This week the polysilicon price declined 5% to $30 per kilogram, so about $80 per kW, as of the Silicon Industry Branch’s Wednesday price update. In contrast, wafer prices increased 4% in a March 3rd price update from LONGi – and in that week polysilicon had declined by a little over 1%. LONGi attributed the wafer price increase to a shortage of the high-purity quartz crucibles which are used – and expended – in the Czochralski process which turns polysilicon into ingots ready to be cut into wafers.
Steep price rises for crucibles were already reached in 2022, with the stock price of the largest Chinese manufacturer tripling from April to July, but until now this was always overshadowed by the polysilicon bottleneck.
In November the crucible price reached $1,700 per item, triple the figure from 12 months before, and last week the crucible price reached $2,800, which is up 25% from last month. The Shanghai Nonferrous Metals reports that the high-purity quartz used to make these crucibles rose 50% in a week to $10,800 per ton. Assuming that crucibles currently take 100 kilograms of sand to make – there has been a trend of larger sizes – that means quartz sand, also in short supply, now constitutes 36% of the crucible price.
It takes something like 2,640 crucibles to produce 1 GW of ingots (according to a source mentioned later in this article), meaning that you currently pay $7.4 million for crucibles to make 1 GW. That is to say $7 per kW, which is 2.8% of the module price, and 0.7% of the cost of a utility-scale project.
So for now this is not a severe problem, but shortage can allow very high prices to be charged, making that 0.7% into several percent, or simply limiting the global wafer output and thus the whole industry.
This crucible supply shortage will last through 2023 and one or two years thereafter, judging by the construction period of new capacity – with one Chinese quartz sand factory construction being predicted to take 36 months, though other capacity is expected to come online in 2023 and 2024.
Even a fast expansion of production capacity will be matched by hundreds of GW of additional latent demand. A LONGi representative told a Securities Daily reporter that some relief can come from simply using lower-quality crucibles, whose production capacity can be expanded more rapidly, and replacing these lower-quality crucibles twice as frequently. Another workaround is to continue the thickness reduction trend for wafers, with TCL Zhonghuan this week changing its offerings to a range consisting of 150μm, 130μm and 110μm.
The crucible has in the past been less than 1% of a module’s cost, whereas polysilicon was 10% of the cost before the 2021 and 2022 price rises which brought it to 40%. Considering polysilicon producers happily charged six times their cost of production, crucible makers may well now charge even more disproportionately, but from a much lower base. The biggest impact may not be for module buyers, but for wafer manufacturers, with large players able to secure affordable supplies while smaller companies lose out. The wafer segment of the supply chain is already quite consolidated in any case.
It should be noted that the supply of these high-purity quartz crucibles is one small fragment of the photovoltaic supply chain which China has yet to decisively dominate. There are two reasons for this – one is a technological barrier owed to the very high purity requirement, the other is the dependence on specific deposits of high-purity quartz (HPQ), with most of the global supply for photovoltaic crucibles coming from the Spruce Pine mining region of the US, and most of the rest coming from India.
This dependence on a small number of specific deposits worldwide would threaten the solar industry’s growth, but it is possible to recycle cracked, impure, used-up crucibles into fresh and pure crucible material instead of discarding them as is the current practice. The cost of such recycling is not high enough to stand in the way of the 1,000 GW solar industry of 2030, but there is little mention of it so far and its adoption may take some time after it is seen as necessary.
According to an article published by the Silicon Industry Branch in November, only three large companies – Unimin in the US, TQC in Norway, and China’s Pacific Quartz produce HPQ sand suitable for photovoltaic crucibles, with a host of upstart Chinese companies making a smaller contribution – with Flat Glass Group among those investing hundreds of millions of dollars in production capacity since the bottleneck began. These companies were expected to produce 55,000 tons of quartz sand in 2022, according to the article, although the final amount was 62,000.
The quartz sand used in crucibles must be 4N5 purity, meaning 99.995%. For 3N purity, there are millions of tons of production capacity, but it’s not pure enough. On the other end of the scale is the even more pure synthetic quartz sand used by the electronics industry but this is both more expensive and may be less durable – used up more swiftly – under the 1,400°C temperature of the Czochralski process, which would exacerbate the cost issue. For the inner face of the crucible which is in contact with polysilicon, a somewhat higher purity of 99.998% is needed, which has led to Western imports being used for that area of crucibles while Chinese product is used for the outer and middle parts.
The Silicon Industry Branch article goes on to state that production of 1 GW of solar wafers in the course of a year requires 88 constantly-running Czochralski furnaces, with crucibles being replaced every 12 days, so 30 crucibles used up per furnace per year, with each crucible weighing 90 kilograms – 90 kilograms of HPQ demand.
If we multiply the relevant numbers from that sentence with an expected wafer demand this year of 500 GW, we get 119,000 tons of demand for quartz sand in 2023, to compare with 2022 output of 62,000, and 100,000 tons production capacity as of August 2022 according to another Chinese industry media source. So 119,000 tons is our estimate for demand, while another prediction from a Chinese agency sees 93,000 tons of demand and a shortfall of 12,000 to 17,000 tons, so they expect only 76,000 to 81,000 tons of production in 2023.
February wafer production grew 15% to 36.4 GW, which annualizes to 436.8 GW, but this is an absolute minimum for 2024 installation demand.