Economies of Scale

Studying the impact of large-area modules on costs

Area-based economies of scale have been demonstrated in the manufacturing of various technologies such as flat panel displays, coated glass for architectural applications and wafer-based semiconductor processes. While the vast majority of solar photovoltaic (PV) modules are  in the range of 1-2 square metres, several companies have attempted to leverage area-based economies of scale in order to reduce their cost. The development of an amorphous silicon (a-Si) SunFab module by Applied Materials is perhaps the most well-known example of such attempts. According to Applied Materials, the modules, which occupied an area of 5.7 square metres, reduced the installation cost of a PV system by more than 20 per cent. However, a few years later, the SunFab line was shut down.

It was unclear whether the struggles SunFab faced were related to a-Si technology, the large-area module format, or market conditions. Despite the fact that very large-area modules have not yet succeeded in the marketplace, interest in the concept has not faded. First Solar, a leading manufacturer of cadmium telluride (CdTe) modules, recently announced its plans to move towards much larger area panels, in order to reduce capex by nearly 40 per cent. Meanwhile, REEL Solar, Inc. has developed a process for electroplating on large areas to enable the manufacture of large CdTe modules. Siva Power, a start-up in copper indium gallium diselenide (CIGS) module manufacturing, is also developing large-area products covering 2 square metres, which are similar to 72-cell multicrystalline silicon (mc-Si) modules, but larger than other leading thin-film products.

In the US, the National Renewable Energy Laboratory (NREL), in one of its research papers, has provided an analysis of the costs of the three leading commercial PV technologies – mc-Si, CdTe and CIGS. It focuses on the rigid glass-glass module architectures used. This is because the per watt cost and the levellised cost of energy (LCOE) are strongly influenced by the performance of modules. It has also examined the potential effect of the module size on its efficiency and energy yield. The following are the key highlights of the research paper.

Analysis of mc-Si modules

NREL estimates that increasing the module size from the current standard 72-cell case to the Gen 10.5 glass size would decrease the manufacturing cost and minimum sustainable pricing (MSP) of a module by $0.035 per watt and $0.04 per watt respectively. About 74 per cent of this reduction would be on account of reduced material costs of the junction box, the potting agent and busbars. While the cost of these materials per module increases with an increase in module area, the rate of this increase is lower than the rate of increase in the module power rating. Since the cost per watt is equal to the cost per module divided by the watts per module, the cost per watt of materials is lower for larger modules. The impact of yield on material costs is small, since NREL’s model assumes a cumulative manufacturing yield loss of approximately 2 per cent, with very little yield loss occurring at the final stages of the manufacturing process.

Lower per watt equipment and facilities costs account for approximately 6 per cent of the total cost reduction. Total equipment costs depend on the cost per tool and the number of tools required to meet the factory throughput requirement (1 GW per year in this case), which is driven by the cycle time for each process. In many operations, little to no increase in cycle time occurs with an increase in the module area. The per tool cost of certain equipment increases with module area, but not linearly; the increase in per tool cost for module assembly equipment is expected to be relatively small. The lower tool count also brings down labour costs as only a certain number of line workers are required to handle one tool. However, the effect of low labour costs on the overall cost savings is small, assuming a highly automated process, which requires relatively unskilled, low-cost labour for module assembly. The electricity usage per module including the flash tester, busbar assembly, laminator and automation equipment increases with the module area, but again, not linearly.

About 66 per cent of the $0.035 per watt manufacturing cost savings are realised with Gen 8 glass. While additional savings could be achieved by using Gen 10 and Gen 10.5 glass, there are decreasing returns to scale. There are two reasons for this. First, for sizes above Gen 8, the per watt costs that scale with the module area (for example, the junction box) are significantly reduced and other costs that do not scale with the module area become more dominant. Second, the increase in auxiliary costs and junction box prices, as well as a very large cell count per module (399 for the Gen 10.5 case), blunts the cost savings. The potential logistics and shipping challenges associated with Gen 10 and Gen 10.5 products discussed above could also affect the overall cost competitiveness of these very large modules.

