Foundries are bracing for a weak fourth quarter with inventory issues likely to last through the first quarter of 2007, according to industry sources.
The sources cited CMOS sensor supplier OmniVision’s (OVTI) decreased orders with Taiwan Semiconductor Mfg. Co. (TSM) as an indication of weak demand for the foundry sector. OmniVision’s orders for TSM in August and September decreased significantly compared to July, the sources added.
They are being optimistic by guessing the glut will only last through the first quarter of 2007. In addition to Omnivision, of course, this week saw a disappointing announcement from Marvell (MRVL), which itself was jumping on a growing heap of disappointments. And the excess capacity is only just now starting to get installed.
One of the major contributors to the industry’s declining profitability (as measured by return on investment) is none other than industry leader Intel (INTC). Rather than taking behaving like the responsible industry leaders in every other cyclical industry and taking the lead in reducing supply, Intel clings tenaciously to Moore’s law in hopes that ever-falling costs will continue to lead to ever-more innovation rather than just lower prices.
Moving along a path of energy-efficient computing, Intel has pledged technology evolution in microprocessors that will improve performance per watt by 300 percent by the end of the decade.
Speaking to attendees at the Intel Developer Forum in San Francisco, Intel president and CEO Paul Otellini said this goal would be achieved through continued evolution of Intel’s semiconductor manufacturing technology.
Today, the company uses 65nm process for microprocessor design and by 2008, it plans to reduce that to a 45nm design process. By the end of 2010, Otellini said, Intel aims to achieve 32nm microprocessor design.
“We’re not going to slow down on Moore’s Law; we have the design and capability to make it happen,” the Intel president said. He was referring to Gordon Moore, co-founder of Intel, who made a popular observation - or forecast - that the number of transistors on integrated circuits doubles every two years.
The problem is, there are no indications that today’s computers aren’t up to the job. Whereas past developments were needed to keep pace with ongoing software developments, all but the most specialized of computing needs can now be handled with a bare-bones setup. And while reducing power consumption is a worthwhile goal, there doesn’t appear to be any particular hurry in doing so.
Advanced Micro Devices (AMD), the rival that has recently been throwing Intel into fits, has no plans to begin producing at 45 nm until 2008. By that time the manufacturing techniques will have been proven and the cost of equipment will have come down thanks to Intel’s relentless push into the technology. And while Intel will have a significant lead to market with the technology, it had a similar lead at 65 nm and managed to lose share anyway. One has to wonder whether the time and expense will be worth the effort or simply help its competitor, which is taking the more rational approach to capacity expansion.
INTC-AMD 1-yr comparison chart: