It seems that the solar PV manufacturing industry is headed for a downturn in 2024, with most upstream sectors likely becoming loss-making.
In this article, I will explain why this is most likely to happen next year; why the current inflated polysilicon supplier margins may be a distraction; what factors could help the industry avoid this; and whether those factors can really halt the first major slump since 2013-2014.
Finally, I will also suggest groups that may benefit from a 2024 PV manufacturing recession.
The recession was set to start in 2020 but was delayed
Technology and manufacturing are hard to predict. But one thing is certain – business cycles. After almost any prolonged economic upswing, an economic contraction of some sort is almost inevitable – a kind of reset where many business plans get impacted.
The solar PV industry is no exception. In fact, it could be argued to be an extreme example of this phenomenon since it relies on government policy, energy sector benchmarks, and the premise of a transition to a green economy for its very survival.
Since 2016, PV manufacturing has been in a boom period: positive margins (no matter how tight) were never seriously questioned, at least for the Top 10 companies across the manufacturing value chain. The risk of investing in new capacity was low; investors were confident about near-term profitability potential for public companies; companies could plan technology upgrades and transitions without worrying new products would be shut out.
Looking back, what’s most surprising is how long this upward stretch actually lasted – about eight years to be precise. It really was an era for the solar industry.
In the weeks leading up to penning this article, while collecting data to support the 2024 outlook, I kept trying to find some magical chart that definitively showed everything happening today reflected the final stretch of the last prolonged upcycle affecting the PV industry ending in the 2013-2014 period after the effective ‘closure’ of European feed-in-tariffs. This was when China was hitting the big red button on gigawatt-scale expansions and shipping product to Europe.
I did dozens of spreadsheet calculations, tracking production metrics like shipments, surplus capacity, inventory, capex, and end-market demand against financial ratios. Ultimately, I gave up. My conclusion was comparing the PV industry from 2013-2023 in any meaningful way was basically like comparing apples to oranges. However, some key indicators are universal, which I will discuss later.
But in doing this work, I suddenly realized upstream manufacturing was supposed to enter its recessionary period two years ago but was saved by two factors no-one could have predicted. Especially the manufacturers themselves desperately trying to figure out what the sales environment would look like two-three years ahead.
Before explaining these two factors, I will use the chart below to illustrate what “downturn” really means, as the term will immediately trigger panic for many but is actually a story of winners and losers in the short-term.
Firstly, everything I’m commenting on here relates to an upstream slump; restricted to module manufacturing and its value-chain stages (cells, wafers, ingots, polysilicon and finally other technologies tied to ultimate module shipment levels: thin films). Also benefitting from sustained investment and profitable manufacturing are materials and equipment suppliers.
In my frame of reference, a downturn represents makers struggling to produce product at cost (usually negative operating profit) and slamming the brakes on controllable variable expenses especially capex and R&D spending. This also overlaps with the inevitable outcome of such an operating environment – manufacturers purely focused on ‘midstream’ (wafers and cells for c-Si manufacturing) getting squeezed out of existence by their suppliers and customers (typically the dominant poly and module players – refer to the case study of what happened in Taiwan).
Now, before predicting how the possible 2024 recession may unfold, let’s look at the two factors I believe only delayed the inevitable slump.
The three-year slide delayed – why?
In Figure 1 (see top wording) I mark 2014-early 2020 as the first period, covering the time after the end of the previous PV industry downturn (lasting around two years) that impacted the sector in 2012-2013.
The PV industry was probably ready for the start of a recessionary period (not necessarily an acute long-term one) sometime around 2020 as expansions took place and growth was seen during 2019. Without the issues outlined in the second and third periods in the chart above, this would have been a logical reset.
