Technology traps: defunding the megaprograms.

My advice to the new Congress on technology policy is to kill government science megaprograms, get out of the technology-subsidy business, and double science and technology funding for universities through thousands of small grants. These priorities are particularly important for Republicans who find big-science wonders hard to resist.

With the possible exception of the Manhattan Project, government science megaprograms have a terrible record of return on the taxpayers' investment. Remember synfuels? This scheme to create gasoline from coal followed the classic, eight-step scenario for wasteful government megaprograms:

1) Scare the hell out of them. (What happens when the oil cartel shuts off the gasoline?)

2) Declare that the program is so big, only the government can pull it off. (Translation: No other sucker could be convinced to invest in this loser.)

3) Get expert advice. (Translation: Listen to oil industry lobbyists who are paid to know that what is good for the oil industry is good for America.)

4) Create a consensus. (Translation: Spread the pork out to enough states to get the bill passed.)

5) Execute. (Translation: Use government funds to hire a large P.R. staff.)

6) Fail.

7) Lose $88 billion.

8) Blame the Republicans for underfunding the project.

Remember the superconducting supercollider (SSC)? I debated a particle physicist from the University of Texas-Arlington on National Public Radio on its merits. He claimed that $12 billion was a cheap price to discover the sixth and elusive "top quark" subatomic particle.

I argued that the genius of the physics community would find a cheaper way to float the top quark in electric and magnetic fields long enough to take its picture. A few weeks later, Congress canceled the SSC. A few weeks after that, the top quark had its first snapshot taken at Chicago's Fermi labs. Then, a Texas entrepreneur proclaimed the $4-billion 10-mile hole in the ground created for the SSC an ideal spot for growing mushrooms.

Boeing and Lockheed have just teamed up to work on Space Shuttle II. What did Space Shuttle I accomplish to justify the next multibillion dollar investment? Certainly, it launched many satellites, but they could have been launched more cheaply with disposable rockets. Indeed, if the American taxpayer had not been forced to subsidize those shuttle satellite launches (wiping out any possible competition that would have had to pay full cost), there might now be a viable private American corporation capable of launching satellites--a boon to the entrepreneurs waiting in line for years for a satellite launch.

NASA has run out of useful work for the shuttle, let alone its successor. So we are bombarded by reports of German and Russian astronauts using the Canadian robot ann to perform ecology experiments. The large P.R. efforts that form in step 5 of all government megascience endeavors have learned that spreading the pork (step 4) now must be both an international and a politically correct endeavor.

Some shuttle experiments--at a cost of about $500 million each--are simply ludicrous. Who cares or will ever care if spiders spin their webs differently in zero gravity? And technology con men are having a field day. One University of Houston professor convinced NASA to spend $2.5 billion on five shuttle flights to make space-grown gallium arsenide (GaAs) semiconductor wafers, the starting material for GaAs computer chips. The flight produced five wafers at a cost of about $100 million each. The promise is that in the near-perfect vacuum of space, the shuttle will produce GaAs semiconductor wafers nearly perfect in crystal structure. Eventually, the space-grown wafer cost is projected to drop to $10,000 per wafer.

I am a member of the board of directors of the largest GaAs chip maker in the United States. Here are the facts:

1) Current terrestrial GaAs wafers cost $500.

2) The hypothesized improvement in the crystal structure of space-grown wafers is irrelevant, since the GaAs chip manufacturing destroys and rebuilds the crystal as part of the process.

3) All GaAs companies would go out of business if their wafers cost $10,000 each.

The basic problem with megaprogram funding is that particle physicists, space scientists, and big-company technology experts can have their way with a lay Congress that barely comprehends the complex technologies it is funding. And even that minimal comprehension comes only when huge sums are expended on ever-increasing congressional staffs.

