The intriguing thing about the chip industry is that it appears to be able to hold contrary opinions simultaneously. Almost nothing you say about it is totally true, almost anything you say about it has some element of truth.
While received wisdom is that you need to spend $1 billion a year on R&D to stay in the wireless IC game, others reckon you can do it for a lot less.
When NXP shuffled off its wireless business, first to STMicroelectronics, and then to Ericsson Mobile Platforms, the reason given was that the combined operation would give the economies of scale needed to compete with the likes of Qualcomm and TI.
Now Freescale has come to the same conclusion and has put up its wireless IC business for sale. One reason is that the pre-2006, pre-private equity, Freescale, is not the same as the Blackstone-owned Freescale today, the main difference being the $700 million it has to find every year to service the debt mountain loaded onto it by its private equity owners.
If you have to spend $700 million before you can spend a penny on R&D, it helps explain why Freescale can't spend $1 billion a year on wireless chip R&D.
But take a start-up company, like Icera Semiconductor, which is competing, like Qualcomm, St-NXP-EMP, TI, Freescale, Mediatek and Infineon at the very heart of the mobile platform - the baseband processor.
I recently asked Simon Knowles, v-p of silicon engineering at Icera, how, on Icera's $75 million spend per year, they managed to stay in the game with the ST's and Qualcomm's of this world which spend $1 billion a year.
"Because we're a small company we are more efficient", replied Knowles, " I spent ten years working at ST, and have no doubt they will spend more money to do something than we will spend to do it."
I definitely agree but I also think it's not perfectly fair to compare $75M to $1B; remember the latter is for companies that also develop application processors, connectivity wireless, analogue, etc. - so the gap is probably at least 2-3 times smaller.
Also Icera has the huge advantage that they only developed one chip per generation so far, and could get away with it because it's seemingly both faster and cheaper than the competition. If you're not in that position, you need to develop a full line-up and that's expensive.
Freescale, with all due respect, I'm not even sure they were competitive at any price point and they moved much too slowly. I wonder who could possibly be interested in their business; Samsung or Marvell comes to mind, but I'm skeptical it'd be a perfect match.