1 Innovation and electricity generation of wind AND solar PV in ChinaLong T. Lam, Ph.D. Technological Change & Entrepreneurship Carnegie Mellon Portugal Program May 2, 2017 Advisors: Prof. Inês Azevedo Prof. Lee Branstetter (Heinz College)

2 Motivation: The Top 6 Carbon EmittersCO2 emissions from fossil-fuel use and cement production in the top 5 emitting countries and EU [1:45] Let me begin by putting Chinese carbon emissions in perspective, using this graph from the European Commissions official 2016 report on global trends in carbon emissions. This graph measures CO2 emission for China, the USA, Western Europe, Russia, India, and Japan. China’s trend line is the most striking feature in this graph. China is now responsible for almost 30% of global emissions. Its annual emissions have more than doubled since the early 2000s, and they are now twice as high as US emissions, and they are more than America’s emissions and Western Europe’s emissions put together. In the fight against global warming, no place is more important than China. In the past decade, the Chinese government concertedly turned its focus to these environmental problems. China has emerged as a powerhouse in the production and deployment of renewable energy hardware. China’s wind turbine installed capacity eclipsed that of the US. Its manufacturing base is the largest in the world. China is responsible for the majority of the solar cell production in the world, and recently surpassed Germany as the country with the highest installed solar PV capacity. [END] EDGAR v4.3.2 FT2015(JRC/PBL 2016: IEA 2014 (suppl. With IEA 2016 for China, BP 2016, NBS 2016, USGS 2016, WSA 2016, NOAA 2016)

3 Agenda Research MotivationChina’s wind power industry: patents and learning rate* Expert elicitation for China’s solar PV industry Grid integration of China’s renewable energy Policy implications and conclusions Acknowledgements Assess China’s wind turbine manufacturing industry’s innovation capabilities and technological progress Examine the successes and shortcomings of China’s policy efforts to integrate renewable energy sources and to transition to a more sustainable energy system Analyze the global solar PV supply chain with a focus on Chinese firms’ innovation capabilities and their role in the global PV ecosystem *: This chapter has been accepted for publication in Energy Policy as Lam, L.T., Branstetter, L., Azevedo, I.M.L., China’s wind industry: leading in deployment, lagging in innovation. Energy Policy 106,

4 China’s Wind Turbine IndustryExplosive growth in China’s wind energy capacity installed capacity increased more than 100 times; cumulative installed capacity in 2016 was 169 GW. Sources of competitiveness: government support and industrial policies; technology transfer; learning; substantial indigenous innovation

5 Chinese Wind Power PatentsPatents granted by different offices China granted thousands of patents Significant fraction to Chinese inventors Chinese inventors obtained small number of international patents: 16 from EPO and 91 from USPTO Leading Chinese firms secured few international patents Chinese international patents are less likely to be cited relative to Western counterparts No. of Patents (Total) No. of Patents (Annual) Patents granted by EPO Exclude utlity and design patents. Inventor’s nationality by address. 16 EPO patents or 0.9% 91 USPTO patents of 1.6% No. of Patents (Total) No. of Patents (Annual)

6 Learning Rates for China’s Wind IndustryDecrease in capital cost drives learning 4.5% learning rate using unit capital cost 4.1% learning rate using LCOE Moderate relative to historical learning rates Denmark 8.8% learning from (100x capacity increase) Germany 12% learning from (60x capacity increase) China is a late-comer in wind turbine sector Technology was adopted from abroad Limited space for technological improvement Use CDM Data

7 Agenda Research MotivationChina’s wind power industry: patents and learning rate Expert elicitation for China’s solar PV industry Grid integration of China’s renewable energy Policy implications and conclusions Acknowledgements Assess China’s wind turbine manufacturing industry’s innovation capabilities and technological progress Examine the successes and shortcomings of China’s policy efforts to integrate renewable energy sources and to transition to a more sustainable energy system Analyze the global solar PV supply chain with a focus on Chinese firms’ innovation capabilities and their role in the global PV ecosystem

8 China’s Solar PV IndustryIndustry is similar to others in manufacturing sector Tech know-how from turn key production lines (de la Tour, 2011) Focus on traditional multicrystalline silicon (mc-Si) technology Small number of international patents Reports of innovation throughout the industry Sharp drop in cell and module production costs and product prices Adoption of new processes and materials Trina mc-Si cell has world’s highest efficiency (Green et al., 2016) [2:00]At first glance, the story of China’s solar PV industry has similar elements to other manufacturing industries in which China has become a leader. The technological barriers to entry in the downstream segments of the solar PV value chain – manufacture of PV cells and modules – were relatively low. Turn-key production lines for cell manufacture could be bought from the capital equipment manufacturers and run with little prior experience. Equipment was not inexpensive, but the local government’s support for these local champions translates to easy access to capital. Assembling modules out of cells and other components were relatively straightforward in the solar PV manufacturing process. In other words, there is little room for Chinese producers to innovate. Number of international patents low. On the other hand, there are reports and stories that extol the innovation capabilities of Chinese firms. Kelly Gallagher discussed in her work that skilled executives from brought home with them the savvy tech know-how, and Popp and his co authors share similar conclusions. There were also partnerships with foreign research institutes. The number of domestic patents has increased in this period. Strikingly, the level and rate at which solar cell and module prices have fallen took many by surprise. At the same time, new innovative processes introduced both in the upstream and downstream promise a further decline in costs, and some Chinese firms are active participants. Martin Green and his associates keep track of the highest efficiency of various solar PV products, and they report that Trina’s polycrystalline cell has the world’s highest efficiency (21.3%)

