SAN FRANCISONvidia Corp. is not getting the 55-nm capacity it needs from silicon foundry giant Taiwan Semiconductor Manufacturing Co. (TSMC), a problem likely to worsen as the graphics chip maker moves to the 40-nm node, according to Doug Freedman, an analyst at American Technology Research.
In a report published Thursday (Sept. 11), Freedman wrote that the "fabless business model seems to be getting stressed out" and that the Nvidia-TSMC relationship "sounds problematic" and "needs to evolve."
In order to minimize its own risk, TSMC is building less capacity as it ramps new technology nodes, according to Freedman. TSMC doesn't get enough visibility from its smaller customers such as Xilinx Inc., Altera Corp. and Broadcom Corp., Freedman wrote, so the foundry cannot take the risk of building new capacity to support larger customers. Nvidia is TSMC's largest customer, he noted.
TSMC's board of directors last month approved a $795 million capital spending plan that includes a push into 45-/40-nm CMOS processes and MEMS.
Last month, Nvidia reported second quarter financial results which included a charge of $196 million to cover a recall of faulty chips. An Nvidia spokesman initially blamed TSMC for the recall, due to a ''weak die/packaging material'' in select devices. The spokesman later recanted, saying Nvidia "worked closely with TSMC on packaging and the material.''
On Tuesday, a New York law firm said it filed a fraud class action lawsuit on behalf of Nvidia investors accusing Nvidia of failing to disclose "unusually high failure rates" or mobile video adapters. The suit, brought by Shalov Stone Bonner & Rocco LLP, names Nvidia, CEO Jen-Hsun Huang and Chief Financial Officer Marvin Burkett and alleges that they issued a series of misrepresentations and omissions relating to the failure rates of the recalled chips.
According to Freedman's report, none of the chips shipped by Nvidia were bad. Some chips were being run at cycle times that were too high for notebooks which led to packages cracking, Freedman said. According to Freedman, the parts Nvidia was shipping were good, but being run in the wrong notebooks.
Freedman also said Nvidia underestimated Advanced Micro Devices Inc.'s latest graphics chip offerings, including the price and performance. Nvidia cut its own prices to stop market share loss, which appears to be working, Freedman wrote.
Last week, Tristan Gerra, an analyst with financial management company Robert W. Baird & Co., said TSMC is seeing orders down substantially in the fourth quarter and could see its manufacturing utilization rates drop to their lowest percentage in five years. Last month, Mehdi Hosseini, an analyst with Friedman Billings Ramsey & Co. Inc., cut his forecast for TSMC due to lower-than-expected demand from foundry customers.