Texas Instruments Inc. last week revealed it is abandoning its efforts in "superchips," which TI executives as recently as eight months ago said held the promise of a possible billion-dollar annual payoff.
TI is cutting its losses and exiting the read-channel and superchip markets and reassigning the associated engineers primarily to broadband-communications applications, said Tom Engibous, the company's chairman, president, and chief executive, at the Banc of America Securities investment conference in San Francisco.
TI had invested heavily in the emerging superchip technology, which integrates a processor, a read channel, support logic, and memory into a single device for use in hard drives. In addition to dedicating 100 engineers to the project, the company made two acquisitions to gain access to complementary technologies.
In March, Del Whitaker, then senior vice president of analog and logic products at TI, who has since retired, told analysts that its superchip was being evaluated by four of the five top disk-drive manufacturers. Each design win would have represented $150 million to $250 million in annual revenue beginning in 2002, he said.
"We didn't win those designs, [partly because] we were kind of slow getting up the learning curve in execution on the transition to CMOS," said Russ Garcia, vice president of marketing at the storage product group of the Dallas-based company.
"Continued investment [in superchips] is relatively high, and we felt it better to redeploy those guys in the communications business, which is a higher-growth market," Garcia said. "We've had to restructure the goal, so it's somewhat disappointing. But at the end of the day, it's a positive thing for TI, as well as the employees who'll be now driving toward the right markets with the best growth."
A handful of companies are attempting to establish a presence in the superchip market, which addresses a relatively limited number of hard-drive manufacturers.
Cirrus Logic Inc., the first to unveil a superchip, has since introduced a second-generation version and is working on a third-generation device, said Peter Hillen, vice president of marketing and business development in the mass-storage division at Cirrus Logic in Austin, Texas.
"With only a few customers and suppliers, it's always a winner-take-all design-win situation," Hillen said. "If you get the socket, you ship millions. If you don't, you're crying in your beer."
Cirrus Logic has design wins for its superchip with Fujitsu, Hitachi, and Western Digital, he said.
Although the exact value of a design win can be debated, "when you consider that the majority of the disk-drive electronics is in the single chip ... the number gets pretty big, pretty quick," Hillen said.
Others seeking to participate in the superchip market include Lucent Technologies' Microelectronics Group, Marvell Technology Group, and STMicroelectronics, according to analysts.
Forward Concepts Co. estimates that worldwide shipments of superchips will reach 20 million units this year and is aimed at about 10% of the total disk-drive market. But in 2004, shipments of the chips will soar to 350 million units and address the whole disk-drive industry.
While Lucent supplies all the individual components that would be integrated in a superchip, and is "positioned to provide one-stop-shopping," it has not yet introduced a superchip implementation, according to a spokesman for the Allentown, Pa., company. Lucent would not comment on its read-channel customers.
Although TI has stopped its read-channel development, it would not be surprising to see the company re-enter the superchip market at some point, possibly by licensing a CMOS read channel, according to Will Strauss, an analyst at Forward Concepts, Tempe, Ariz.
A key component to creating a superchip is the production of a CMOS-based read-channel IC. Read channels have traditionally been produced in bipolar and BiCMOS processes. TI said its efforts to move to a pure CMOS implementation fell behind.
"It came down to a combination of execution and market conditions, and it made sense to change the business direction," Garcia said.
TI's two acquisitions targeted at furthering its efforts in mass storage were Silicon Systems Inc., a designer and manufacturer of application-specific mixed-signal devices, and Intersect Technologies Inc., a developer of software, tools, and device drivers.
TI plans to maintain a significant presence in mass storage, Garcia said. The company will continue to supply preamps, servo drivers, and DSPs for hard-drive applications, as well as read channels, preamps, and servo controllers for removable storage, and DVDs.
The move away from hard-drive read-channel development "will not change our revenue picture at all for 2000 or 2001, but will change our profitability more to the positive," Garcia said. "We'll actually grow revenue in mass storage."
TI's failure to gain major design wins for its superchip forced the company to make a hard decision to make the best use of an increasingly limited resource, he said.
"When you look at the marketplace today, mixed-signal engineers and high-speed-digital engineers are really hard to come by," Garcia said. "It ends up being a capacity thing, and you want to place the high-value engineers in the most productive businesses."