9 Expert Elicitation In retrospect Future prospectsRank the importance of different components in reducing c-Si module and system costs Identify important technological and non-technological factors Future prospects Estimate efficiency and cost of utility-scale c-Si, thin-film PV, concentrating PV, organic PV, and emerging technologies by 2030 Estimate cost for utility-scale c-Si PV systems in China 16 experts from industry, academia, and national labs All but three are Chinese nationals Most experts have had professional experience outside of China

10 Important Factors Technology adoption and improvement throughout the supply chain Drop in polysilicon price was critical Some advances were indigenous, e.g. seed-assisted growth method Non-technological factors were critical as well “Market formation policies” around the world: FIT, RPS, net-metering laws, ITC (Gallagher, 2014) Economies of scale, agglomeration effects, learning-by-doing, human capital mobilization, vertical integration (Yu et al., 2011; Goodrich et al., 2013; Luo et al., 2013, Gallagher, 2014) Open and modular nature of c-Si PV

11 2030 C-Si Module Cost Explain what the three technologies areEfficiency Mono-Si 21.2%, 4% higher than today mc-Si: 20.3% Novel: 23%, some experts expect that in the most optimistic scenatrios, novel modules can be close to the theoretical maximum Estimations are fairly consistent with some exceptions. Experts F and K expect significantly higher efficiency for mono-Si and mc-Si modules, while most experts see limited prospects for this development Cost Mono-Si module costs range from $0.16/W to $0.46/W, with an average of the best estimates around $0.27/W, compared to PRICE 64 cents/W in early Currently, it’s decreased by 30% to 41 cents. This is due to oversupply. Mc-Si 24 cents/W, 15 cents/W lower than current price. Novel: 30 cents/ W The estimated ranges are more narrow than previous studies (Curtright et al 2008). Efficiencies for silicon PV modules have improved over the past decade, but the theoretical efficiency ceiling remains unchanged. Similarly, decrease in module cost production was dramatic, but future system cost reductions will depend more on non-module components.

12 2030 C-Si PV System Cost 120 100 80 60 Cost, US ¢ /Watt 40 20novel Cost, US ¢ /Watt mono multi A B D F G H I K L O

13 Agenda Research MotivationChina’s wind power industry: patents and learning rate Expert elicitation for China’s solar PV industry Grid integration of China’s renewable energy Policy implications and conclusions Acknowledgements Assess China’s wind turbine manufacturing industry’s innovation capabilities and technological progress Examine the successes and shortcomings of China’s policy efforts to integrate renewable energy sources and to transition to a more sustainable energy system Analyze the global solar PV supply chain with a focus on Chinese firms’ innovation capabilities and their role in the global PV ecosystem *:This chapter has been published as Lam, L.T., Branstetter, L., Azevedo, I.M.L., China's wind electricity and cost of carbon mitigation are more expensive than anticipated. Environ. Res. Lett. 11, 1–11.

14 China’s Electricity Generation from RenewablesChina installed more wind turbines than the U.S. but generated much less electricity. Pervasive grid connection and curtailment problems [1:30] Define curtailment While working on the first study, I noticed something interesting. While China’s installed capacity surpassed that of the US in 2010, the US turbines continue to churn out more electricity. In 2014, when the latest data is available, China’s wind base had as much as 115GW of installed capacity, 75% more than the US, but it only generated 156TWh. By comparison, the US generated 180 TWh in the same year. What explains the gap? A high fraction of China’s turbines are not connected to the grid. And a high fraction of connected turbines are unable to sell their electricity to the grid due to curtailment. Signs that indicate the severity of China’s grid connection and curtailment problems began to emerge in the late 2000s and these issues persist until today. By my estimate, if China were to connect all of its wind turbines and place them in use at the same rate as 2014, it could generate almost 40% more electricity from wind, the equivalent of installing about 32 GW capacity. So I am in interested in learning more about the successes and shortcomings of China’s efforts to integrate renewables into its grid system. [END]

15 Measuring Performance of China’s Wind TurbinesCapacity factor Ex-ante, estimated in CDM Project Design Document (CF ex-ante) Ex-post using actual generation and cumulative installed capacity (CF ex-post, installed) or cumulative grid-connected capacity (CF ex-post, connected) Utilization factor (UF): portion of hours turbines are in use in a year Often included in national reports Levelized cost of electricity (LCOE) and Cost of Carbon Mitigation (CCM) are estimated under the same four scenarios (CF ex-ante; UF; CF ex-post, installed; CF ex-post, connected)

16 Low Capacity Factors Large discrepancies between estimated and actual performance

17 Levelized Cost of ElectricityLCOE is one-half to two times more expensive than estimated Gaps have narrowed in recent years [1:30] Here are the results. I find that wind curtailment has been a persistent problem in China. Provinces with high wind penetration rates have most difficulties in integration. While there were signs of improvement in 2014, last year data show that the problem is far from over, with as much as 20% of the electricity produced by wind turbines was curtailed. This is not a good sign as China pushes forward with its ambitious target for solar and wind. China’s wind capacity factors were lower than what developers anticipated, and they exhibit a downward trend. LCOE and CCM are much higher than anticipated. I find that the learning rate is around 3.5%-4.5%, which is low relative to the historical learning rate. [END]

18 Cost of Carbon MitigationCCM is four to six times higher than anticipated CCM is defined as the difference in LCOE of the base fuel and wind divided by the emissions factor.

