May 02, 2008
Big Photovoltaic Price Drop Due To Large Silicon Supplies?

Government policies spurred a rapidly rising demand for photovoltaics and these policies caused a rise in market prices. But an article in Technology Review reports some photovoltaics industry analysts are predicting a large drop in PV costs owing to rapidly growing capacity for making silicon crystals. The end of polysilicon shortages could cause PV costs to drop in half.

"It takes about two or three years to add capacity," says Travis Bradford, an industry analyst for the Prometheus Institute. The shortage has been severe enough to drive up silicon prices to more than 10 times normal levels, to $450 a kilogram, adds Ted Sullivan, an analyst at Lux Research.

The added silicon production capacity is now starting to begin operations. While only 15,000 tons of silicon were available for use in solar cells in 2005, by 2010, this number could grow to 123,000 tons, Sullivan says. And that will allow existing and planned production of solar panels to ramp up, increasing supply. "What that means, practically, is that [solar] module prices are going to come down pretty dramatically in the next two or three years," Bradford says.

Last week I linked to a PV stock analyst claiming a big drop in polysilicon costs is coming real soon now. But I'm still left wondering: How far above market prices is the manufacturing cost of polysilicon?

The rise in PV prices in the last several years might finally reverse itself. We can hope so. We need price relief from rising oil and natural gas prices. That price relief can only come in the form of cheaper substitutes.

Update: Japanese PV maker Sharp claims a new thin film PV plant will create PV cells for half the current costs.

In 2007, Sharp started operations at its Toyama plant in Japan for the manufacture of silicon for solar cells and more recently, in February 2008, it announced a collaboration with a production equipment company to develop equipment for manufacturing thin-film solar cells, giving the company a foothold in everything from raw materials to devices across a range of technologies, including polycrystalline and thin-film. In 2005, Sharp began mass production of tandem thin-film solar cells, for instance.

Sharp adds that the end of the 2009 financial year (March 2010) will see the start of operations at its new thin-film solar cell plant in Sakai City, Osaka prefecture in Japan, which will have an annual capacity of about 1 GW, the cost of generating solar power will be about half current levels in 2010. This, says Sharp, will be equivalent to around ¥23/kWh (US¢22/kWh), which is close to the current cost of domestic electricity.

The solar PV market has gotten so big with so many players and technological approaches that substantial price declines seem likely just due to the number of competing teams.

Share |      Randall Parker, 2008 May 02 11:29 PM  Energy Solar

Paul F. Dietz said at May 3, 2008 5:21 AM:

Historically, silicon PV module prices have declined along an experience curve with an exponent of around -.2 (unit prices go as cumulative production raised to this power). If I understand correctly, this trend stopped a few years ago as silicon cost increased. But the underlying improvement in the manufacturing cost very well may have continued, making the cost of the silicon a larger fraction of the total cost. So, if the silicon cost problem is rectified, I expect overall costs could decline dramatically.

Perry E. Metzger said at May 3, 2008 8:03 AM:

I suspect that people are quite wrong about the shortages having eased significantly.

In order to meet demand, polysilicon supplies are going to have to keep doubling annually for the next 15 years or so. Prices are certainly going to drop, but not so much that the polysilicon business isn't going to keep growing radically for a long time to come.

Engineer-Poet said at May 3, 2008 9:01 AM:

If demand is going to be that firm, polysilicon ought to be a really good investment opportunity.

I like SRI's scheme to produce silicon from phosphate-production waste:
Na2SiF6 + Δ -> 2 NaF + SiF4(g)

SiF4 + 4Na -> 4 NaF + Si (l)

Wash to remove NaF.

(filter here removes subscripts. rats.)

kurt9 said at May 3, 2008 10:48 AM:

Polysilicon material is in very short supply. There are new polysilicon plants coming on-line in China in the near future. However, the industry players are being very careful not to crash the price of silicon material. They do not build a new plant until they get firm orders for the expected output of the new plant. This means that anyone trying to source silicon material (like myself) have to get enough aggregate orders (25-50 metric tons per month) so as to get on a buyers list for one of the new plants coming on line.

Greg said at May 3, 2008 10:56 AM:


If supplies are going to increase 100% a year for the next 15 years, then in 2023 we'll have about 1 ton of silicone per each living human. It will be enough to generate 1 MW energy per person in a sunny day Don't you think it's a bit too much?

