Solar costs don’t decline like computer costs. They decline more like a combination of television screen costs and car horsepower costs, trends you probably don’t even notice. Surprised? You should be. Because as a VC, let me say that we’ve tricked you.
We’ve tricked you because, along with the rest of the tech industry, we’ve convinced you that “technology” is something that doubles in price performance every 18 months. And true, silicon-based performance has been following this trend line, also known as Moore’s Law (more aptly called Moore’s Theorem) for years.
The nature of Moore’s Law is based on the compression of tiny traces on silicon chips. These traces are like racetracks for electrons, sending them around where we want them to go. As the racetrack gets narrower and hairpins and curves back more and more on itself, we can get a longer racetrack on the same acreage.
More racetrack per acre means, on semiconductor chips, more transistors on the same size chip. Computing performance goes up, chip cost stays the same (roughly, the equipment to make the chips gets more expensive, which takes away some of the gains). It nets out to what we’ve been observing: Moore’s Theorem, er, Law — 2x every 18 months or a 60% reduction in $ / performance each year.
Solar doesn’t work that way. It’s different physics. Electricity is generated from every solar substrate (usually silicon) by photons jarring electrons loose to run willy-nilly over the surface of the solar panel. The electrons are captured by metal lines on the surface and channeled into wires connected to the panel.
The problem is, there is a theoretical limit to how much sunlight can dislodge how many electrons, and for a single crystal silicon substrate it’s about 33%. Current lab designs are at about 22% (at SunPower, one of the efficiency leaders), with plans (depending on the company) to increase this to about 23 – 24% over the next 5 years. This is clearly not 2x improvement in efficiency in 18 months, it’s less than 1% improvement per year.
So all the improvements in solar power/cost performance have to come from manufacturing improvements. We’re investing hard in that area, and there are signs of improvement. We’d be delighted with 5% per year cost reduction of the solar cell. That’s a good start, but remember that solar panels still have to be installed, and installation accounts for 50% of total cost of an installed solar system.
Rosy forecasts from the last decade are that installed solar panels could reduce cost to reach “grid parity” (equal to the cost of our current power source prices), in just a few months. But since I listen to businessmen trying to sell me product or raise venture capital for a living, I went somewhere more impartial: the economists at the Energy Information Administration of the US Government and their March 2009 data.
It appears that the EIA’s optimistic scenarios for solar (the “Falling costs” columns) are still far from gas turbine costs TODAY on an installed basis, even in 30 years. These numbers are in constant $ per Kw of base power.
So in the most aggressive “falling costs” case, Photovoltaics (that’s solar energy to you and me) experience a nearly 4% per year cost decline. That reduction doesn’t compare to computer chip declines of 60% per year. It’s somewhere between the cost decline of televisions per square inch 1980-2004 (9% per year) and $ per car horsepower decline (1%, my calculations) in the same period. It’s a manufactured good, not a technology product.
It’s going to be more expensive for solar power for some time to come. But starting today, according to the climate guys, is not soon enough. We need more, sooner. If we’re going to build it now, then it’s going to be expensive.