19 Conclusions and Policy Implications (1)Remarkable improvements with few fundamental breakthroughs Unprecedented production level through economies of scale and learning “Platform for production development” (Nahm and Steinfeld, 2014) Policies are important in creating demand Dependency on policy support will continue in the near future Challenges ahead Overcapacity; short-term focus on profit; delays in advanced cell adoption Heavy focus on c-Si at the expense of other PV technologies The standard Al-BSF design had been invented decades earlier, and physics principles behind photovoltaic much earlier. Advanced cell designs like PERT, IBC, and HIT predated China’s entrance to the PV market A prominent PV researcher remarked that Chinese PV makers “have made small if any actual process cost reductions at Southwest Airlines margins” As long as a new innovation stays within the silicon paradigm, a firm has to rely on China’s manufacturing infrastructure as a “platform for product development”(Nahm and Steinfeld 2014). Equipment manufacturers need to design new “drop-in” machines that are compatible to existing production lines. In this sense, China has essentially become the test bed for new silicon PV technologies. While these policies are not identical in their goals, they all have created substantial demand, creating market conditions in which solar PV can compete with traditional energy sources. Solar PV’s dependency on support policy will continue in the near future in China, and likely elsewhere around the world. However, it may be some years before solar PV can directly compete with coal on a cost basis, especially when taking into account integration costs.

20 Conclusions and Policy Implications (2)China’s experiences can offer policy insights to India and other developing economies Long-term policy and financial commitment Comprehensive planning: installation, generation, and distribution Tension in dual goals of deployment and employment in manufacturing Opportunities in energy storage and micro-grids Trade is not a zero-sum game Tariffs hurt Chinese PV makers and U.S. polysilicon producers U.S. customers and PV installers benefit from low prices Tariffs have not necessarily enhanced competitiveness of U.S. PV makers Chinese firms have entered all stages of the supply chain, producing most of the installed solar modules around the world. Meanwhile, production costs are at record lows continued declines in module and system costs coupled with steady efficiency gain will allowing solar photovoltaic to be competitive with traditional energy resources like coal in China will remain the mainstream product for large-scale electricity generation application in the near future, though the industry’s overinvestment, overcapacity, and singular short-term focus on silicon PV may create an opening for emerging technologies

21 Acknowledgements Profs. Inês Azevedo, Lee Branstetter, Kelly Sims Gallagher, Francisco Veloso Participating PV experts Dr. Ahmed Abdullah (UCSD) Ana Paola Giordano, Tatiana Marques, António Moreira (CLSBE) Prof. Granger Morgan (CMU) Prof. Joseph Yuan, Sun Shuang (Tsinghua University) Dr. Robert Margolis (NREL) Prof. Sally Xu (Peking University) Dr. Yang Liu (IEA) Carnegie Mellon Portugal Program Portuguese Foundation for Science and Technology AWEA Blakemore Freeman Foundation CEDM, CEIC, Scott Energy Institute (CMU) CMU Library First China-US PV Youth Forum NSF East Asia Pacific Summer Institute SPPM (Tsinghua University)

22 References (1/2) Anadon, L.D., Bunn, M., Chan, G., Chan, M., Jones, C., Kempener, R., Lee, A., Logar, N., Narayanamurti, V., Transforming US energy innovation. Harvard Kennedy School. Arrow, K., The economic implication of learning-by-doing. Review of Economic Studies 29, 155–173. Baker, E., Bosetti, V., Anadon, L.D., Henrion, M., Reis, L.A., Future costs of key low-carbon energy technologies: Harmonization and aggregation of energy technology expert elicitation data. Energy Policy 80, 219–232. doi: /j.enpol Bettencourt LMA, Trancik JE, Kaur J, Determinants of the pace of global innovation in energy technologies, PLoS One, 2013, Vol. 8, e67864 Bosetti, V., Catenacci, M., Fiorese, G., Verdolini, E., The future prospect of PV and CSP solar technologies: An expert elicitation survey. Energy Policy 49, 308–317. doi: /j.enpol Branstetter, L., Li, G., Veloso, F., The rise of international co-invention, in: Jaffe, A.B., Jones, B.F. (Eds.), The Changing Frontier: Rethinking Science and Innovation Policy. University of Chicago Press. Davidson, M., Politics of Power in China: Institutional Bottlenecks to Reducing Wind Curtailment Through Improved Transmission. International Association for Energy Economics (IAEE) Gallagher, K.S., The globalization of clean energy technology – Lessons from China. MIT Press. Print. Gallagher, K.S., Zhang, F., Climate technology & development case study: Innovation and technology transfer across global value chains: Evidence from China's PV industry Goodrich, A., Powell, D.M., James, T.L., Woodhouse, M., Buonassisi, T., Assessing the drivers of regional trends in solar photovoltaic manufacturing. Energy Environ. Sci. 6, doi: /c3ee40701b Green, M.A., Emery, K., Hishikawa, Y., Warta, W., Dunlop, E.D., Solar cell efficiency tables (version 48). Prog. Photovolt: Res. Appl. 24, 905–913. doi: /pip.2788 Inman, M., How Low Will Photovoltaic Prices Go? An Expert Discussion. Kang, J., Yuan, J., Hu, Z., Xu, Y., Review on wind power development and relevant policies in China during the 11th Five-Year-Plan period. Renewable and Sustainable Energy Reviews 16, 1907–1915. doi: /j.rser Kahrl, F., Wang, X., Integrating Renewable Energy Into Power Systems in China: A Technical Primer – Power System Operations. Regulatory Assistance Project, Beijing. la Tour, de, A., Glachant, M., Ménière, Y., Innovation and international technology transfer: The case of the Chinese photovoltaic industry. Energy Policy 39, 761–770. doi: /j.enpol Lewis, J., Green Innovation in China: China's Wind Power Industry and the Global Transition to a Low-Carbon Economy. Columbia University Press, New York.