Aron said at May 3, 2008 2:48 PM:

A quick back-of-the-envelope calculation for home-use at the stated prices.

1) 4000$ \ Kw capacity price for solar cell.
2) 6 KwH per day average production per 1KW capacity as average over year.
3) 6 cents per KwH for grid power.

It thus takes 30 years to re-collect the upfront cost:

4000$ / (.06$) = 67K kwH breakeven / (365 * 6 kwH) = ~30 years

And that would seemingly ignore the cost of maintenance (or replacement) and installation which has to be material. There are fixed costs admittedly in your power bill but you still have to have grid to supplement the solar so these are a sunk cost. Throw in the opportunity cost of that capital at 6%-8% a year and you have a serious problem it would seem.

HellKaiserRyo said at May 3, 2008 4:09 PM:

"If supplies are going to increase 100% a year for the next 15 years, then in 2023 we'll have about 1 ton of silicone per each living human. It will be enough to generate 1 MW energy per person in a sunny day Don't you think it's a bit too much?"

It is 1/15 of the total amount of solar energy on Earth. This includes that amount hitting the land and oceans. We use about 15 terawatts of energy in 2004. That is about 6.7 petawatts.

Engineer-Poet said at May 3, 2008 8:48 PM:

Aron:  don't forget that the price of electricity may rise steeply, and it's considerably higher than 6¢/kWh in many places, including California.  Afternoon peak rates are higher still.  When you include the savings in infrastructure that PV permits (no transmission or distribution for generation sited at the load), PV is already competitive against grid power in some places.

That said, $4000/kW is the current price, not the current cost.  If we consider $45/kg to be the cost of production of bulk PV-grade silicon and something like Evergreen Solar's string ribbon process can turn it into 100-micron cells at a 100% markup (100 tons/month or more), one kilo of Si makes 4.29m² of cells; at 18% efficiency and 1 kW insolation, that's about 770 Wpeak for about $90 in silicon, or about 12¢/W for the cells.  There appears to be a lot of room for price cuts in silicon PV.

Aron said at May 4, 2008 1:37 AM:

6 cents is a BS number. I got that off my power bill for a non-summer month. I see that the national average is closer to 10 cents and I'm sure there is several cents to standard deviate from there. The ideal target figure is really defined as what a reasonably large fraction of people are paying in sunny states.

But I really wonder whether residential PV isn't totally a red herring anyway. Isn't the most cost-effective solution going to remain the centralized one?

I would think something like the stirling system is both more cost-efficient and wholly undesirable for a normal household. :p

M. Simon said at May 4, 2008 5:33 AM:

PV makes sense for peak power shaving in summer months for demand metered operations. For the rest it is just a hobby.

juandos said at May 4, 2008 5:50 AM:

Engineer-Poet says: "Aron: don't forget that the price of electricity may rise steeply, and it's considerably higher than 6¢/kWh in many places, including California. Afternoon peak rates are higher still. When you include the savings in infrastructure that PV permits (no transmission or distribution for generation sited at the load), PV is already competitive against grid power in some places"...

How much of the California price is due to their high taxes?

BTW just how efficient are photo voltaics compared to gas, coal, or nuclear WITHOUT governement (also known as taxpayer extortion) supports?

jms said at May 4, 2008 5:54 AM:

I think that the killer app would be a roof-mounted solar system
that is affordable and just powerful enough to run the air
conditioner. There would be an enormous psychological attraction
to "free" air conditioning, and widespread adoption of such a
system on a per-household basis would largely reduce the
electricity summer peak.

I'm not too concerned about people doing the math and figuring
out that the investment would take 30 years to recoup. People
make this sort of decision from their hearts, not their heads.

Jacob said at May 4, 2008 7:03 AM:

"People make this sort of decision from their hearts, not their heads."

Sure, those that have more money than brains. There are lots of them.

BlogDog said at May 4, 2008 7:17 AM:

OK, how does the investor win in this? What companies are worth the investment to make some money for all us little guys out here?

ajacksonian said at May 4, 2008 7:57 AM:

This the battle between the crystalline silicon people and the copper indium gallium selenide thin film people. While the former tout higher efficiencies, the production cost is sky-high, while the latter show lower efficiencies (Nanosolar publicizing around 14% or so) but production costs lowered by roll-to-roll printing on flexible substrate. Each has their place in the market depending upon up-front cost, needed conversion ratio and life expectancy. If you need higher efficiency in less space and can justify the cost, go for crystalline. If not and can do with lower efficiencies and costs, then thin-film looks good. In-between is the race to find the best market niches for each technology and 'better' doesn't always win in these things as what works is a cost/benefit trade-off.