Analysis of CdTe modules

Since First Solar is the only high-volume manufacturer of CdTe modules, sufficient data for a bottom-up cost analysis of Gen 8, Gen 10 and Gen 10.5 sizes is not available. NREL has estimated the potential effect of the Gen 6 case on cost by using the available information on the S6 product.

If the reported capex is achieved, there could be reductions of $0.10 per watt in the manufacturing cost and up to $0.14 per watt in minimum sustainable price (MSP). Around 80 per cent of the total manufacturing cost savings come from reductions in material costs (46 per cent) and capex (34 per cent). As with mc-Si, the per watt material costs in CdTe modules are reduced because the cost of materials is either fixed or increases at a lower rate than the rated module power output. The reduction in junction box material costs is even more pronounced for CdTe as compared to mc-Si.

The decrease in capex with larger modules is significant – $0.4 per watt per year as against $0.66 per watt per year. Unfortunately, because sufficient data could not be obtained on the capex for each tool in the Gen 6 case, we have little insight into the main drivers of this reduction. It is likely that the mechanisms observed for mc-Si apply to CdTe as well – cycle time (per watt) decreases, thus reducing the number of tools required to meet the throughput requirement, while the price per tool increases at a slower rate than the throughput (watt per year) per tool. As the capex required for manufacturing CdTe modules is much higher than that required for assembling mc-Si modules, the potential for capex savings in the former case is higher.

Assuming that the annual maintenance costs of CdTe modules are a fixed percentage of the overall capex, lower capex would result in proportionally lower maintenance costs. Smaller savings have also been observed in labour costs because of the decrease in the number of equipment stations required to meet throughput requirements, as well as in electricity costs due to an increase in electricity usage.

Analysis of CIGS modules

CIGS modules hold the greatest potential for cost reduction in larger areas, with savings of up to $0.136 per watt in manufacturing costs and $0.147 per watt in MSP. There are three distinct steps in the CIGS manufacturing process. These require sputtering of the molybdenum back contact layer, the intrinsic and aluminium-doped zinc oxide front contact stack, and the copper, indium and gallium precursors – all of which lead to a reduction in the cost of ownership as the module area increases. These cost reductions pertain to equipment, electricity usage and labour requirements per watt, as well as the potential for increased substrate collection efficiency and material utilisation, which depends on the configuration of the modules in the tool and the tool design. Very large-area sputtering tools are already employed in the manufacture of some displays and architectural glass coatings. In addition to this, the same economies of scale, as described above for CdTe and mc-Si, exist in lamination, and  junction box and busbar attachment for CIGS modules.

As in the case of mc-Si modules, when the size extends beyond Gen 8, the cost of the junction box and busbars decreases (the former is reduced from $0.048 per watt in the reference case to $0.009 per watt in the Gen 8 case), and other costs that do not reduce with module area begin to dominate. Around 71 per cent of the manufacturing cost reduction is achieved with Gen 6 sizes, and 91 per cent savings can be realised with Gen 8 modules.

These cost savings are dependent on the ability to maintain the manufacturing yield and efficiency as the area increases. Currently, there is no data available on manufacturing yields for CIGS processes over very large substrate areas. Savings are most sensitive to differences in the selenisation process, including the price per tool and the throughput. According to NREL, for batch processes using currently available tools, throughputs that are at least similar to those of the reference case could be achieved, because a similar total module area could be packed into the batch furnace. However, sputtering and selenisation have not been demonstrated on these very large-area modules, and it is unclear whether modifications in the selenisation process will be needed to maintain good performance. Any process changes that may be required could affect the throughput at each step.

