IC startups chase new tech
By Rick Merritt, EE Times, (12/06/02 04:46 p.m. EST)
siliconstrategies.com
BURLINGAME, Calif. - Despite the worst climate in recent memory for semiconductor startups, a fresh crop of entrepreneurs arrived here last week proffering chips they said would spur the development of multiprotocol radios, full-motion video on cell phones and wideband wireless LANs, among other next-generation products. Other startups are chasing breakthroughs in underlying process technology issues, such as power consumption, said speakers at the MicroVentures conference here.
“In the last year I have seen some of the best companies I've seen in the last four years. They are solving problems in basic science,” said Rob Chaplinsky, general partner at Mohr Davidow Ventures (Menlo Park, Calif.).
“The mountain ranges are stacked up one behind another,” said Paul Saffo, industry pundit and director of the Institute for the Future (Menlo Park), providing an image of the opportunities that lie ahead. “I think the unintended consequence of the dot-com bust is [that] we have created the largest generation of entrepreneurs this country has ever seen.”
But Saffo was equally quick to trot out an old joke to describe the angst of the investment community. “These days, venture capitalists sleep like babies: They sleep for two hours and then get up crying,” he quipped.
Indeed, there's plenty to cry about. Many startups are watching their valuations decline in lockstep with the stock market as they hit each new round of funding. In part, that's because investors estimate most startups will at best garner something far short of $200 million in a successful initial public offering or acquisition in the current downturn.
As valuations fall, initial investors face the prospect of reinvesting in a company at a loss or seeing their investment wiped out altogether. That has scared off traditional angel and seed fund investors, and it has created a climate where few new startups are forming. The intrepid few that do form strive to keep their “burn rates” of monthly cash consumption to a minimum, since investors typically expect startups to generate sustainable revenue on a total investment of less than $20 million.
In a sign of the times, the MicroVentures conference, which once attracted as many as 600 registrants, drew only 300 this year. Less than 150 people showed up for many of the plenary sessions. And of nearly 70 startups making presentations, only four were formed in 2002.
“There's a lot of unhappiness going around,” said Tracy Richardson, president of startup Stargen Inc. (Marlborough, Mass.), who said he dreads the prospect of a new round of funding coming up. “You have to be No. 1 or [No.] 2 in your category or it's very hard to get funding,” said Tom Sennhauser, president and chief executive officer of storage processor startup Astute Networks (San Diego), his third startup company.
Wireless diving board>/B>
In seeking a growth niche, many startups appear to be riding on the coattails of the few remaining hot markets, especially 802.11 wireless LANs and a new crop of multiprotocol cell phones capable of taking photographs. Indeed, WLANs and cellular were among the main jumping-off points for a number of the startups at the conference.
Airgo Networks (Palo Alto, Calif.), for example, sketched its plans to increase 802.11 data rates and ranges by a factor of 10 based on multi-antenna technology developed in part by its founder, Greg Raleigh, at an earlier startup called Clarity Wireless. “We have new science based on multiple-antenna technology to achieve an order-of-magnitude performance gain without power or cost penalties,” said Raleigh.
Airgo struck another common theme by saying that its chips will ease the security and setup issues that plague 802.11 nets. The company-which has snagged Lucent Technologies' pioneering 802.11 engineers in the Netherlands as well as engineers from Philips and business managers from Atheros-will not detail products for a few months.
Similarly, startup Engim (Acton, Mass.) is claiming “a fundamental breakthrough in chip design” that will allow its four-chip set to provide 50 times more bandwidth for 802.11 access points than existing devices. Engim plans to close a round of funding by June, ship its first products in the following quarter and apply its technology to CDMA and DSL networks in the more distant future, said CEO Art Buckland.
Like Airgo, Engim has amassed an impressive team, including top technologists from Texas Instruments and MIT. Jerry Fishman, chief executive officer of Analog Devices Inc., is a member of the board.
