Mannerisms

All three participants in Crolles2, NXP, Freescale Semiconductor and STMicroelectronics have said they will no longer be pursuing basic CMOS process R&D after the 45nm process node when that research is completed at the end of this year.

The whingeing has come from Freescale and ST, not from any European employees of either company, but from their PR people in the US complaining about reports.

This week Freescale's American PR said the company had merely exchanged one consortium doing basic process R&D for another - the IBM cluster. But the press announcement she sent me about Freescale joining the IBM cluster was dated January and only referred to 45nm R&D. The basic research into that node was completed before January.

Earlier this year, ST's American PR simply refused to believe the company had pulled out of doing basic process R&D at Crolles2, until confronted with direct quotations from ST's CEO and COO confirming the move.

Why do the PRs whinge? Maybe they're ashamed of pulling out of basic process R&D. After all it's axiomatic in the semiconductor business that R&D is the last thing you cut. You even cut executive remuneration before you cut the R&D budget.

That's because semiconductor companies depend on a constant stream of new products to maintain revenues and profits. "You'll never get well on the old products", says Gordon Moore, founder and former CEO of Intel.

A constant stream of new products can only come from extensive investment in R&D. That's why semiconductor companies hate to tell their customers they are cutting the R&D. Too often it has been the kiss of death.

"What will happen in a few years' time when all the processes need to be tweaked to deal with image sensors and RF devices? Can this be done with foundry partners, just by passing the manual over the fence?" asked Philippe Geyres recently. Until December last year, Geyres was one of the top ranking executives at ST.

At Renesas, where the strategy is not to cut basic process R&D, there is another reason for still investing in it - time to market. "For Renesas, our strategy is that when it comes to R&D we will continue on our own development", says Satoru ito, CEO of Renesas, "that doesn't mean we will do all our own manufacturing. We need enough capability to manage subcontract manufacturing. To serve our business cycles and business spread it is necessary for us to be able to do SOC design, and pilot line production, and to have a number of options for mass production. That is our strategy,"

So, for Renesas, investing in basic process technology is both a necessary investment for understanding the foundry processes and also a valuable option for getting products into mass production quickly.

Those companies cutting back on involvement in basic process R&D may be whingeing because their companies are embarrassed about cutting the R&D.

It is also possible they are embarrassed about collapsing the basic process R&D project at Crolles2 and thereby ending nearly 25 years of collaborative pan-European involvement in pursuing independent semiconductor process R&D, supported financially by the public authorities.

Once off the Moore's Law treadmill, it takes a mammoth investment to get back in. And the public authorities are going to look askance in the future at anyone trying to get public money to do basic microelectronics process research.

Now that's something worth whingeing about.