The second period covers Q2 2020 until Q3 2022. I term this “Recession Delay Period 1”. During this time, the solar sector changed direction. In short, end-market demand spiked tremendously; renewables and climate change became real issues globally with everyone wanting solar tomorrow. Module supply was bottle-necked by polysilicon availability. Polysilicon prices skyrocketed, also explaining why the polysilicon margin (Poly gm in chart) rose from around 15% in early 2020 to over 75% by the end of 2022.
Normally, a surge in polysilicon prices would have had profound implications on module supplier profits (polysilicon costs rose three-four fold). However, as mentioned earlier, margins for module suppliers (at least the leading global Chinese player(s)) were barely impacted during the second period. This was because solar demand far outpaced supply levels, and module suppliers could increase their Average Selling Price (ASP) for modules. In short, module costs and ASPs increased almost simultaneously, keeping margins (for leading Chinese player(s)) around 15%.
During this time, the module market was very much a seller’s market to the extent module suppliers would routinely overnight tear up old supply agreements because the next day they could resign at a far higher ASPs with other customers under agreements 10X the size of old deals.
This second quarter 2020 to third quarter 2022 period did not actually last a long time. It certainly would come crashing down but when?
By the end of 2022, this had started to unfold. The reason was immense new capacity across the value-chain coming online in China, not just from new entrants but existing manufacturers (like Daqo) as well. These producers were in effect adding surplus capacity, potentially eroding their own margins.
We are currently in what I call the third period above (Recession Delay 2) starting third quarter 2022.
No end-market for any technology could feasibly absorb 30-40% supply surplus overnight, and PV is no exception. Almost overnight, therefore, the entire sector flipped to being a buyer’s market, with polysilicon pricing collapsing. Daqo’s margin nosedives from third quarter 2022 to end of 2023, clearly shown in earlier charts. For Daqo, this is not an issue having the lowest cost structure is hugely beneficial. But for other poly makers, it means danger. Western suppliers faced possibility of return to the dreary days of years past with poly pricing below US$10/kg, given no corresponding spikes in US module pricing.
From third quarter 2022, module ASPs started a continual decline, especially in H2’23 to new lows below 20 US cents/Watt (current ex-US global pricing).
But module margins do not seem to have changed much complexity-wise. The reason is the rate of cost declines (sometimes even faster than the module pricing drops) keeping pace with how quickly module prices reduce. And all these cost reductions stem from lower polysilicon pricing, not any major shifts occurring in ingot, wafer, cell or module assembly costs (more later).
As evident from the chart, module margins by second half 2023 are already higher than Daqo’s polysilicon margin.
The averted manufacturing recession therefore is entirely due to the cost reduction amplitude from polysilicon outpacing the module pricing drops.
But now we are at the point where the module supplier has no more get out jail free cards to play. This is why I strongly believe from 2024 onwards, PV manufacturing is highly likely to enter its next recessionary phase, unless other delaying factors to a recession appear.
Prices still falling but costs have almost nowhere to hide
Currently, module pricing continues to decline, a factor of the surplus capacity, oversupply, and inventory build-up affecting the two key end-markets of Europe and the US right now. End-market demand is not a binary on-off switch that can simply be activated by shipping large volumes to regions or having fire-sale offers – it relies to some extent on hope for a magic demand elasticity.
In fact, this last point seems to be where many module suppliers today are pinning hopes for a 2024 saving grace. To some extent, one can understand the reasoning. Projects in the multi-gigawatt range being offered at around 15 US cents and financed by institutions will to some degree make projects economical that previously were not, almost overnight. If only life were that simple! Reality is large ground-mount type projects (essential to consume such vast capacities) tend to proceed on schedule due to planning, permitting, grid availability issues and so on. While tweaks in module pricing clearly help, there are few single factors that could accelerate projects by 12-18 months.
It is therefore completely reasonable to expect pricing will continue declining, for at least the next six months. Certain markets or sub-segments of these markets will show some price elasticity to be sure. For example, US utility-scale ground-mount projects needing Western poly and non-China wafer-cell or buyers of modules in 2022 that signed contracts at higher prices and are reluctant to renegotiate/review lower alternate quotes.
