In October, the company said it hoped to get the chips into volume production by December. Originally, the chips were expected to come out in commercially acceptable volumes in June.
"We certainly hoped to satisfy the critical demand for the holiday season in larger volumes than we have," said Murray Goldman, Transmeta's CEO. "Although shipments increased in November, we fell substantially short of our production goals."
Transmeta shares on Wednesday fell 65 cents, or 24 percent, to $2.05. In the past year, the stock has tumbled 93 percent.
Wednesday's announcement is the latest in a string of bad news for a company that was founded in 1995 by backers who believed they had learned from the failures of other would-be Intel rivals, from Cyrix to Centaur.
But manufacturing delays, intense competition from Intel, and an economic downturn in Asia--where many Transmeta-powered products are sold--have pounded the company this year. Since its successful public stock offering a year ago, the company has switched CEOs twice, its stock price has plunged, and its chips remain scarce in the United States.
The quarterly drop in revenue that the company disclosed Wednesday follows disappointing results in the previous two quarters; revenue slid from $18.6 million in the first quarter, to $10.5 million in the second and to $5 million in the third.
The rapid decline will likely spark debates over Transmeta's survival as an independent company. Some have speculated that it could become an acquisition target.
The 5800 has become a financial and strategic debacle for the company. At the beginning of 2001, the company was busy signing up new customers for its Crusoe 5600 and seeing revenue increase by 50 percent quarter to quarter.
Although revenue began to drop in the second quarter because of the weak economy in Japan--where the vast majority of the company's chips are sold--Transmeta said that a range of new notebooks containing the 5800 would come out in the latter half of 2001. Not only would the 5800 be faster than the 5600, the company said, but also it would be cheaper to produce.
Getting the chip out the door in volume, though, has proven to be extremely difficult. Many of the early 5800 chips failed critical usability tests, which forced Transmeta to postpone the commercial release. Because of delays with the 5800, Toshiba earlier this year canceled a notebook built around the processor for the U.S. market.
The 5800 is actually quite similar to the 5600. The difference is that the 5800 is made on the 130-nanometer manufacturing process, which means it has smaller transistors than the 5600. The 5600 is made on the 180-nanometer process. IBM manufactured the 5600; Taiwan Semiconductor Manufacturing Co. makes the 5800.
The 5500 is a version of the 5800 for Internet appliances.
In all likelihood, the problem was caused by Transmeta's design, rather than anything TSMC did, said Kevin Krewell, an analyst at the Microprocessor Report.
To fix the final problem, Goldman said, Transmeta has come out with a new metal mask set, a device used to connect layers of silicon with metal.
Typically, the chip designer, not the manufacturing foundry, controls the mask set design. "More than likely when there is a metallization change, there is a design bug they are patching around," Krewell said.
Goldman noted: "TSMC has given us very strong support."
Despite the problems, Goldman said Fujitsu released a notebook containing the 5800 chip in Japan. Fujitsu is using the limited number of 5800 chips already produced by Transmeta but testing them thoroughly, a Transmeta representative said.
Although the company's technology continues to get high marks from manufacturers and analysts, plummeting sales are making many question Transmeta's long-term viability.
"This points out the problem of being a one-product company," Krewell said. "I really believe they would be better off as part of a larger company."