Effect of module area on BoS and total system installation costs

Based on the available interviews and data at NREL, the authors believe that the installation of large modules would require a machine to assist in the lifting and placing of modules. In NREL’s model, it is assumed that the cost of this machine is equal to that of a standard crane truck used in construction plus a 20 per cent premium for robots, end effectors, etc. to interface with the module. The results assume that machine-assisted module mounting takes almost the same amount of time as manual module mounting. If this process could be speeded up, additional cost savings could be realised. However, the labour costs associated with module mounting constitute a small fraction of the overall labour costs, so these savings are likely to be modest.

Material cost savings are observed by replacing typical clamps for glass-glass modules, which cost around $2.65 per module, with flanges and adhesives that are estimated to cost $0.80 per module. The reduced module count and modified system architecture also result in labour and structural/electrical balance of system (BoS) savings. The reduced module count results in a significant proportional reduction in the module mounting time, but because much of the labour resource is spent on other tasks such as installing the structure, racks and the inverter, the savings as a percentage of the total labour costs are modest. There is also a small additional cost associated with the use of an additional machine to assist in large module placement.

Decreasing returns to scale are observed for both mc-Si and CIGS modules beyond Gen 8, but some additional module cost savings are still realised moving from Gen 8 to Gen 10 glass size. Increasing the size from Gen 10 to Gen 10.5 results in less than $0.005 per watt of additional savings.

Potential effect of increased module size on energy yield

Based on interviews with industry experts in module performance and reliability, the authors believe that the performance of standard mc-Si modules will not be significantly affected by size. With modules employing cells that were discarded prior to assembly, there may be uneven degradation of cells. There may also be greater variations in the output of the various cells within a larger module due to partial shading. The effect of these factors on energy yield, if observed, can be mitigated in wafer-based module architectures by rewiring the modules and employing additional bypass diodes, or module-level power electronics.

The module area might affect the energy production of monolithic thin-film modules in many more ways. First, nameplate efficiency could be affected if non-uniformities in the manufacturing processes over large areas are not addressed. The process employed for the vast majority of CIGS and CdTe manufacturing does not allow for cell binning, which reduces the ability to mitigate the effect of any cell-to-cell non-uniformities. It is currently unclear how difficult it will be to achieve high uniformity over large areas with the vapour phase transport process for CdTe and the selenisation process for CIGS. In addition, the non-uniform cell output as the result of manufacturing non-uniformities or temperature non-uniformities in the field may reduce the energy yield. In the current monolithic architectures, this cannot be mitigated with the use of bypass additional diodes.


The exact manufacturing cost savings realised will depend on the steps and factory layout involved, whether large modules are manufactured in a new facility or in an upgraded, existing facility, and what equipment is used. Savings achieved with large-sized modules will also depend on the price of materials at any given point in time. For Gen 10 and Gen 10.5 glass, significant challenges related to logistics and shipping will need to be addressed in order to achieve low costs.

The effect of module area on efficiency and energy production, especially on thin-film modules, is not well understood. Further research is required to verify the effects of module area on energy output and thus LCOE.

There may be other barriers, apart from cost per watt, to manufacturing very large modules. For thin-films in particular, a very high capex would be required to build up the necessary capacity for manufacturing large modules at a competitive scale, given the sizable manufacturing capacity that already exists for mc-Si modules. In the highly competitive PV market, smaller capex investments such as those in upgrading existing mc-Si lines are preferred, making it difficult to invest significant amounts in research and development, which encourages relatively short-term thinking. Meanwhile, downstream suppliers may be slow to adopt large-area modules if it requires additional investment at their end. With the increase in module size, new investments would be required from equipment manufacturers to develop new tools and processes suitable for large areas. In some cases, technology can be borrowed from other industries, as with sputtering equipment currently used in display manufacturing. n

Based on NREL’s paper, “An Analysis of the Cost and Performance of Photovoltaic Systems as a Function of Module Area”, by Kelsey A.W. Horowitz, Ran Fu, Xingshu Sun, Tim Silverman, Mike Woodhouse and Muhammad A. Alam


Enter your email address