Teri Viswanath: Welcome to Power Plays, a CoBank Knowledge Exchange podcast series, an audio program where we connect you with top energy and environmental innovators who share their insights, market observations, and experience. On today's special Earth Day program, we're going to explore the key catalyst that will drive the next decade of cost declines for solar power.
The economics already suggest the sun will be the largest source of energy for the world by 2050. This topic seems like the right one to tackle on this particular day. I'm your host, Teri Viswanath, the lead economist covering power, energy, and water for CoBank. I'm joined by my co-host Tamra Reynolds, CoBank's regional vice president of electric distribution. Hi, Tamra.
Tamra Reynolds: Hey Teri, I'm glad we had a chance to connect today. Having been involved as a banker financing utility-scale solar projects for the last decade, I've had a front-row seat to observe the falling costs of solar modules and I'm really excited to talk about today's program. To go deep into this topic, we invited Simon Price, the founder, and CEO of Exawatt to join our discussion. Simon and his team in Sheffield, England are globally respected experts with deep technical understanding of all facets of the PV value chain from material supply and manufacturing to downstream project development. Simon, welcome to our program.
Simon Price: Thanks for inviting me. It's good to be here.
Tamra: Simon, from an electric co-op’s perspective, we usually think about costs at the module level, but in our briefing call with you last week, you talked a little bit about the segments of the upstream solar supply chain and how different innovations in these segments has really been responsible for driving down the costs. For starters, it might be helpful if you could provide our audience with an overview of that.
Simon: Yes, absolutely. The lion's share of the PV module market globally, something like 95% comes from solar cells made from crystalline silicon, and the rest comes from what we would call thin-film modules. There's really only one significant manufacturer of those globally and that's First Solar in the US. Today I think we're going to be focusing on the crystalline silicon segment, and there's essentially four main steps in that process and one critical material input at the start of those processes, which is polysilicon.
Those four steps you have, the first one is ingots crystallization. You're essentially melting chunks of polysilicon in a furnace and then recrystallizing them to form a silicon ingot. Then you have the wafering step where you take those ingots that you've grown and you slice them into thin wafers, about 180 microns thick, so a little over five wafers per millimeter of silicon. Then you have the cell manufacturing step and that's where the magic happens is that's where you take your wafers and you convert them into electricity-generating devices through various processing steps.
Then finally, you have the module assembly step, and that's really just taking those cells that you've built and connecting them together in series, wrapping them up in an encapsulant of some kind, putting a sheet of glass on the front and on the back, either a weatherproof back sheet or another piece of glass. Then you generally put an aluminum frame around the edge although we don't always do that, and a junction box on the back, and that's how you connect the modules together and then you're done, that's your fully assembled module.
Another important distinction in terms of the process is back at the ingot step. You can grow silicon ingots in two crystalline forms of silicon. One is called monocrystalline, so it's a single crystal. The other is called multi-crystalline, which as the name implies is multiple crystals of silicon, smaller crystals that they're all joined together and you can see them in terms of grain boundaries, if you like, between these little grains of silicon.
At Exawatt, what we did back in 2015 is we made some forecasts based on some technical and economic inputs and modeling that we've done, that mono which was less than 20% of the market at the time, but ultimately take over from multi, which was about 80%. Multi wasn't quite as high performance as mono, but it was much cheaper to make and to build into modules, but we could see some ways that mono would catch up on cost. If you have two modules that cost the same, one mono with higher performance and the other multi with lower performance, the higher performance mono module is going to win in the market because it produces more electricity so it reduces the levelized cost of energy at the system level.
Teri: That's super helpful. Thanks, Simon, for that background. I actually think we have to go back and think about that market. You talked about that perspective back in 2015 and I think we've seen a relatively important change occur since that time and really thinking about the sum of the parts. Let's talk a little bit about it and unpack that a bit more. Tell me what's changed generally since 2015.
Simon: There's certainly been a lot of consolidation since 2015, although much of that consolidation has happened relatively recently. Back in the mid-2010s, they were a couple of 100 at least module sellers from small shops to fairly large global companies that are still around now. There's been some consolidation and upstream there's been quite a lot. In fact, it was always more consolidated the further upstream you went. The number of polysilicon makers was and remains relatively small and dominated by a few giants as you get to the module level where module assembly is relatively cheap to build that you tend to find a broader group.