Timing-wise, the chances of module pricing recovering look equally dim. The calendar year-end in January annually tends to have pricing crashes, and 2024 will be no exception as China itself is the largest shipping end-market and has seen typical calendar year-end resets before (where demand falls off a cliff, factory capacity adjusts and pricing goes down).
The sad part is things look set to become even worse from a pricing standpoint.
If the number of manufacturers had remained the same, then all that would be needed is each maker to adjust production and supply, controlling inventory to avert a crisis. Unfortunately, this has never been a strength for Chinese manufacturing.
Even if there were a chance for large Chinese poly makers to agree (some cartel type arrangement) to limit supply and keep pricing robust, they are doing exactly the opposite. Each company continues adding new poly capacity relentlessly chasing max production, become obsessed with market share growth. Ultimately, this simply erodes their own margins.
This is before considering all the new entities entering the industry, again all Chinese.
If someone had deliberately tried to get pricing to a point that decimated margins, thereby luring other domestic (or overseas) rivals to enter who could then be responsible for crashing pricing overnight and removing any ability to compete this would be impressive and warrant praise. However, I doubt I am giving anyone enough credit in this case! It requires collective (or cartel type) thinking, which I don’t think exists in China right now.
Factoring surplus ingot, wafer, cell and module capacity currently in China, the oversupplied module situation is frankly already out of hand domestically. What concerns me more is whether upstream markets in 2024 will be loss-making.
In recent years, countless new entities in China entered the value-chain, inadvertently building 10-20GW wafer factories almost overnight. These firms have zero industry knowledge or background and lack technical competence and the customer channels to sell product. This is a recipe for disaster in any industry.
Even with inventory building and pricing in freefall the last few months, the obsession with PV manufacturing activity continues in China.
In China, the 100% domestic equipment supply-chain is equally obsessed with the growth narrative thanks to upcoming IPOs and being newly listed. Piling on homogenous c-Si plants does not seem smart right now for China.
In the section below, I will predict possible price and cost trajectories and when decisions (especially capex) may change leading to 2024.
Breaking even by mid 2024?
Figure 2 outlines what 2024 could look like, assuming several factors. Of course, any one factor may not play out. Remember this is just one potential outlook, and many others are out there depending on what you assume the PV world in 2024 resembles, not to mention unknowns (like pandemic and wars in Europe the past few years).
But from the PV manufacturer standpoint, I have summarized what I believe 2024 will most likely see. The chart below again tracks the key metrics that can support the thesis.
Predicting pricing is close to impossible! In my 20 years watching the PV sector no-one has ever got pricing forecasts correct. Something unexpected always happens. Whether completely irrational firesales (like huge losses) or buys (purchasing at any price) that cannot be anticipated.
So here I will keep things simple. I assume pricing (based on reporting from large Chinese module suppliers on global average ASPs across all end-markets, US$ per watt) will decline from today’s levels (around US$0.18-0.19 per watt) to US$0.14-0.15 per watt by the end of 2024. While intra-year seasonal changes cannot be predicted for the year, I have just analyzed based on a linear decline during this period. This is enough for the analysis that follows.
The key here is it forms a baseline from which possible cost trajectories can be considered and cost trajectories are in fact easier to predict than pricing.
For the vertically-integrated monosilicon module supplier (making some/all ingots, wafers, cells and modules in-house) there are 5 key cost elements that will impact profit/loss ability on modules. In the analysis below, I will look at the absolute lowest-cost scenario. If the lowest-cost integrated player cannot make money then things only look worse for others.
Polysilicon cost is based on purchasing from another (e.g. Daqo) that has a benchmark cost and looks to profit. Ingots, wafers and cells become zero-sum games just considering the production costs as there is no requirement to profit for these internally-produced items. Module assembly has a production cost (currently the highest across the value chain) which is ultimately where the margin (profit/loss) comes from as these firms only count module sales as company revenue.
