Darpa Buys Into Tiny Computer Chips Firm

A new law allows Pentagon to become venture capitalist to support domestic firms deemed vital to national security SANTA CLARA, CALIF. -- When Jerry E. Crowley threw a lunch last fall for potential investors, he was looking for a few risk-taking venture capitalists to help out his company, Gazelle Microcircuits Inc., a small startup electronics firm here in Silicon Valley. But instead of someone from the private sector, Crowley managed to tap a totally unexpected source of funds: the Department of Defense (DOD). What's more, the money came not in the form of a traditional research-and-development grant but as a direct investment that could, if Crowley's company takes off, actually yield a profit for the Pentagon. The novel agreement is between Gazelle and the Defense Advanced Research Projects Agency, or DARPA, the Pentagon's R&D arm. It was made possible by an experimental program passed by Congress last year that allows the agency to make direct investments in technologies that it believes are critical to the military. If the program succeeds, it could serve as a model for how the government funnels badly needed research dollars into a range of struggling high-tech firms. But that's an important if. Shortly after the agreement was announced, the Defense Department transferred DARPA's director, Craig Fields, to another post in the Pentagon. Published reports indicated that the move was due to White House opposition to Fields' support for threatened industries. "We haven't been directed not to make future investments," says a DARPA source about the program that funded Gazelle. "But we have been told to make certain that the standard contract possibilities are exhausted before we undertake any new agreements." Arati Prabhakar, the young Indian-born physicist who promoted DARPA's pact with Gazelle, says she never thought she would be in the job she now holds. "The last thing on my mind was to work for the federal government," she says. But her initial misgivings have long since vanished. At DARPA, where she is acting deputy director of the defense sciences office, she says, "you get the opportunity to use technology to solve real problems, to do useful things. In that sense, it's not at all like the typical government agency." p In her early 30s, Prabhakar works with the 14 project managers in the defense sciences office, encouraging progress in fields as far-flung as microelectronics and optics to neural networks and biochemical sensors. "My job is to make sure things happen," she says. "To get the technology out of the lab into the hands of users." Prabhakar, who came to the United States at the age of three and grew up in Texas, received her Ph.D. in applied physics at the California Institute of Technology, working under physicist Tom McGill on deep-level electronic defects in semiconductors. Upon graduation from Caltech in 1985, she became a congressional fellow in the Office of Technology Assessment. When DARPA beckoned a year and a half later, she grabbed at the opportunity because it is, as she explains, "an extremely result- oriented organization." Although McGill, her Caltech mentor, happened to be at the Crowley luncheon last October, she knew all about Gazelle long before he ever mentioned the company to her. In fact, she says, "You could say that it was DARPA funding that got the gallium arsenide industry started in the first place by funding the research." --F.G. Gazelle, as its name implies, specializes in speed. It was founded in August 1986 by two young chip-makers, Andrew Graham, now 35, and David MacMillan, 34, who wanted to develop new ultrafast techniques for processing and transmitting data - though not by using conventional silicon chips. Instead, they opted for circuitry made of gallium arsenide (GaAs), which is an inherently faster and less energy-draining semiconductor but is more difficult to work with and requires very special processing skills. When they tried to raise money, venture capitalists told them they had a good idea but needed to bring in a more seasoned executive as CEO. Crowley, 52, filled the bill. An electrical engineer by training, he had led programs for a number of electronics companies, including Motorola, Signetics, and National Semiconductor, and had started up the United States subsidiary of Oki, the large Japanese semiconductor manufacturer. "It's an intriguing technology for which there isn't necessarily a large market," says Harvard University chemist George M. Whitesides, who attended the luncheon at which Crowley spoke. "So it's no surprise that Gazelle was strapped for financing." Nonetheless, Whitesides immediately recognized its value. As an occasional adviser to DARPA, he knew that the company's work involved just the sort of high-tech capability that the Pentagon badly wants to keep in the United States. Gallium arsenide is useful in a broad range of ground-based technologies, from parallel processing computers to high-definition graphics terminals. And it is ideal for satellites, Whitesides explains, "be- cause it is radiation hard." Circuitry made from gallium arsenide components is much more likely than silicon to survive bombardment by cosmic rays and other particles in space. In 1988 Gazelle first capitalized on this technology with a programmable GaAs logic control device that, at a switching speed of 10 nanoseconds, was billed as the world's fastest. It was designed for use in personal computers and workstations. Last November, Gazelle announced a second product: a family of GaAs chips for use in high-speed communications between or within computer systems. After last fall's luncheon, White-sides and another guest, physicist Tom McGill of the California Institute of Technology, told Crowley that he