I guess from a technology perspective, we began to see this possibility in 2015 that these two processes I mentioned, the monocrystalline and the multicrystalline would begin to converge from a pricing perspective. We expected that mono wafers would remain more expensive, but we expected that by the time you get to the module level, the cost per watt, so the cost per unit of power that that module can generate would start to approach parity between those two technologies.
At the time what we said was quite contentious, some would even say heretical, we made a tour in early 2015 of some of the major Chinese manufacturers. Most of the senior execs we talked to at those manufacturers were very skeptical about what we were saying. One or two were not. In fact, one, particularly a company called LONGi said to us, what you're proposing is what we're doing and they were fairly small manufacturers at the time, maybe just about on the fringes of the top 10, but probably a little outside. As of now, they are either the first or the second largest manufacturer in global PV.
They did what we predicted the industry would do, and they were the first to do it and they've become a leader as a result. I think the only thing we really got wrong in our forecasting back in 2015 is that we said that the tipping point would happen five years ahead. We said that within five years, the market share of mono would go from 20% to 50% and in fact, what happened is that the mono achieved 50% market share in mid-2018 and mono is now well on the way to complete domination in the 90%-range for market share now and likely to exceed that.
Tamra: Simon from a consumer perspective, I think that the marketplace has really grown accustomed to this rapid rate of technology innovation, massive supply chain reinvestment in the new processes, and of course falling costs. There seems to be a general trend that bigger is better, maybe like watts per module, for example. Are these factors going to continue to drive down costs?
Simon: Yes, I think it's important that we talk about that trend and what it means for costs at the consumer end. The rise of these mega factories that's really happened over the last five or so years has been an important factor in reaching scale. To some extent, collocating those facilities, those different parts of the supply chain has helped too.
In the US First Solar, for example, which has facilities in Ohio is located within the same state as three of the top 10 largest glass manufacturers in the nation. You see some of this co-location in China too, although in general the wafers and the polysilicon tend to have been made more in the north and west of the country where the energy costs are the lowest in the country. In general, you can say that bigger is better from a manufacturing scale side. From a module perspective, bigger isn't necessarily better.
The way that we tend to look at manufacturing is on, as I mentioned, cost per watt or performance per watt. What that means is that you think of it as a fraction with a numerator and denominator. The dollars is the number of dollars it takes to make that module, including all of your capital costs of depreciation, all the other things, the energy, labor, and materials that go into making that module. Then you divide that total cost by the number of watts that the module can produce.
When I say that, what I mean is what's its rated power under what we call STC, standard test conditions. The sunlight overhead about 1000 watts of sunlight per meter squared shining down on that module at a tester. That dollar per watt number either in cost or price, or you can break it down as you go up the supply chain into the components, that's what we focus on and that's how we look at where the technology has to go and which technologies are likely to win.
Our thesis generally is that for a technology to be successful, it has to increase performance, so the efficiency of the module, the power per unit area, if you like, but not increase the cost per watt. Although that seems like an obvious thing to say in some ways, you'd be amazed how many companies have gone bankrupt by not paying attention to that fundamental principle and I'm happy to get into that in more detail if you'd like to.
Teri: Simon, I think that's just the part which is we have this idea whether it's watts per module and I also think right now the discussion around wafer. I think the watts per module you talked about that, this had been sort of bragging rights for module manufacturers and now we have this new discussion, which seems like the old discussion we just had about wafer sides. I don't know, is that the thing that pushes us slower in the next five years, is it the wafer discussion?
Simon: I would say not really. In all of module manufacturing, it's an optimization game and the focus of energy or effort in the various steps tends to follow where the highest cost is. Back in the late 2000s, polysilicon was by far the highest cost component of a module. On a dollar per kilogram basis, the spot price of polysilicon in late 2008 was something like $350, $400, and occasionally even higher than that per kilogram.
Today, polysilicon costs maybe $5 or $6 per kilogram to manufacture and sells for between 8 and 10, perhaps a little bit higher last year due to some supply demand effects, but it's below $10 a kilogram and it's going to stay there. Polysilicon isn't the driver that it used to be. Generally speaking, as modules improve, the bulk materials in those modules start to become more to dominate the cost structure. The reason for that is that some of those other materials like glass and the aluminum and the backsheet and the encapsulant and the junction box, those are fairly without wanting to be too rude, they're dumb products in a way.
