fab for their excvlusive use, but a well thought out plan for a regional fab supporting local regional companies could possibly be made to work.
now, in-house manufacturing is looking more attractive. During the next downturn (which is always around the corner) the losses of the fabless/fab-lite will be less severe than those with huge fabs running at 60% utilization. Bit too early to say fab-less isn’t the way to go in my opinion… Especially for the smaller guys it doesn’t make sense in the long turn to make the huge investments involved. Why develop in-house processes if you know you will be running behind?