Whenever anyone mentions 'printing press' with any technology, you are hearing the words: lower cost per unit approaching that of just the cost of materials. It was revolutionary for Gutenberg and remains so to this day for semiconductors. If you can print it, you drive the cost down and hard.

Randall Parker said at May 4, 2008 9:34 AM:


Whether PV makes sense depends on where you far for 2 reasons: Retail costs per kwh vary in the US by about a factor of 3. Second, solar radiation varies by a similar amount.

If you are in high electric cost but high solar radiance SoCal then PV makes a lot more sense on a house than if in lower electric cost Washington state with much lower solar radiance. SoCal is really the ideal place for PV in the United States.

Todd said at May 4, 2008 9:50 AM:

Peter Robinson has a 5 part interview with TJ Rodgers on his solar company, beginning here.

Rodgers predicts the price will be so affordable, most new house construction in sunny climates will have solar panel shingles on them by default.

david foster said at May 4, 2008 9:53 AM:

The inherent storage capability of solar-thermal seems to me to be a big advantage over PV cells, unless there are *major* improvements in batteries or other energy-storage technologies.

Engineer-Poet said at May 4, 2008 10:12 AM:

ajacksonian:  CIGS is cheaper today, but IIUC the world supplies of both indium and gallium are limited.  If that's true, CIGS will not scale.

Silicon is a large fraction of Earth's crust, and can be refined from silicates as well as quartz.  It is safe to say that silicon is effectively inexhaustible, and SiPV will scale.

Randall Parker said at May 4, 2008 10:29 AM:


Yes, I've read the claims that indium and gallium supplies are the bottleneck. I think we need either silicon or carbon based PV to really scale.

My question: What is the energy cost of creating silicon crystals?

david foster,

In the early stages of PV usage storage capacity isn't a problem. PV cuts peak demand for fossil fuels-based electric power. Only once PV capacity exceeds afternoon fossil fuels-based electric usage does storage become desirable.

Granted, if PV costs drop a lot in the next few years then that problem could develop pretty quickly. Also, electric demand peaks late in the afternoon and evening. That doesn't match solar radiation peak at noon. So solar thermal offers real advantages.

But can solar thermal costs drop as much as PV costs? My impression is that solar thermal costs less now but PV is inherently more amenable to cost cutting.

kent said at May 4, 2008 10:59 AM:

RE: Aron's back of the envelope calculations. . .

You were conservative in your expenses, but right on the money for payback.
I've been reviewing the numbers every year or so, but it just isn't cost
effective with 6.6 cent/Kwh electricity.

But everyone leaves out a separate issue of backup power. As minority
environmental groups increasingly win greater control of regulated
power agencies through citizen initiatives and torts, the reliability
of the grid will likely go down. California was a similar example of
lowered reliability for different reasons. As we continue to NOT
build our infrastructure for the future generations in the name of
'conserving our way to sustainability', decreased reliability in utility
services will be a natural result.

I wouldn't consider building a new home without a backup generator. If
solar could compete with the fixed cost of a backup diesel generator of
the same capacity, it (might) be a superior option due to reduced maintenance
and fuel cost.

Engineer-Poet said at May 4, 2008 3:09 PM:

We can calculate the basic energy cost of silicon from the SRI process:  it's the heat required to dissociate sodium fluorosilicate, plus 4 moles of sodium metal per mole of silicon.  There will be some extras in purification of SiF4 gas, washing NaF out of the product, etc. but they are going to be relatively small.

I don't know the exact figures for any of the inputs, but once you have them it's a spreadsheet exercise.  Evergreen Solar's string-ribbon process doesn't produce single crystals (neither do amorphous processes), so the energy cost of those isn't necessarily relevant.

Kent:  It's feasible to run devices off a Prius as a backup generator.  Perhaps you should aim at a PV-fed battery bank (dump excess to the grid?) with a Prius making up for any deficit.  This lets you get use out of all the components except the stationary battery every day.

