NEW YORK -- Intel CEO Craig Barrett will pay homage to "convergence" in his keynote address at the Supercomm trade show in Atlanta on Tuesday, but it will be the chip giant's upcoming 90-nm (0.09-micron) CMOS process technology that will enable the company to dominate the communications semiconductor market, according to an executive at the Santa Clara, Calif.-based firm.
Predicting a "capacity famine" in cutting-edge processes when the "business turns around," Intel Executive VP and Communications Group GM Sean Maloney said in an exclusive interview here today that Intel's "main thrust" in the communications market is R&D "technology breakthroughs" in IC manufacturing--especially 90-nm CMOS processes--for electronics, photonics and laser applications.
"There's a big push to achieve breakthrough performance in 90-nm [for our communications chips]," Maloney said. "We're looking to do more and more [at the 90-nm node] as we drive through the 0.35- and 0.25-micron geometries. We're looking to leapfrog a [device geometry] generation," he said.
Maloney's disclosure of Intel's intent to turn its 90-nm process technology into a communications workhorse is believed to be the first mention to date of its application beyond microprocessors.
So far, Intel has recently disclosed a fully functional 52-megabit SRAM, based on its new 90-nm process. It has also talked about using the 90-nm process for its next-generation Pentium 4 processor line--code-named Prescott, which will have close to 100 million transistors. Intel is expected to roll out Prescott in early 2003, with other 90-nm chips to follow next year, according to the company.
No slowdown in Internet
Meanwhile, the Intel executive also dropped hints about the company's plans to launch a new, advanced R&D center for communication OEMs, although he declined to elaborate. And, despite the massive downturn in the technology and communications sectors, Maloney also said that Intel has seen no sign of decline in Internet traffic.
"Traffic is undiminished and still appears to be doubling every year," Maloney said, noting that the current pace of manufacturing technology will not be able to keep up with market growth when the business turns around--something Intel is planning for now and predicting will take place between 2003 and 2004.
While acknowledging that "the industry is on its ass," Maloney warned that the curtailing of capital investment during the current business downturn runs the risk of catching chip makers unprepared for the eventual upswing in the market.
Cutting-edge process shortage
He, in fact, believes that chip makers are already unprepared to meet OEM demand for cutting-edge chip processes. "We currently have a shortage of 0.13-micron manufacturing capacity--and this is the middle of the worst downturn in the industry's history," the Intel executive said. "What's going to happen when it recovers?"
And what about the massive investments foundries like Taiwan Semiconductor Manufacturing Co. Ltd. (TSMC) are currently making in next-generation process technology? "That's good. But it's not going to be enough," Maloney warned.
Not surprisingly, he also noted that integrated device manufacturers (IDMs) like Intel still have some advantages over silicon foundries in the market. In the past, most communications-equipment makers depended upon fabless semiconductor makers, many of which had their devices manufactured in vacant foundry capacity designed primarily for graphics chips, he said.
There are other pressing issues in the communications IC market as well. According to Maloney, Intel will be there with new processes and manufacturing techniques, which will do three things that communications-equipment makers want most: "cost reduce, cost reduce, cost reduce."
And leading the process charge will be manufacturing improvements in what Maloney calls "post recessionary technologies." These include four key product segments: 802.11-enabled wireless local area networks, optical, network processors, and Ethernet.
Intel's process prowess and its intent to leapfrog a current generation of semiconductor manufacturing technology will focus on massive cost reduction, building a common production platform, and integration, he said.
The Intel executive also observed that the communications semiconductor sector has been served "by a wide range of process technologies, including low-volume indium phosphide, silicon germanium, BiCMOS and a mish-mash of CMOS." He said that 90-nm technology has some inherent advantages over the current mix of solutions. "90-nm is better for error correction, Internet access technology, power reduction and cost reduction," he said.
Chip directions for 90-nm node
Where will Intel bring its 90-nm process and manufacturing technology to bear first? The company will focus on mixed-signal/radio-frequency (RF) integration and high-speed digital CMOS for network processing, according to Maloney.
Clearly in Intel's sights are current 10-gigabyte-per-second optical systems, which cost in the $10,000-to-$20,000 price range. "These costs have got to come down [100 times] and we are not going to get there through conventional manufacturing methods," Maloney observed.
A key ingredient in Intel's ability to integrate across the full spectrum of next generation communications manufacturing needs came last week with the $50 million purchase of the network tunable laser business and technology from New Focus Inc. As part of the agreement, approximately 40 employees from New Focus have joined Intel. Intel will also license technology and supply certain products to New Focus.
When combined with Intel's existing optical packaging expertise, the acquisition of network tunable laser technology will enable Intel to offer small form factor, low-cost tunable optical transceivers to accelerate the deployment of dense wavelength division multiplexing (DWDM). The acquisition augments the optical networking business that Intel has built over the past several years and will complement Intel's world-class silicon design and manufacturing capabilities.
R&D center in the works
The Intel VP declined to elaborate further on its process and R&D efforts, but said the company would soon culminate in a major announcement about a new development and manufacturing center focused on supplying communications OEMs. Given the influx of low-cost manufacturing into the Pacific Rim, and especially to China, would it be a safe bet to say Intel is considering building its new development and manufacturing center in China--perhaps Shanghai, according to industry observers.
"Several locations are under evaluation, but we have no plans as yet," Maloney said. Currently, Intel has R&D centers in Beijing, where it does human interface and speech recognition research, a communications development center in Shanghai, as well as other development facilities in Europe and Israel.