Nvidia cuts forecast, blames 130-nm foundry costs
By Mark LaPedus
Semiconductor Business News
07/28/2003, 5:45 PM ET
SANTA CLARA, Calif. -- Nvidia Corp. here today (July 28) announced it would meet its sales forecast for the quarter, but the company lowered its gross-margin expectations for the period. The graphics chip maker blamed the issues on higher-than-expected product costs for its 0.13-micron foundry technology.
Nvidia did not identify its foundry partners, but the company uses two main providers--Taiwan Semiconductor Manufacturing Co. Ltd. (TSMC) and IBM Corp.'s Microelectronics Division.
The Santa Clara-based graphics chip maker expects to report sales of approximately $455-to-$460 million for the second quarter of fiscal 2004, which ended July 27. This range is within the guidance provided by the company on its last quarterly conference call.
But gross margins are expected to be slightly lower than the company's original guidance. "This was a result of higher than anticipated product costs attributed to 0.13 micron semiconductor process technology," according to Nvidia.
At present, Nvidia's foundry partners are TSMC and IBM. TSMC provides the bulk of Nvidia's foundry requirements, while IBM is making higher-end chips for the company, according to analysts.
In March, IBM scored a major coup in the silicon foundry business, announcing a multi-year alliance with Nvidia. Under the terms, IBM will make Nvidia's next-generation GeForce graphics processor line, based on its 0.13-micron technology. IBM will produce the chips within its new 300-mm fab in East Fishkill, starting in the summer of 2003 ( see March 25 story ).
TSMC, however, is expected to process 95 percent of Nvidia's wafers in 2003, according to a report from Pacific Crest Securities Inc., an investment banking firm in Portland ( see March 25 story ).
In any event, the shift towards 0.13-micron technology has been a challenge for the foundries and prompted Semico Research Corp. to lower its forecast for this industry in 2003 ( see today's story ).
Chuck Byers, head of branding for TSMC, in a recent interview claimed that the company's 0.13-micron technology is superior than IBM's process. But sources believe that TSMC's 0.13-micron yields actually ranges anywhere from 40 percent to 70 percent, depending on the product line.
And despite a recent slew of foundry customer announcements, IBM's foundry unit has been hurt by a "slow improvement in yields" at 0.13-micron technologies, according to SG Cowen Securities Corp. ( see July 17 story ).
Both TSMC and IBM are also reportedly suffering from lackluster yields for chips, based on low-k dielectrics technology. "The yields are about 30 percent at the foundries for chips with low-k," said Michael McConnell, an analyst with Pacific Crest Securities in Portland. "The foundries have a lot of work to do with low-k," he told Silicon Strategies.
For IC makers, however, there are few options for 0.13-micron foundry capacity in the marketplace. "While UMC and Chartered Semiconductor are viable options for 0.13-micron production, we believe their leading-edge yields remain suspect and highly risky for customers," McConnell wrote in report issued last week, referring to Taiwan's United Microelectronics Corp. and Chartered Semiconductor Manufacturing Pte. Ltd. |