23 References (2/2) Li C.B., Li P., Feng, X., Analysis of wind power generation operation management risk in China. Renewable Energy –75 Li, X., Hubacek, K., Siu, Y.L., Wind power in China - Dream or reality? Energy 37, 51–60. doi: /j.energy Liu, Y., Kokko, A., Wind power in China Policy and development challenges. Energy Policy 38, 5520–5529. doi: /j.enpol Luo, S., Lovely, M.E., Popp, D., Intellectual Returnees as Drivers of Indigenous Innovation: Evidence from the Chinese Photovoltaic Industry. NBER Working Paper. Morgan, M.G., Henrion, M., Small, M., Uncertainty: A Guide to Dealing with Uncertainty in Quantitative Risk and Policy Analysis. Cambridge University Press. Nahm, J., Steinfeld, E.S., Scale-up nation: China's specialization in innovative manufacturing. World Development 54, doi: /j.worlddev Olivier, J.G.J., Janssens-Maenhout, G., Muntean, M., Peters, J.A.H.W., Trends in global CO2 emissions. PBL Netherlands Environmental Assessment Agency and Joint Research Centre. Pei, W., Chen, Y., Sheng, K., Deng, W., Du, Y., Qi, Z., Kong, L., Temporal-spatial analysis and improvement measures of Chinese power system for wind power curtailment problem. Renewable and Sustainable Energy Reviews 49, 148–168. doi: /j.rser Qiu, Y., Anadon, L.D., The price of wind power in China during its expansion: technology adoption, learning-by-doing, economies of scale, and manufacturing localization. Energy Economics 34, 772–785. doi: /j.eneco Ru P, Zhi Q, Zhang F, Zhong X, Li J, Su J (2012) Behind the development of technology: The transition of innovation modes in China’s wind turbine manufacturing industry. Energy Policy 43(C), 58–69. doi: /j.enpol Schuman, S., Lin, A., China’s Renewable Energy Law and its impact on renewable power in China: Progress, challenges and recommendations for improving implementation. Energy Policy 51, 89–109. doi: /j.enpol Tang, T., Popp, D., The Learning Process and Technological Change in Wind Power: Evidence from China's CDM Wind Projects. NBER working paper. Wang, F., Yin, H., Li, S., China’s renewable energy policy: Commitments and challenges. Energy Policy 38, 1872–1878. doi: /j.enpol Yao, X., Liu, Y., Qu, S., When will wind energy achieve grid parity in China? – Connecting technological learning and climate finance. Applied Energy 160, 697–704. doi: /j.apenergy Yu, C.F., Van Sark, W., Alsema, E.A., Unraveling the photovoltaic technology learning curve by incorporation of input price changes and scale effects. Renewable and Sustainable Energy Reviews 324–337. doi: /j.rser Zhao, Z.-Y., Chang, R.-D., Chen, Y.-L., What hinder the further development of wind power in China?—A socio-technical barrier study. Energy Policy 88, 465–476. doi: /j.enpol

24 Additional Slides

25 Wind patents by producersFirm Year Founded Ownership structure 2015 Cumulative Capacity (MW) EPO USPTO PCT/WIPO Apps Patents Foreign Goldwind 1998 SOE on stock exchange 31130 6 1 10 5 15 7 3 Sinovel 2006 16240 21 22 9 Guodian United Power 2007 SOE 14450 2 Dongfang 2004 10660 Mingyang Public 10110 Shanghai Electric 7330 XEMC Windpower 7040 19 4 Envision Private 6890 38 72 28 11 CSIC Chongqing 5300 Windey (Yunda) 2001 4160 Total 85 108 35 58 25

26 Main Patenting Routes

27 Poisson Model 𝜇 𝑖 is estimated from observed characteristics:𝜇 𝑖 =𝑒𝑥𝑝 𝑥 𝑖 𝛽 , 𝑖=1,…,𝑁 The log-likelihood is: 𝑙𝑛𝐿 𝛽 = 𝑖=1 𝑁 { 𝑦 𝑖 𝑥 𝑖 ′ 𝛽− exp 𝑥 𝑖 ′ 𝛽 −𝑙𝑛 𝑦 𝑖 ! } We assume that the observed count for observation i is drawn from a Poisson distribution with mean 𝜇 𝑖 , and In our case, these characteristics include the patent’s grant year and its nationality. The Poisson maximum likelihood is the solution to the nonlinear equations corresponding to the first-order condition for maximum likelihood.

28 Negative Binomial However, the Poisson distribution assumes equidispersion, or equality of mean and variance. Citation frequency data often exhibit overdispersion, and we can adjust for this by using a negative binomial regression model, which corrects the overdisperson by incorporating an error term u that follows a gamma distribution NB: which the sample variance is larger than the sample mean where mu > 0 is the mean of Y and a lpha> 0 is the heterogeneity parameter The values of a and b that maximize ln LHa, bL will be the maximum likelihood estimates we seek,