TTT said at May 4, 2008 3:25 PM:

If Silicon capacity was this acute of a problem for this long, then it is not just PVs that will fall in price, but ALL seminconductors. That means LCD TVs, PCs, networking equipment, automobile electronics, etc. will all drop.

That would be enough for another tech boom.

BTW, here is a solar energy cost curve through 2020, and a solar intensity map.

Randall Parker said at May 4, 2008 3:34 PM:


Computer chips are very small. PV covering a roof uses orders of magnitude more silicon than a PC uses. So silicon costs matter far less to the computer industry than to the PV industry.

TTT said at May 4, 2008 4:03 PM:


Not necessarity, if microprocessors are sold by the hundreds of millions each year.

What percentage of silicon is used by PVs, and what percentage is used by the semiconductor chip industry?

Randall Parker said at May 4, 2008 4:30 PM:


Silicon wafer costs represent a far smaller fraction of the cost of a computer than of a rooftop PV installation.

John Moore said at May 4, 2008 5:37 PM:

If PV is to be used large scale, some form of storage is required. Furthermore, one must consider the elements of grid power cost if the grid is the "storage."

PV generated at peak usage times and pumped into the grid is fine, but at off-peak times, the grid is already using low cost power. If a significant proportion of the power is coming from PV, then the grid suppliers have to still build enough capacity to provide peak power usage on cloudy days.

Hence a true cost calculation requires analysis of grid power costs, with various government interventions removed - or, it has to include a storage system, and storage is very expensive.

Finally, the power electronics to convert the PV power to something usable are not free. Right now, wholesale, it costs $1/peak-watt. That's a lot. In volume it would come down, but probably not to less than 40 cents/peak watt.

Engineer-Poet said at May 5, 2008 6:46 AM:

There are two ideal sinks for PV energy in the sunny parts of the USA:

  • EV and PHEV battery packs
  • Ice-storage A/C systems
These could accept an enormous amount of power (at least as much as peak load for several hours) and offset demand for hours to days.

david foster said at May 5, 2008 7:44 AM:


Storage is ideally needed over a day or more, not just a few hours. Consider: a sunny but cool day (like yesterday, where I am) followed by an overcast, hot, and muggy day. More A/C load on the second day than the first.

Whatever the form of solar, a major problem is that all large-scale projects will be greatly slowed down by litigation--affecting both the land used for the project itself and the rights-of-way for the transmission lines. I fully expect to see protest signs that say "Land for people, not for profits."

Innovation Catalyst said at May 5, 2008 9:22 AM:

If the EESTOR units are more than vapor, you might see PV home storage devices. Otherwise in net metering states your 'storage medium' is the grid itself.

TheOldMan said at May 5, 2008 10:16 AM:

"PV makes sense for peak power shaving in summer months for demand metered operations." Precisely! My 5.5kW system pumps out the most power during 12-18:00 which is just when PGE would charge me $0.51/kWh peak but since I'm pushing it back into the system, that is what they pay me. With all the variables in the mix, I calculated a 10.5 yr payback; my $3000+ annual electric bill is now ~$70/year.

Keep the coal and nuclear plants running for dependable baseline power and use PV to trim the peaks.

Wm. Wangard said at May 5, 2008 8:05 PM:

Yes, if the government restricts the production of electricity by conventional means: coal, nuclear, and hydroelectric, then the cost of solar just might become competitive.

Engineer-Poet said at May 8, 2008 7:31 PM:

Hydroelectric is maxed out, and coal is deteriorating in quality even as remaining seams become harder to get to.  Meanwhile, solar and wind follow a steadily decreasing cost curve.  Nuclear is the only conventional energy source with lots of room for growth.

Scott said at May 29, 2008 5:20 PM:

For the extreme example, look at current Hawaii electric costs. The total charge per KWH in Hawaii (from my May 08 electric bill in Kona, HI) was 36.6 cents per KWH and that does not include those fees which are not tied to usage levels. As virtually all the electricity on the island is generated with oil, I don't anticipate it's going to be going down much in the future.

Randall Parker said at May 29, 2008 8:29 PM:


Hawaii seems like the best place in the United States for solar. That use of oil for electricity is going to drive up your rates even higher in coming years. I would expect PV to pay back pretty quickly in Hawaii.

Post a comment
Name (not anon or anonymous):
Email Address:
Remember info?

Go Read More Posts On FuturePundit
Site Traffic Info
The contents of this site are copyright ©