They're bulk materials, they're commodities, and they're somewhat difficult to cost-reduce. Unless you can find a way to make the glass thinner, for example, you can't really reduce the amount you use and the price of glass, absolute occasional blips due to supply demand effects, hasn't really changed that much over the years. It's reached a low price in China and it's not getting much lower. The way that you improve your cost per watt, the way that you reduce your cost per watt over time is by focusing more on the watts than on the cost. You improve performance. That's why mono got to where it's got. It's got there based on its performance, not on its cost really.
When it comes to the question you asked about wafers and whether larger is better in that sense, we're in that optimization stage right now. Different manufacturers are innovating with different wafer sizes. They are getting bigger and there are some scale effects in manufacturing from getting bigger. There are also ways of deploying those wafers differently into the modules so you can cut those wafers into pieces, for example. The first step is what you would call a half-cut cell. You take your cell and you scribe it down the middle, and then you have two cells, and then you piece those together, you string those together in the module, and you can get some advantages in performance from doing that.
There are various techniques for extracting more performance out of a module by changing the size of the wafer or the way that you configure the wafers or cut them, and therefore the size of the cells that you're making from those wafers. It's a little bit too soon to say which one will be best. My feeling is that there'll be some continued optimization, but different manufacturers will promote their solution to that problem as the solution and we're not quite there yet.
When we've looked at our cost forecasts in general, although there are small, subtle differences between the full integrated cost of modules that are built from these different types and sizes of wafers, by far the biggest determinant of change and I wouldn't say noise, but the window of whether the high, low cost might be in the next few years. By far the biggest determinant of that is the ongoing cost reduction due to improve performance and due to those bulk materials that I mentioned, it's not really from the size of the wafers.
Teri: That's interesting. Maybe the market has come to have this expectation that this is just what we do. We have these massive drops in module costs every year, and that's what we've become accustomed to. When we think about maybe we've gone as far as we can go with lowering costs for the first-generation solar technology and in the next two to three years, guide us on expectation for cost of declines.
Simon: There will be ongoing cost declines on a dollar per watt level. That's absolutely true. In the current generation of modules, which I would call the mono PERC generation, monocrystalline and PERC being the name that we give to the cell that has a passivated emitter and rear. The rear side of the cell has also been improved relative to previous cell generations.
That mono PERC generation is starting to approach the limits of performance, but it's not there yet and we think it will probably hit its limits in three to five years. At that point, there will be a move to new technologies, most likely based on N-type and then potentially beyond that-- when I say N-type, I mean N-type silicon, rather than the P-type silicon that most current modules are made from.
Then beyond N-type, there may be other structures, you may have heard in the industry discussion of perovskites for example, essentially layer that goes on the surface or can we put on the surface of a traditional crystalline cell to improve the efficiency of that cell. There are other formats that might come along, but we're talking more than five years and probably 10 years before those formats start to dominate.
Teri: Not even commercial, really. [crosstalk]
Simon: Not yet, but there are some good numbers coming out from the manufacturers that are doing R&D on those. We can't rule them out, but I would say the next three to five years is taking the horse we're on and just extracting the most performance out of it. Then the next few years after that is probably moving to a subtle changes within the current technology, really.
Moving to N-type silicon, potentially changing the cell construction so there are two candidates in the cell construction. One is to call heterojunction, which is a technology that's been around for a long time, but it's now almost ready for the big time. Another one is called TOPCon, which is another approach to creating a high efficiency cell. We currently expect them to do battle in the next three to five years, something like that.
Tamra: Simon, from a global perspective, where are we now with regard to the manufacturing costs for panels and where might we be headed?
Simon: Okay, that's a good question. One thing I didn't note earlier is that almost all of the work we do at Exawatt is focused on manufacturing cost, not price. We do, do some supply demand analysis, and we do some price forecasting, but we always do it on what we call the should cost basis. We figure out for a given manufacturer, and we usually benchmark it to the leading fully integrated manufacturer, because that's the lowest cost, what will be their manufacturing cost of goods sold or cogs and we use the US accounting definition for that.
That's the manufacturing cost of the module. It's the bill of materials, it's the energy, the labor, and it's the depreciation. What's that cost of goods sold going to be? We currently believe looking at our numbers as they're coming in now, that the cost of goods sold for a mono PERC module, which is the market leader, the dominant platform if you like in the industry, at the end of last year was in the 18.5 to 19.5 cents per watt range.