For its part, startup Elliptic Semiconductor (Kanata, Ont.) is designing intellectual property to offload TCP and security processing for 802.11 networks with a core that it says will consume 100 times less power than a general-purpose processor handling those functions. The block will be available in August as a hard core with interfaces to ARM processors and Texas Instruments DSPs, according to executives at the company.
After establishing a foothold in 802.11 chips, Elliptic hopes to design audio/video-processing blocks in 2004, said president Gord Harding.
Startup TelASIC Communications (El Segundo, Calif.), meanwhile, claims that next summer it will unveil the first software-defined radio chip for cellular basestations. The RF chip, made in an IBM Corp. silicon germanium pro-cess, can span CDMA, TDMA and GSM air interfaces and is based on technology pioneered by Raytheon, said CEO Tony Giraudo.
Targeting cellular handsets, startup Ashvattha Semiconductor Inc. (San Diego) said it will deliver an RF transceiver this summer that can simultaneously handle GSM/GPRS, global positioning system and Bluetooth traffic. The 19-person company just won a patent for its isolation structures.
Synthesizer put on-chip
Ashvattha's 10 x 10-mm MME 2700-like the TelASIC device, built in an IBM SiGe process-will sport interfaces to TI, Analog Devices and Philips baseband processors. It requires one external crystal oscillator and a power amplifier. “One of the big problems was getting the synthesizer on-chip, and we were able to do that,” said CEO Kartik Sridharan. If the company is successful, it should see a revenue ramp in 2004, when Ashvattha hopes to launch a follow-on transceiver that supports 802.11.
Also eyeing cellular, Alphamosaic (Cambridge, England) said it will bring CIF-resolution video to handsets at 30 frames/ second, with a compression chip that dissipates 50 mW.
In the increasingly crowded arena of storage networks, startup Ivivity Inc. (Norcross, Ga.) publicly detailed its processor for the first time here. The DiSX 2000 can handle Fibre Channel, Ethernet or iSCSI traffic at rates up to 10 Gbits/s. The sub-$1,000 chip terminates TCP and iSCSI and accelerates a host of storage applications but does not include media-access control or physical-layer functions.
Co-founder Zulfiqar Qazilbash claims the chip is being designed into a storage array and two storage appliances by OEM development partners.
Startup Tarari (San Diego) also emerged from stealth mode at the conference. Vice president of engineering Jeff Carmichael said the company got a thumbs-up from Intel chief Craig Barrett to be spun out of the former Intel communications systems group when it became clear the effort to develop security and XML hardware acceleration boards would not get sufficient market visibility inside the microprocessor giant. Tarari is to formally launch its first products this week.
Along with the activity in wireless, the MicroVentures conference heard startups attempting to forge breakthroughs in chip-manufacturing technologies. Andy Rappaport, a partner at August Capital (Menlo Park. Calif.), said three of his last five deals were focused on process technology. “The ability to capitalize on Moore's Law is becoming increasingly difficult,” Rappaport said. “What used to be second- or third-order effects are now first-order problems. The plain-vanilla CMOS process is really at its limits.”
Not that the biggest chip makers aren't trying. “We think we have attacks on most of the central challenges to get down to 10- to 20-nanometer process technology,” said David Tennenhouse, director of research for Intel Corp. and a keynoter at the event. Intel has two long-term research projects to whittle down the increasing power requirements of high-end microprocessors. Now the company hopes to help crystallize, by 2006, industry work on issues involved in getting to 2- to 5-nm technology, Tennenhouse said.
At the systems level, Tennenhouse cited opportunities in picture-enabled cell phones and wireless networks to help drive Intel's vision of ubiquitous computing. “This year we will have 16 million networked cameras. Where will we store all those pictures?” he asked. “As for 802.11, it is not the ubiquitous-computing network we wanted, but it's a good place to start.”
In Intel's vision of ubiquitous computing, embedded sensor networks feed data directly to computers without human intervention. The company is preparing to invest in and license sensor network technology to Crossbow Technology Inc. (San Jose, Calif.). It is also working with other startups, including Ember Corp. (Boston), to define low-power wireless networks as part of the 802.15.4 standards group, Tennenhouse said. |