29 Results: Patent Citation Analysis NB Poisson Germany 2.322*** 2.300*** 2.157*** 2.168*** 2.239*** 2.221*** (0.297) (0.292) (0.284) (0.305) (0.298) Japan 2.256*** 2.251*** 2.379*** 2.353*** 2.043*** 2.052*** (0.303) (0.326) (0.323) (0.287) US 3.009*** 3.078*** 3.223*** 3.234*** 2.943*** 2.979*** (0.392) (0.403) (0.430) (0.433) (0.407) (0.411) Denmark 1.658*** 1.671*** 1.674*** 1.696*** 1.577*** 1.604*** (0.203) (0.204) (0.210) (0.212) (0.206) (0.208) ROW 2.530*** 2.554*** 2.548*** 2.592*** 2.387*** 2.433*** (0.334) (0.339) Constant 0.000 0.140*** 0.137*** 0.109*** 0.110*** (0.024) (0.021) Year Dummies Y Exposure Observations 3328 2700 2748 Pseudo Log- likelihood Next I want to see what is the likelihood of a Chinese patent being cited compared it its foreign counterparts. I use NB and Poisson model, with Chinese patents as the base group. Use WIPO data for citation analysis to try to overcome home country bias. There are 156 patents whose first inventors are Chinese nationals. Between 1980 and 2014, the likelihood of a Chinese wind turbine patent being cited by subsequent patents is less than that of a German, Japanese, Danish, or American patent, and this trend is significant and robust. For example, German wind patents are associated with approximately 2.3 times higher citation rate than Chinese patents, and U.S. patents are three times more likely to be cited than Chinese patents. The results are consistent and robust for both models and across different time specifications We include year fixed effects to account for the fixed differences in the number of citations across the patent year cohorts and a time exposure term to account for the time elapsed since a patent was first published

30 Results: Learning Rate 12 11 10 9 8 7 6 5 Capital Cost (mRMB/MW) Variable (1) (2) (3) Cumulative Capacity -0.051*** -0.060*** -0.066*** (-0.012) (0.008) (0.007) Plant’s load factor -0.607*** (0.036) Constant -0.387*** -1.213 2.527*** (-0.131) (0.099) (0.074) Year Effect Y Province Effect R-Squared 0.613 0.716 0.604 Observations 1477 *** p<0.01, ** p<0.05, * p<0.1 0.36 0.34 0.32 0.3 0.28 0.26 0.24 0.22 0.2 0.18 0.16 Capacity Factor Capital cost in 2004 currency value Capacity Factor as defined by Reported are learning parameter

31 Learning Rate EstimationCt = aNtα 𝐿𝑒𝑎𝑟𝑛𝑖𝑛𝑔 𝑟𝑎𝑡𝑒= 𝐶 𝑡 − 𝐶 𝑡2 𝐶 𝑡 ↔ 𝑎( 𝑁 𝑡 𝛼 − 𝑎(2 𝑁 𝑡 𝛼 𝑎( 𝑁 𝑡 𝛼 ↔1− 2 𝛼 𝐶 𝑡 and 𝑁 𝑡 are unit costs of wind turbine and cumulative installed capacity at time t Take log of first equation => alpha is the coefficient in the econmetrics results

32 Global Solar PV Installation[1:15] This figure shows the global solar PV installation. China’s cell production ranked number 1 in the world in 2007, but in 2007 the Chinese market for PV was virtually non-existent. The center of the PV global market was Europe. The effect of the Eurozone crisis is clear between 2011 and 2012, where the previous exponential growth rate came to a halt. But this also was the turning point for the Chinese market, where domestic demand picked up the slack left by Europe. The government previously announced that the target installed capacity is 10GW by 2015, 35GW by This target has been revised upward several times. (now: 110GW by 2020). Currently China has 77GW installed capacity, stripping Germany of its status as the world leader. SolarPower Europe, 2015

33 Results: Technological FactorsStage Key Factors Polysilicon Investment and scaling up of production plants; hydrochloronation technology upgrade; increase number of seed rods in furnace; reduction in electricity use; investment in FBR technology Ingot/Wafer Seed-assisted growth method using crystalline Si and quartz; diamond wire sawing; larger furnace and larger ingots; black Silicon; direct wafer Cell Improved efficiency; improved silver paste recipe; efficiency use of silver paste; higher number of busbars; high-efficiency cells (PERC/L/T, IBC, HIT) Module Domestic production and reduction of material use of key components (EVA sheets, glass, backsheets); replacement of TPT backsheets Equipment Indigenization of equipment for Al BSF; automation; gradual domestication of key equipment for high-efficiency cells Diamond wires, an alternative wafer slicing technology, can address a number of issues inherent to multi-wire slurry sawing. There is less kerfloss. Research has been demonstrated that diamond wires, which are coated with abrasive diamond grits by resin bonding, can saw wafers at a higher speed than slurry sawing, and the wafer’s damaged layer is thinner. slice a kilogram of silicon into 60 wafers, compared to steel wire’s 51, delivering substantially more efficient production and effectively decreasing the module cost about around 2.6 cents/W. Large mono-Si wafer makers have all adopted diamond wires in their new lines, and some are still retrofitting their slurry sawing equipment with diamond wires in existing factories.

34 Results: Technological FactorsStage Key Factors Polysilicon Investment and scaling up of production plants; hydrochloronation technology upgrade; increase number of seed rods in furnace; reduction in electricity use; investment in FBR technology Ingot/Wafer Seed-assisted growth method using crystalline Si and quartz; diamond wire sawing; larger furnace and larger ingots; black Silicon; direct wafer Cell Improved efficiency; improved silver paste recipe; efficiency use of silver paste; higher number of busbars; high-efficiency cells (PERC/L/T, IBC, HIT) Module Domestic production and reduction of material use of key components (EVA sheets, glass, backsheets); replacement of TPT backsheets Equipment Indigenization of equipment for Al BSF; automation; gradual domestication of key equipment for high-efficiency cells Early Chinese PV entrants purchased turnkey production lines from Western equipment makers, who in turn trained local employees to operate the machines. First, equipment is sold as-is, with each machine a technical black box. Should a piece of equipment fail, the provider could take months to acknowledge and address the issue, time that PV makers could not afford. Second, the rush of PV investment due to burgeoning demand in the 2000s created a backlog of orders for incumbent Western equipment makers. Third, international equipment was expensive. Chinese PV makers, especially smaller ones, were more willing to purchase domestic equipment. Investment costs for setting up a production line have dropped significantly. An expert from a leading Chinese equipment maker estimated that a 25 MW production line in the mid 2000s costed million yuan RMB to set up, but currently a production line with the same capacity costs about 40million yuan RMB.

