Débat problématique énergétique / effet de serre / climat, etc.
2005 Solar Year-end Review & 2006 Solar Industry Forecast
Polysilicon Supply Constraint Limiting Industry Growth
by Jesse W. Pichel and Ming Yang, Research Analysts, Piper Jaffray
January 11, 2006

Overextended demand for raw silicon (pictured here) is expected to limit solar photovoltaic market growth in 2006-2007 worldwide.

Photo: Lara Skinner
    The rapid increase in solar cell production in 2004, thanks to expanding subsidies worldwide and rising Integrated Circuit (IC) unit volumes triggered a polysilicon shortage forcing solar manufacturers to pay higher prices to secure silicon supply in 2005. Going into 2007, the industry can expect more of the same.
    "We estimate that solar manufacturers met 80% to 90¨% of its 2005 production plans due to polysilicon stockpiles from 2001/2002, resulting in a 30% solar industry growth over 2004 to 1656 MW in 2005. But the picture is bleak for 2006 given that stockpiles are depleted..."
    2005 witnessed higher polysilicon costs: The contract price of polysilicon at $60/kg in 2005 doubled from $30/kg in 2003. For companies using traditional mono- or polycrystalline silicon wafers in modules (91% of industry), the polysilicon feedstock represents 25% of the module BOM (bill of material) in 2005. Despite higher prices, only 80-90% of planned production was met in 2005.
    For 2006, anticipate limited growth and margin degradation: For 2006 and into 2007 we believe the greatest risk to the solar industry remains the polysilicon shortage and resulting price increases that may limit growth and/or degrade margins beginning in 2H06. Only companies that have secured allocation can grow; only those that have fixed price contracts will maintain margins. 
The situation should intensify into 2007. Contract prices are anticipated to reach 80 per kg in 2007, and the spot price will remain over $100 per kg. Our supply chain checks confirm that polysilicon contracts are sold out through 2007. We anticipate that polysilicon feedstock will rise from 25% of BOM to 40% by 2007. Only 60-65% of planned production will likely be met.
    The shortage is most pronounced in 2006, and will cap solar industry growth at 5%: We estimate that solar manufacturers met 80% to 90% of its 2005 production plans due to polysilicon stock piles from 2001/2002, resulting in a 30% solar industry growth over 2004 to 1656 MW in 2005. But the picture is bleak for 2006 given that stockpiles are depleted -- we estimate only 13,000 metric tons of polysilicon will be available for solar cell production. Despite advances in technology that increases cell efficiency and reduced polysilicon use, the 13,000 metric ton translates to a mere ~1,500 MW of crystalline solar cell production. Thus, we believe the solar industry overall will only grow 5%in 2006 to ~1738 MW of total solar cell production. We have detailed our polysilicon feedstock production estimates with several polysilicon/wafer/cell manufacturers and industry consultants. All agreed with a realistic scenario of feedstock CAGR of ~12% through 2007.

     We have detailed our assumptions in the exhibit below:

Piper Jaffray Solar Industry Production Estimates, 2003-2010E

Source: Piper Jaffray Estimates

Polysilicon Background

    Approximately 94% of solar cells are manufactured using crystalline silicon as the primary raw material. For companies using traditional mono- or polycrystalline silicon wafers in modules (91% of industry), this is essentially the same ultra-pure silicon material used to manufacture ICs. Historically, the solar industry has purchased off-spec material that is rejected by the IC industry, as semiconductors require much higher purity silicon. However, as the solar industry has grown, its demand has surpassed the off-spec silicon production. As a result, the solar industry has been forced to buy IC grade silicon. Currently, SGS is the only producer of solar grade silicon in substantial volumes.     The polysilicon manufacturing process is highly capital intensive and requires investments of $200-$250 million for a 3,000 metric ton capacity that takes 24 months to ramp. Five major manufacturers constitute 88% of the world's polysilicon production. These are Hemlock, Tokuyama, Wacker, REC (subsidiary SGS and ASiMI), and MEMC. The world capacity is estimated at 30,000 metric tons in 2005.
    In 2004, about 65% of the polysilicon production was used to manufacture semiconductors, with the balance being consumed by solar cells. Due to the semiconductor down cycle in 2001 that saw polysilicon prices decline below cost to $24/kg, polysilicon manufacturers have been unwilling to add capacity without purchase agreements.


Source: Tokuyama
Polysilicon R&D and Capacity Expansion
    The polysilicon industry is enjoying record industry profits. Additionally, for the first time solar manufacturers are pre-paying for supply (thanks to recent IPOs) and thus funding poly capacity expansion that should eliminate the shortage in 2008. Wacker, Tokuyama, and REC have launched programs to develop processes for manufacturing granular silicon (fluidized bed reactor for Wacker and REC and vapor to liquid deposition (VLD) reactor for Tokuyama). Tokuyama is building a 200-ton half commercial VLD pilot plant in Japan, while Wacker already has a 100-ton FBR pilot plant in Germany. REC is also looking to build a 200-ton pilot plant in Moses Lake, WA. In terms of capacity expansion, Wacker is currently expanding its facility in Germany, Hemlock is adding 3,000 ton of capacity, Tokuyama is expanding 400 tons in Japan, while REC has a goal to increase SGS to 2,500 tons per year. However, most production will not come online until 2008.

The Raw Polysilicon Feedstock Manufacturing Process
    The process for making polysilicon feedstock is commonly referred to as the Siemens process using a CVD reactor and silane or trichlorosilane gas. The entire industry uses this CVD process with the exception of MEMC in Pasadena, Texas, which uses a silane fluid bed reactor that produces granular polysilicon. (REC at Moses Lake, WA and Wacker in Germany are both working on fluid bed reactors as is Schumacher Technology). 

Granular polysilicon, which fluid bed reactors produce, is desirable since it can be easily melted to top off the crystal growing crucible, allowing a longer silicon ingot crystal without the need to shut down the furnace. Furthermore, granular poly may enable innovations in high-speed, high-volume solar cell and module manufacturing.

The Case For "Virtual" Integration
    While PV manufacturers are accelerating manufacturing process cost improvements to mitigate rising raw material costs, we believe that the greatest cost improvement for the PV industry can be attained by ensuring a consistent, low-cost supply of polysilicon. We suggest an industry consortium that would mitigate risk in constructing new PV poly capacity. The latest manufacturing techniques for polysilicon production are fluid bed reactors including tribromosilane (SiHBr3) fluid bed reactors and continuous substrate fabrication such as the continuous melt replenishment (CMR) process. According to industry sources, a $200 million investment could generate 3,000 ton of Electronic Grade polysilicon per annum, and supply polysilicon at $20 per kilo. Furthermore, any excess production could be sold into the IC wafer supply chain. We believe that a select few solar wafer manufacturers will adopt a virtual integration approach by investing proceeds from recent financings. In our opinion, this will enable a sustained competitive advantage.