For core process R&D there are currently two choices: IBM and TSMC. The other Japanese companies will have to jump one way or the other, or, maybe, form their own consortium possible with UMC of Taiwan. Historically, Toshiba has always had close links with IBM, going back to the Siemens-Motorola-Toshiba-IBM process alliance of the early 1990’s, and today it announced it is joining the the IBM camp.
This will be a blow for the other Japanese semiconductor companies because Toshiba is by far the most financially secure of the Japanese manufacturers, and could have been the core for a Japanese consortium. However, previous attempts to combine resources and manufacturing facilities have failed in Japan and, with Renesas and Matsushita going off on their own collaboration, Toshiba probably felt it had no choice. For all semiconductor IDMs there’s a pretty bleak choice. Invest $6 billion in a 32nm fab, or go to the foundries. With Infineon, NXP, STMicroelectronics and Texas Instruments overtly or impliedly indicating that they will go to foundry, that leaves Intel, Matsushita/Renesas, AMD, half a dozen memory companies and four or five foundries as the only people with 32nm fabs.
The foundries reaction – to cut back on capex – is not a good omen for the soon-to-be-fabless companies. It means the price of wafers will go up. Finally, for everyone with a 32nm fab there’s going to be a new problem. If 450mm wafers are adopted, and the companies which buy most of the world’s manufacturing equipment are pushing hard for 450mm manufacturing equipment to be developed, then there’s the problem that only seven fabs will be needed to make the world’s total demand for transistors.
For the equipment industry to have only seven customers is not an appealing prospect. So the structure of the semiconductor industry is evolving faster than ever with the number of centres doing basic process development liable to shrink to three or four; the number of advanced CMOS fab owners likely to shrink to under ten and, if 450mm wafers ever get to be used, the total number of advanced CMOS fabs in the world shrinking into single figures.
These apocalyptic figures for the future of the semiconductor industry suggest one thing: that fundamental change is in the offing. A possibility is the nanoprinting technique pioneered in IBM’s Zurich R&D centre which has printed 60nm gold dots. That capability translates into a 100,000 dpi printing process compared to printing processes today of 1,500 dpi. Such a technique applied to the manufacturing of ICs could fundamentally change the future of semiconductor lithography. Printed ICs could take us back to the days when every semiconductor start-up had its own manufacturing facility.