35 Thin Film TechnologiesThin film technologies most promising and can challenge silicon on efficiency and cost Amorphous Si already ”out” Cadmium Telluride (CdTe) and Copper-Indium-Gallium-Selenide (CIGS) most viable Continue to improve in efficiency and cost, though China not very active in this technology [39] [0:45] [43:30] Experts state that silicon PV will continue to be the mainstream electricity supply technology for at least the next ten years. Some experts expect that demand for non-silicon PV technologies will remain relatively small, but others offer a more positive outlook, believing demand will rise. Differences in technological characteristics and application requirements will result in wide segmentation of PV products. Furthermore, China may continue to be a strong player in silicon cell and panel production, but the development of non-silicon technologies will be geographically diverse. Among non-Silicon technologies, thin film is most promising technology for large scale electricity generation application. Efficiency for cadmium telluride (CdTe) modules will reach 22% by 2030 with an average production cost of $0.27/W. Lab record: 18.6% For comparison, modules made by the world’s largest CdTe maker, First Solar, have a 16.8% efficiency and cost $0.40/W Copper-indium-gallium-selenide (CIGS) modules will achieve lower efficiency and cost reduction than CdTe. CIGS technology may not be able to compete with silicon-based PV on a cost basis, but CIGS modules are lightweight and highly flexible, allowing them to compete in other market segments. Efficiency: 23% (vs. 13.8%). Record 17.5%. Cost 29.9 cents/W. Solar Frontier, the largest CI(G)S incumbent, has a target production cost of 42 US¢ per Watt by 2017.

36 Concentrator Photovoltaic (CPV)General skepticism toward CPV’s future viability Collapse of polysilicon price makes low concentrator PV less attractive High concentrator PV uses high-efficiency multi-junction cells, but high system costs May make sense in sunny region close to big load center with high electricity price One expert did not see commercial viability for CPV systems and declined to provide CPV’s future costs and efficiency. The collapse in polysilicon price made low concentrator photovoltaic (LCPV) system less attractive. because the precipitous fall in pricing of traditional silicon PV panels but challenges in tracking and alignment mechanisms as well a that CPV systems could be competitive in sunny regions that are close to big population centers with high electricity price. For example, CPV systems could be installed in North Africa providing electricity across the Mediterranean to Europe high system costs remain

37 Excitonic and PerovskiteContinued gain in efficiency for dye-sensitized solar cells and organic PV Reliability issues Perovskite most promising among emerging technologies Efficiency up six times since introduction Would not be commercially ready in the near future

38 Trade Dispute: AD/CVD Tariffs TimelineSeptember 9, 2016 Trade Dispute: AD/CVD Tariffs Timeline 10/2011: SolarWorld filed complaint with the USDOC and USITC Focus on the US and China relationships, for sake of time 3 rounds: 2 rounds of Chinese modules sold in the US Chinese tariff on US polysilicon sold in China 2012 duties, focusing on products with Chinese cells, mostly range from 23-34% 10/2012: Final ruling from DoC for Chinese cells (23.75%– 254.66%)

39 Trade Dispute: Effects of US First Round TariffsSeptember 9, 2016 Trade Dispute: Effects of US First Round Tariffs Most major Chinese panel makers are still able to supply the US profitably “It’s not what we like to do, and it does impact our margin,” Canadian Solar chief executive Shawn Qu said in November. Source: Yahoo! Finance Final USDoC ruling: Oct 2012

40 Trade Dispute: AD/CDV Tariffs TimelineSeptember 9, 2016 Trade Dispute: AD/CDV Tariffs Timeline 12/2013: US companies filed second petition Final DOC tariffs issued in Dec. ‘14 range from 11-30% for Taiwanese companies and 75-91% for most Chinese companies 12/2014: Final ruling of the second round for Chinese wafers, cells, modules (23.74 –258.57%)

41 Trade Dispute: Effects of US Second Round TariffsSeptember 9, 2016 Trade Dispute: Effects of US Second Round Tariffs Canadian Solar Sales Sales (1,000s USD) 2011 192,381 2012 254,097 2013 215,262 2014 604,537 2015 903,748 Decision was anticipated; stocks prices held Expansion of capacities outside of China Source: OSIRIS Source: Yahoo! Finance Jan 2015

42 Trade Dispute: AD/CVD Tariff TimelineSeptember 9, 2016 Trade Dispute: AD/CVD Tariff Timeline 07/2012: China launched an investigation on US renewable energy program 12/2014: WTO decided US tariffs breached rules In Jan. ‘14 China finalized AD and countervailing tariffs on imported poly from US & South Korea for 5 years Avg. US tariff: ~55% S. Korea tariff ranges from 2.4% % 01/2014: China’s final ruling on imported polysilicon from US (~55%) & South Korea (2.4% %)

43 Trade Dispute: Effects of China Tariffs on PolysiliconSeptember 9, 2016 Trade Dispute: Effects of China Tariffs on Polysilicon Source: GTM Research & SEIA. “U.S. Solar Market Insight: Q ” Chinese anti-dumping tariffs enacted Poly utilization rate down from 91% in Q to 36% in Q1 2016