We expect that by the end of this year, so the end of 2021, that those same costs of goods sold for the same module-based with similar wafers inside will be between 16 and 17.5 cents but it could fall as low as just below 15 cents in what we call the squeeze case, which is the case where the market is severely oversupplied and then demand falls temporarily for whatever reason. We found that in the past we used to do a base high and low set of forecasts and in the end what we found is that generally for solar it's base low and squeeze. Because you do get these moments where the industry is just fundamentally oversupplied. We're not currently in quite that situation, but we have been once or twice in the last few years.
Tamra: Simon, that's super helpful. In terms of a cost perspective, I was just thinking back to your comment you just made, which was about the technology, where it might be heading. We're getting really good with this first-generation technology but I think before we wrap up the program, I'd love to hear your thoughts on the next-generation technology. When does it go from test case to pilot to commercial in what that next chapter of solar generation might look like?
Simon: Sure. I should say before I could move on to that, as I mentioned, we focus on cost. Those costs I gave you, you shouldn't expect those to turn up as prices in the US market. Those are factory gate costs in China for the large integrated manufacturers. You've still got to ship them to the US, and there's the shipping on this profit margins, and what have you along the way. Those are just the cost. It tells you what the Chinese manufacturers baseline is basically.
As I mentioned earlier, we think in the next few years it's likely to become a battle between-- well, it's really between three technologies. It's the existing incumbent dominant mono PERC modules that we see today in the market, and then it's the merging new platforms, the N-type platforms, heterojunction, and TOPCon. At the moment, those technologies while being superior in performance are still significantly more expensive, the mono PERCs, so the application set if you like, the number of use cases where it makes sense for them is fairly small right now. For them to become dominant at a global level they have to start to match mono PERC on cost.
One of the ways they do that is by ongoing cost reductions in those technologies as they scale, but the other one is by an expanding performance advantage. An expanding cell and module efficiency advantage. If I look at our efficiency forecasts at the moment, we have mono PERC on average at about 20% efficiency right now in late 2020, early 2021. Heterojunction is a little bit ahead of that and TOPCon is about the same. By 2025, we think that mono PERC will increase in efficiency and maybe get to above 21%, maybe 21.5, but heterojunction and TOPCon, probably by that point will be approaching 23%. That's a big efficiency advantage.
That when you combine it with ongoing cost reductions as those technologies scale, we believe is likely to propel one or both of them to eventual dominance over mono PERC, but it's a little too early to call which and there are reasons why mono PERC might do better than we forecast. It has a history of surprising us in a positive way.
Tamra: It goes back to the original conversation we just started with, which was the mono versus multi. We continue to see these these technologies, this combination of competition for the same reasons.
Simon: Yes, absolutely. I would say with mono and multi, what we saw was staring us in the face even though nobody agreed with us at the time. In this case, it's not as clear cut. People have been talking about N-type being the next winner for as long as I've been in PV, which was in the late 2000s. We've said for the longest time, it's not going to happen yet. It's five years away, it's five years away. We said back in 2015, N-type will come but not until mono PERC run out of room. That date that we set for N-type way back, that's moved out in time further ahead.
The incumbent technologies in PV often have a way of hanging around a lot longer than you would expect. Engineers managed to find ways to extract more performance and reduce and take out more and more cost. Don't bet against that.
Tamra: Well, Simon, thanks for joining us on today's program. We really appreciate your insights. We'd like to invite you back soon to walk us through Exawatt's outlook for battery storage.
Simon: That'd be great. Thank you very much.
Tamra: Teri, I just had a few comments about our interview with Simon. First off, Simon is terrific for breaking down a relatively complex topic. I appreciate his insights. I think the point that I was struck by is that we're still really deploying what most consider to be first-generation technology.
Teri: You're absolutely correct. Simon's point that within the space of just a few years, the dominant mono PERC technology that might potentially reach just above 21% efficiency, maybe 21.5, it might be displaced by heterojunction and TOPCon. These technologies will be approaching 23%, and that really feels like another breakthrough for the industry.
Tamra: Teri, in our conversation with Simon prior to the podcast, he discussed even greater breakthroughs on the horizon. Next generation technologies with new materials that will capture and convert a greater percentage of sunlight enabling us to reach closer to 27% efficiency. These new efficient modules will probably be commercial in the next decade.
Teri: Out of the lab and soon to the market. We hope you've enjoyed this episode of Power Plays, a CoBank Knowledge Exchange podcast series. Please join us next month when we will visit in depth with Rural Electric Cooperative thought leaders. Thanks again.