Predicting polysilicon-to-module costs for the next 12 months is a relatively simple task with 15 years of data and trends to provide support, assuming no short-term spikes occur in prices of raw materials (silver, glass etc.) or consumables not controlled by the PV industry.
Has the industry been learning to cut costs?
Within this part of my research, I decided it was time to finally look at the cost reduction within the mono-Si value chain in a way I had contemplated for 10 years. Specifically – how much do the non-polysilicon costs truly come down year-after-year? I envisioned the presumed “learning curve” of the PV industry. This chart I must have seen thousands of times over the years and discussed frequently.
Every year, the International PV Technology Roadmap (ITRPV) report is published. It is an extremely valuable compilation containing indicators and graphs tracking key parameters across the (almost exclusively mono c-Si) PV manufacturing process. One of the most cited charts in the ITRPV report is the so-called ‘Learning Curve’ which tracks the functional relationship of PV industry average module selling prices versus cumulative installed PV capacity each year (on a logarithmic-log basis). Underlying this curve is the assumption declining module ASPs year-after-year across the total industry are linked to cumulative learning.
Overall, the finding sounds great but module average selling prices tend to have multiple determinants (policy, under/over capacity etc.) and whether cumulative installed PV capacity is the best proxy for tracking annual ASP declines could be debated. How much of the year-over-year price declines (or shifts for that matter) are due to existing cumulative installation scale?
For many years, I had thought about plotting annual mono c-Si costs versus annual mono production each year instead. From a cost standpoint, this could show ‘learning’ more clearly. Considering the issues now around whether 2024 may usher a sector downturn (and poly costs/pricing potentially equalizing/bottoming out), how far away are we from ingot/wafer/cell/module manufacturing/processing costs being limited only by underlying material costs (silver, glass etc.)?
Figure 3 shows (what I believe is a first) a learning curve for c-Si module production costs declining linked to annual mono-Si production levels over the period 2013-2023.
A few interesting observations by studying Figure 3 closely show non-poly costs (ingotàmodules) have indeed come down each year looking back 10 years. Going back further than this serves little purpose. It is in the last decade Chinese vertically-integrated firms have been setting the benchmarks for costs, which have become the global base for the whole industry.
Without the 2023 data, one might conclude from the chart some constant relationship exists between annual mono capacity and lower non-poly costs. But of course, this cannot logically be true indefinitely. At some point the benchmark costs become an outcome of buying consumables many of which the PV industry has no control over regardless of how cheap labor and electricity are made in China or broader South East Asia.
The 2023 deviation may be the first in many years signaling Chinese c-Si makers need new measures for cutting costs further. In 2023, driving down (non-poly) costs has simply not been imperative, largely because the polysilicon costs declined so much and gross margins at the module layer could still be maintained in the mid-teen US$/watt range.
Indeed, during 2023, the main variable driving down costs (poly+non-poly) has clearly been shown in earlier charts to be the polysilicon cost (sales pricing) with wild gyrations this year. For 2024, I am assuming players like Daqo (and most of the Chinese poly industry) will have sales pricing below today’s but still not constituting any major threat to operational losses. For argument’s sake, I took 15% margin (for the poly suppliers) to forecast 2024 average selling price for polysilicon and then folded in the power demand/cost trends to extrapolate the final polysilicon costs for module makers procuring material.
This leads to questions around cost-cutting activities for 2024 (and 2025) of course assuming manufacturing does indeed face a recession next year. Huge strides in cost cuts were made across China PV during the last recessionary phase, establishing an almost 100% domestic materials/equipment ecosystem specialized for silicon PV manufacturing. Glass was thinned, silver lines made narrower, factories scaled in size, automation increased as were supplier margins squeezed.
This time, such tricks are not there for deployment as these activities are now largely done. Can or should glass realistically not reduce thickness further? Is it time to take wafer