44 PV Mfgs. % of Sales Revenue by Region, 2015PV manufacturers have varying degree of regional exposure The majority of revenue from First Solar and Sunpower comes from the U.S. with virtually no penetration in the Chinese market Many of the publicly traded Chinese companies have large, but no overwhelming revenues from their domestic market – demand could be coming from private companies Shunfeng-Suntech generated 58% of its revenue in China in 2015 Note: not all companies separate revenue into each geographic location represented in graphic. In those instances, all non-separated numbers are classified in “other” unless otherwise stated. Sources: Company figures based on Q4 ’15 (and previous) SEC filings by the respective companies. JA Solar and ReneSola have no filed their annual reports yet so numbers are reflected. Jinko Solar US numbers represent revenues the company received from all of “America” as U.S. was not broken out separately.

45 Trade Dispute: Recent DevelopmentsSeptember 9, 2016 Trade Dispute: Recent Developments US/EU organizations that have come out against the tariffs SolarPower Europe (formerly European Photovoltaic Industry Association), representing 1.3 million European jobs and 130,000 European companies related to PV PV modules could be sold 20% cheaper in the EU without trade restrictions on Chinese panels, according to a study commissioned by Solar Alliance for Europe (SAFE) Australia canceled anti-dumping investigations Trade uncertainties under the new administration

46 Trade Dispute: Response from Chinese FirmsSeptember 9, 2016 Trade Dispute: Response from Chinese Firms Go Big: build up international manufacturing capacity Pros: develop local production capacity in emerging markets; circumvent anti-dumping tariffs Cons: tariffs may change to apply to cells/modules manufactured in SE Asia; challenges in developing local supply chain Maintain a Presence: acquisition in US/EU markets Pros: less risk; can obtain new tech manufacturing capability Cons: limited ability to grow market presence; risks in acquision; no benefits of inexpensive SE Asian manufacturing Stay Home: focus on growing Chinese market Pros: Not subject to trade uncertainty; proximity to market Cons: lack of diversification; missing out on growing markets Go Big: Canadian Solar, NARI Group Maintain: SF-PV/Suntech (Suniva), CSUN (400MW of module in Sactown) Stay home: small manufacturers like Beijing Sunlectric; Suntech

47 What did experts miss before?-“People do not really understand the detailed operation of this industry and did not taking into consideration the contribution by China.” -The cost structure for crystalline Si is different. “People may be familiar with electronics industry, but PV industry has a wider scope than just crystalline Si. Although the core component is crystalline Si, but it’s not the biggest part in terms of cost.” -Price of polysilicon -Scale of Chinese production -The domestic manufacturing and installation of equipment -The rush of private investment -Break up of polysilicon monopoly and partially the rate of technological improvement -“They did not consider the manufacturing in China. Chinese and German investors would invest massively in manufacturing and lower the costs that fast. If the manufacturing had stayed in Germany, the price would have stayed at 4-5 dollars/W.” -Polysilicon supply increased; China broke up and decreased the monopolistic profits -Technological improvements: thinner wafer, thinner silver electrode, higher efficiency -Wrong judgment on the scale of polysilicon material expansion -Rapid improvement in efficiency thanks to technological improvements (equipment as well as other manufacturing processes) -Economies of scale

48 CdTe https://www.energy.gov/eere/sunshot/cadmium-telluride

49 CIGS https://www.energy.gov/eere/sunshot/copper-indium-gallium-diselenide

50 OPV https://www.energy.gov/eere/sunshot/organic-photovoltaics-research

51 Perovskite Solar Cell https://www.energy.gov/eere/sunshot/hybrid-organic-inorganic-halide-perovskite-solar-cells

52 Perovskite Structure

53 Probability of System Cost…Expert <4RMB/W (%) >6RMB/W (%) A 100 B D 95-100 0-5 F G 50-60 H 20-40 I 30 K 90-100 L 80-90 10-20 O 40 25

54 The number of foreign patents awarded to Chinese inventors is increasing rapidlyTotal number of USPTO patent grants Yet India and China are already innovating, as is evidenced by the rapidly rising number of patents awarded by the USPTO to inventors resident in China and India. While patent levels remain relatively low, the rates of growth are exponential.

55 US firms aggressively patent their inventions in other major markets…And even more striking indicator of the relatively low quality of indigenous Chinese firms’ domestic patents is provided in this next set of slides. There is a long tradition in the innovation research community of using the propensity of firms to patent their inventions in major foreign markets as a reflection of the perceived value of their patents. And US firms are pretty aggressive about patenting a large fraction of their inventions in other major markets like Europe and Japan. This graph traces out , for the top 100 US based inventing firms, the weighted average fraction of their US patent grants for which they also seek patents in Europe (point), patents in Japan (point), patents in either Europe or Japan (as well as the US), and patents in all three so-called triadic markets. The percentages are pretty high. These US firms seek protection for nearly one third of their US patents in at least one major economy outside the US.

