Practical Chip Design

In an uncharacteristically non-integrated-circuit keynote topic, the Hot Chips conference this week offered up Dick Swanson, co-founder, president and chief technical officer of SunPower, to give a brief history of the company. Along the way Swanson outlined the technical evolution that brought the company from, more or less literally, the middle of the desert to a rooftop near you, and may well take it back again. And he gave some examples of the kind of numbers-bending hype that is gradually raising suspicions about the entire sector.

In 1985 SunPower spun out of Stanford University, according to Swanson, convinced that silicon wafer-based solar cells could never be competitive on cost and that the future lay with concentrator technology, in which a large amount of light is focused on a relatively small area of high-output photovoltaic structures. The company envisioned that the lower costs of concentrator cells per square meter of collected ambient light would enable vast solar farms in the desert. But a number of things got in the way.

One was a program at the Jet Propulsion Laboratories that drove the cost of wafer-based cells so low that neither concentrators nor the other favorite of the day, thin-film silicon cells, could catch up. The other was the failure of huge solar farms to emerge, perhaps because by the time they were becoming technically feasible, the Reagan administration had slashed federal subsidies to the solar industry, and there was no economic proposition for building large generating capacity so far from the potential load.

So SunPower found itself with a working concentrator technology capable of very high efficiency, but at very high cost. Each time the company just about ran out of money, Swanson related, a project came along. DoD funding carried the company for a while. When that ended, the first private client was Honda, wanting to win a race for solar-powered cars in Australia. They won using SunPower cells, but it was a one-time sale.

Next was the NASA Helios project, a prototype to create solar-powered electric aircraft that could operate at very high altitudes over long periods of time, in effect offering an alternative to low-earth-orbit satellites. Another technical success, another one-time sale, as the project was cancelled after successfully setting a record for the highest altitude reached by a non-rocket-powered aircraft at 100,000 feet.

Swanson then described his door-to-door campaign to get venture funding for the company, all of which failed. Venture capitalists saw no future in photovoltaics. During this time, the company abandoned its attachment to concentrators and began developing high-efficiency wafer-based cells. At last, with literally no alternative left but to lay off the staff, Swanson approached TJ Rodgers of Cypress Semiconductor, who, Swanson said, listened to his pitch and wrote him a personal check for $750,000 on the spot.

Rodgers and Cypress went on to make about $150 million investment in the company, infusing what had been a very academic organization with Cypress's bare-knuckles manufacturing ethos, building a factory, and turning the wafer cell array into a production product. They were rewarded when the company went public at $18 a share—50 percent above the listing price Credit Suisse had insisted on, and closed the day at $27. (As of this writing, the SunPower is quoted at just under $97, down from a 52-week high of $164.49.)

The rest of the story would have been yet another solar start-up, Swanson suggested somewhat facetiously, except for one coincidence. The light-gathering efficiency of the SunPower cell was high enough that its surface appeared black instead of metallic. That aesthetic point, rather than any of the technical parameters, was the swing factor in convincing many home owners to put arrays of the cells on their roofs. And so as point-of-use panels for private homes and businesses suddenly became the killer application for photovoltaic power, SunPower came to dominate that market.

From here, Swanson said, the future is continuous focus on securing more wafer supplies to ease the current shortage, improving efficiency—the company recently claimed to have achieved 23.4 percent—and reducing costs. That last part is pivotal, because to date, Swanson admits, the industry's value proposition for most applications still depends upon substantial government subsidies. "We are making a Faustian bargain," he explained. "Subsidize us for another five years, and by the end of that time we will have reduced the costs enough to eliminate the need for further subsidies."

Of course like Faust, the industry has every hope of escaping their end of the deal on the off chance that the promise doesn't work out. And this optimism sometimes bubbles over into some rather fast and loose computations.

For example, Swanson described the company's frustration that about half of the retail-level cost of a solar rooftop system today is installation cost, not cell cost. This has led the company to explore prefabricated modules, which in turn has led them to reconsider the big farms in the desert. Swanson said that in a demonstration project, SunPower was able to install a farm at such a rate that, in less than the time it takes to construct a nuclear plant, they could have installed the equivalent capacity of cells. This of course neglects the current shortage of wafers. It also appears to bypass in the comparison the fact that the nuclear plant, even based on the sorry operating record of early pressurized-water reactors, would have a much higher duty cycle, and it appears to neglect the dependence of solar cells on weather, angle of incidence, and time of exposure. But the numbers make nice reading.

Swanson made an interesting point in this regard about costs. While for a residential customer the measure of solar cost is parity with grid power—assuming he can sell excess power back to a skeptical utility company—the equation is quite different for a utility company contemplating a large-scale solar farm. There, the question is the marginal cost of other sources of new generating capacity, not parity with the average cost of existing capacity. And Swanson claimed that solar was already, averaged over the estimated 25-year life of the cells, the least expensive source of new capacity.

But that's if you assume substantial increases over the next 25 years in costs for coal, gas, and other alternatives—likely, but far from guaranteed in a subsidy-riddled market. And it also assumes that the utility can actually do something with the power when it's being generated. Solar of course reaches peak generating capacity during a few hours of the day in the summer when the angle of incidence of sunlight on the panels' surface is closest to 90 degrees. Unfortunately, except in a few markets like Phoenix where people tend to run their air conditioners all day every day, that is not the peak demand time for the utilities. People heat a lot of water—and their houses—in the morning, and they turn on the AC and their appliances when they get home from work. If there is a gradual shift to electric vehicles, most of the charging will presumably occur at night, when the solar capacity is zero.

This brings up the question of storage, as one questioner in the Hot Chips audience pointed out. Without substantial cheap, efficient energy storage capacity, solar in many areas just doesn't make sense as a significant proportion of generating capacity, and it cannot be used to offset growth in demand, because it cannot be trusted to be there during the peak load.

Swanson admitted that storage is a very difficult problem, but said that at this point it is simply not needed, with solar even in its poster-child country of Germany only contributing a couple of percent of total generating capacity. He suggested that when solar reached five percent or more—perhaps in 20 years—large-scale storage would be necessary to maintain stability of the electric grid. He did not address how much the lack of storage discounts the value of solar capacity for the utility that must buy the plant, duplicate the capacity, then perhaps watch it sit nearly idle during its peak-generation periods because the added capacity isn't necessary right then.

Addressing other questions from the floor, Swanson dismissed suggestions that photovoltaics never produce as much energy as was required to manufacture them. He traced this idea to a 1975 Scientific American article that, he said, might have been accurate at the time, but had been rendered obsolete by continued improvements in wafer manufacturing. Without giving any specifics on his analysis, he said that SunPower believes the energy break-even point is now about two years—this mostly dominated by the energy required to manufacture the wafers. In response to another question, he said that the best available data on cell useful life comes from a 30-year-old installation at California's Rancho Seco, which is still running, and the cells are "just fine. So a 25-year operating life is conservative."

Actually, according to the owner's Web site, the first portion of the Rancho Seco installation was put in place in 1984, which would have been less than 25 years ago. Many of the panels have been added since. Swanson did say that the known failure modes on the panels include failure of the encapsulation, solarization of the glass, browning of the plastic, and simple failure of solder joints.