IDM or Fab-Lite?

The current debate about the efficacy of going from IDM to fab-lite was addressed some years ago by the founder and former CEO of Bookham Technology, Andrew Rickman.


“I was reading in the paper that a McKinsey report said there would be a complete stratification of the chip industry, that manufacturing would become a cheap man’s activity, and all the intellectual power would go into the design of things. Manufacturing would just become a service, and there’d be all these virtual companies,” says Rickman.

“ARM’s based on that virtual model. But I’d say: ‘No, manufacturing’s not finished’. You won’t create an Intel or a Microsoft based on the ARM model,” says Rickman, “the companies that have the highest profitability are those which own both the process and the design.”

“If you look at the ARM model, you’ve got a good net margin,” continues Rickman, “but if you look at absolute profit from the capability that exists within that company, you’re limited, in terms of revenue, of what you can achieve, because a huge value is actually being delivered by Taiwan Semiconductors or somebody else.”

“If you look at a company like Maxim or Intel, then they have the ability to make a lot more money at the end of the day, even though their gross margin, or net margin, may not be any more impressive.”


Comments

2 comments

  1. If you’re big enough to afford a fab *and* keep it full then you can make more profit than using a foundry — but at 28nm and below very few companies meet these requirements.
    At 20nm you probably need to push $20b worth of chips through the fab over its lifetime to justify building the fab, and who knows what the number will be at 14nm?
    If your devices can be made using a much lower investment “boutique” technology (like larger-geometry SiGe) then the picture obviously changes — or at least, until the same function can be done in standard CMOS and integrated into an SoC…
    In the end it’s the IP and what it can deliver to the customer that really matters. If the IP is simple and easy to copy it’s of little value because anyone can do it; if the entry barrier is high either for technical reasons (design difficulty) or commercial ones (market entry cost) then the IP has real value — but then there will also tend to be only a few suppliers who dominate the market and make most of the money.

  2. Yes yes yes. He’s right.

Leave a Reply

Your email address will not be published. Required fields are marked *

*