56 But the top 100 indigenous Chinese applicants patent only a small fraction of their inventions outside China But if we look at the top 100 indigenous Chinese patenting firms, we see that they seek foreign patent protection for a much, much lower fraction of their domestic invention. Measured at market exchange rates, China is still less than half the size of the US economy and substantially less than half the size of the Western European economy. A Chinese inventor who has really succeeded in creating valuable new-to-the-world technology should have a strong incentive to patent the idea in at least one of the much larger economies. Yet Chinese firms seek US protection for less than 6% of their domestically patented ideas, they seek protection for only about 4% in Europe, and only 1% in Japan. Chinese firms seek patent protection in 2 of the 3 triadic markets for only about 3.8% of their domestic grants, and they seek patent protection in all 3 markets for less than 1% of their domestic patents. And if we look at the number of domestic patent grants for which Chinese firms pursue the foreign patent application process all the way to the successful receipt of a foreign patent grant, the numbers are vanishingly small. So far, fewer than 3% of Chinese invention patent grants have been successfully patented in ANY triadic market, fewer than 1% have been patented in 2 out of the 3. And these are the results for the most aggressive, most successful inventing firms in China -- the top If we were to redo this for the top 500 or the top 1000, we suspect the percentages would fall sharply. The simple reality is that Chinese inventors regard the vast majority of their own domestic inventions – something on the order of 95% -- as simply not worth patenting outside of China. That is a striking vote of no confidence, or at least a vote of low confidence, in the quality of Chinese invention – delivered by the inventors themselves!

57 We are not the only ones who question the value of Chinese patent grantsBrian Wright and his students have found that Chinese indigenous inventors inflate their patent applications to meet local government targets… …And to benefit from local government subsidies (Lei et al., 2015) Domestic patents of low quality can also be an asset in an evolving legal system that struggles to distinguish between a good patent and a bad patent The number and growth rate of domestic patenting may (substantially) overstate the true innovation of indigenous Chinese firms Now, if most of these indigenous patents are of low quality, then why do Chinese firms go through the trouble of taking them out? Well, the work of Bronwyn’s Berkeley colleague, Brian Wright, and his students, suggests a partial answer. Chinese firms churn out vast reams of low quality patents because their government pushes them to do so, with a combination of subsidy carrots and administrative sticks. If someone is actually willing to pay you for crappy patents, you’ll take out more of them. And those who have followed the history of IP litigation in China understand that, in China’s very imperfect legal system, vast reams of low quality patents can have legal value even when they have no or next to know real technological value. This is even more true in China than it has ever been true in the U.S. All these considerations help us understand why the number of indigenous patents has expanded so quickly – AND why it is a mistake to regard all of them or even most of them as evidence of a real innovation boom in China. They also point to the usefulness of using China’s foreign patents – if what we are really interested in is a measure of honest-to-goodness advances in the global state of the art. Going forward, we certainly intend to pay more attention to China’s domestic patents than we do in our current draft, and we will describe the outlines of Chinese inventive activity traced out by these data. But we doubt this will overturn the view of Chinese innovation we’ve developed in the current draft based on their US patents. Lei Z, Sun Z, Wright, B (2015) Patent Subsidy and Patent Filing in China. University of California, Berkeley, Mimeo. Li X (2012) Behind the Recent Surge of Chinese Patenting: An Institutional View. Research Policy 41(1),

58 Chinese domestic patent data suggest an explosion of innovation…But before I even start, I need to contend with an important question. What about the tidal wave of patents being issued in China itself? For those of you who’ve seen the numbers, they’re mindboggling. Just in 2011 alone, China’s state intellectual property office granted nearly 1 million patents of various kinds. And the majority of these go to purely indigenous inventors How could we possibly base our inference about innovation in China on the relatively small number of US patents Chinese inventors have received, when the number of indigenous patents grants is so many orders of magnitude larger? The first thing we need to realize is that the overwhelming majority of these domestic patent grants are actually utility models or design patents. Neither requires a substantive examination, and neither require a significant technical advance over the existing state of the art. You simply cannot think of these as patents in the standard sense of the word, as we normally use it in the innovation literature.

59 But the numbers of “true” patent grants (invention patents) are much smaller…When we focus on Chinese so-called “invention patents,” which do require an examination and do, in principle, require an advance over the existing technical state of the art, we see significantly smaller numbers and significantly less growth. Total grant numbers drop by about 80% in recent years. This consideration alone helps bring measured Chinese innovation from the heights of Olympus down toward the plains of Thessaly. To dig into what these domestic invention patent grant data might tell us about innovation in China, Guangwei and I bought the micro data from the Chinese patent office. Obtaining and cleaning these data have been expensive, in terms of time and money.

60 A significant fraction of Chinese invention patents are awarded to foreign inventors…The first thing we learn as we dig into the invention patent grant data is that, historically, a very large fraction of grants have gone to foreign inventors, who have been seeking to protect their intellectual property, developed abroad, in the Chinese market. In fact, if we focus only on firms’ patent applications, it is only now that the fraction of patents granted to domestic inventors is passing the 50% mark. That being said, the Chinese patent data appear to point to more indigenous inventive activity than what we find in the U.S. patent data. But, as we all know, simple counts of patents can be misleading. And that is certainly true here. A range of indicators suggest that the quality of Chinese patents awarded to indigenous firms is quite low.

61 Indigenous firms allow their domestic patents to expire much earlier than foreign firms doFirst, we should point out that many Chinese inventors never even request a formal examination of their patent applications and this is part of the reason why the grant to application ratio is significantly lower in China than it is in the US, the EPO, or the JPO. Another indication of the low quality of Chinese domestic patent grants is illustrated by this slide. There is a long tradition in the innovation research community of using the willingness of firms to pay renewal fees for their patents as an indicator of patent quality. This graph shows the average age at which patents expire for failure to pay fees. The fact that domestic firms allow their Chinese patents to expire much more quickly than foreign firms suggests that their patents are worth less. And the low rate of renewal by indigenous inventors is a major factor dragging average renewal rates in China far below those of any major jurisdiction, especially at greater patent ages.