After eliminating the big-science megaprograms, Congress should attack the technology subsidies that Secretary of Labor Robert Reich reasonably calls "corporate welfare." The corporate subsidy most often touted as a success by the Clinton administration (yes, they speak on both sides of the issue) is Sematech, the Austin-based semiconductor research facility that has been given $1 billion in two five-year grants so far. A reasonably well-ran organization, Sematech recently announced it would not seek a third $500-million grant. (Of course, the original Sematech promise was that it would not come back to Congress the second time.) The Clinton administration believes Sematech should be replicated in other industries. But its record is not one that warrants replication:

* Sematech has as members only 12 of America's 200 semiconductor companies.

* Two of Sematech's original 14 members quit because even with their dues halved by government subsidy they could not justify the investment.

* The big companies that control Sematech's board designed the consortium's dues structure to prevent small, entrepreneurial companies from joining. A $20-million chip company that may someday be the next Intel must pay 5 percent of revenue, while Intel itself pays only 0.15 percent of its revenue--a 33-to-1 ratio, which is the primary reason so few companies joined Sematech originally. Of course, Intel, which makes over $1 billion a quarter in pre-tax profits, needs the subsidy a lot less than the small companies that were excluded. But the political system provides the opposite results: Only big companies can muster the lobbying resources to convince Congress to subsidize them. And why would they share the pork with the upstarts?

* Sematech used its government subsidy to attack directly the other 100-plus American chip companies that were not Sematech members. After the checks were signed and the TV lights turned off, Sematech began granting funds to companies that make the critical equipment for the production of computer chips--in return for contracts to hold back the most advanced equipment from all but Sematech members for up to one year. (The deals, which Sematech denied repeatedly, were discovered during a lawsuit.) It is no wonder that Sematech insisted on and received antitrust immunity as part of its funding legislation.

If Sematech's silicon-chip subsidy represents the Clinton/Gore model for government subsidies, it's up to the new Republican Congress to stop its replication. Let's not copy a system that allows well-heeled corporations to use their lobbying clout to entrench themselves with taxpayer subsidies, to the detriment of new companies with new ideas.

The flow of bright, well-educated technologists into industry is much more important to American high-tech businesses than are subsidies to prop up ailing giants. And by cutting out science megaprograms and corporate technology subsidies, the new Congress can both cut the federal budget and free up funds to increase university research funding.

Many Silicon Valley venture capitalists-no friends of big government--believe that the defunct DARPA (Defense Advanced Research Projects Agency) was one of the most effective government technology programs. They credit it with funding such winning pre-venture-capital investments as the UNIX computer operating system work done by Sun Micro-systems founder Bill Joy.

DARPA funded my doctoral studies on transistor physics at Stanford. The high-performance chips I worked on may or may not have improved national defense, but I became one of the hundreds of DARPA-funded Ph.D.s who flooded into Silicon Valley from Stanford and Berkeley.

What caused an unlikely agency like DARPA to provide decent return on government investment?

* DARPA conducted classified military research, which kept Congress on a need-to-know basis. Thus DARPA projects avoided having to spread the pork or to hire a P.R. staff to maintain viability.

* DARPA contracts were awarded by competent technical experts on a merit basis without much political consideration. DARPA also had a "customer," the Pentagon, that had at least a long-run interest in the usefulness of what it funded.

* DARPA tended to fund a large number of small programs, rather than wasteful megaprojects. The agency was on the right side of the economic tradeoff that demands the sacrifice of 1,000 chances to fund the next Bill Joy/Sun Microsystems in order to fund one superconducting supercollider.

Unfortunately, today's ARPA, the non-defense version of the old DARPA, is drifting back into politics. Members of Congress fantasize about "dual use" (military and commercial) technology, with the hope of picking losers and winners, the latter preferably in their districts. There are debates about where the "retraining" funds should be spent when military programs are shut down.

Some of this is inevitable--ARPA's mission is hazier and more politicized than DARPA's. But the agency's best chance for success is if Congress leaves it alone, allowing it to set technical priorities and give out thousands of small grants to universities based only on a peer-review meritocracy.

The new Congress has an opportunity to shrink the federal government and simultaneously help America's technology industries. It involves getting politics out of the laboratory and supporting education on a non-partisan, merit basis.

T.J. Rodgers is president and chief executive officer of Cypress Semiconductor.