ought to go to DARPA if he really needed a quick infusion of research funding. But the businessman-engineer, who has started up a number of companies, was skeptical about dealing with the Pentagon. "All our people felt it would take too long and too much red tape to get money from it," he recalls. Eventually, he decided to give it a try. But he wanted the government to know he wouldn't stand for any hassle. "I told them that if I didn't get a decision in three months, I'd have to go here," he says, making a none-too-subtle allusion to the Japanese. Not only are the Japanese keenly interested in gallium arsenide as part of their national goal to expand their data works, but they have already approached Gazelle about investing in the firm. To Crowley's astonishment, he got a reply from DARPA well before his self-imposed deadline. It agreed to invest a total of $4 million in Gazelle in four installments. Under the arrangement, Gazelle will design and produce (through subcontractors) GaSa chips and related systems, including software, that can operate at 1 gigabit (a billion bits) or more per second. This is 100 times faster than conventional silicon-based technologies and ten times faster than some fiber-optic channels. In return, DARPA can ask either for royalties on the products that result from the R&D or for cash payments based on the future value of Gazelle's stock. The agency also has a veto power on sale of the technology or even of the company itself. In the latter case, it would be given a chance to seek a buyer within the U.S. DARPA isn't really interested in running a company, says Arati Prabhakar, the acting deputy director of the agency's defense sciences office and the deal's principal architect (see accompanying story). "Our motivation was not to obtain equity in the company," says Prabhakar, "but to ensure that we keep access to certain advanced technologies at our fingertips at the lowest possible cost." Alluding to the sharply rising costs of new defense programs, such as the Stealth bomber, she adds, "We certainly aren't going to build new systems if every node costs another order of magnitude more." What made the unusual deal possible was legislation passed quietly last year at the behest of Sen. Sam Nunn (D-Ga.), chairman of the Armed Services Committee, as a rider to the fiscal 1990 budget. It lets the Pentagon, for the first time, take an entrepreneurial role in promoting technologies in the national interest by making direct investments in companies. In theory, since these investments may lead to profits, the program could eventually be self-financing. According to Richard L. Dunn, DARPA's general counsel, inspiration for the legislation came from an informal group of former high-ranking military officers, led by retired Adm. William Hauser and retired Gen. Bernard Schrieber. "They were worried that the Defense Department wasn't getting access to the best and the brightest researchers," says Dunn. Some of these concerns were expressed publicly in a paper written by a member of the group, Burt Edelson, a former NASA official and retired Navy captain, who is now a fellow at the Johns Hopkins University School for Advanced International Studies. Edelson, who subsequently met with Nunn, argued that DARPA could serve as a model for a civilian technology agency. As a result of these conversations, Nunn asked his staff to draft legislation that would lift some of the restraints that have kept DARPA from investing in the technologies of smaller research-minded companies. Says Dunn, "It's been a general rule in government that if any agency makes any money it goes into the Treasury." Congressional sources acknowledge that some lawmakers were uneasy about letting the Pentagon engage in money-making activities that might give it too much independence from legislative oversight. But after receiving assurances from DARPA that it would make frequent reports on its investment activities, Congress agreed to try out the program for two years. DARPA will be allowed to invest $25 million during the current fiscal year in firms undertaking promising R&D, and another $25 million in 1991. The money, moreover, will come out of DARPA's existing $1 billion budget, rather than from any supplemental appropriation - an idea that Congress killed off at the last moment. Prabhakar says that she picked Gazelle as a test case because she was impressed by its gallium arsenide technology and found it broadly applicable to a number of military needs, including command and control systems. Whitesides concurs: "Gazelle has a team that is especially good at turning new concepts into designs," he says. Despite its credentials, however, Gazelle was an unlikely candidate to receive military backing under existing programs. It and most small high-tech companies don't have the large staffs needed to submit the detailed proposals and documentation that DOD requires of its contractors. Crowley, who had managed until now to raise some $13 million from private investors - including $1 million from Harvard's venture capital firm - is as pleased as Prabhakar with the deal, which he signed at an April 6 ceremony in Washington. "It's great for the U.S. taxpayer, and it's great for DARPA," he says. "Rather than go to Congress and fight for every dollar, they can now develop their own pool of money and use it for longer-term investments, developing the kind of continuity that the Japanese give their research programs." Prabhakar agrees that it gives DARPA a new weapon in its fight to keep the U.S. competitive. "If we make a profit, it's a bonus," she says. "We'll put it into more R&D." Frederic Golden is the West Coast correspondent for The Scientist.

Darpa Buys Into Tiny Computer Chips Firm

Interested in reading more?

Become a Member of