In our Q3 2022 global solar capacity figures paper released several months ago, we estimated that around 330 GW of solar modules had been manufactured worldwide in the course of the full year of 2022, which implied around 330 GW would be built in 2023. At the time we tried to shy away from the inevitable conclusion, arguing that perhaps various difficulties, such as those in the supply chain, plus transmission constraints, would limit things to only 280 GW or so this year.
Since then we’ve come round to the obvious, that (roughly speaking) the 330 GW manufactured last year will be built this year. As of last week we’re backed up by Trendforce, which has published analysis arguing that 351 GW will be installed worldwide in 2023, with a 53.4% growth rate. It has 202.5 GW of this being built in Asia-Pacific. And certainly China by itself can build 120 GW. Trendforce has Europe at 68.6 GW this year, but 2022 import figures cause us to expect more like 90 GW. The US, as described in another article in this issue, will be at 40 GW. That’s 250 GW just from US-EU-China. In many places there are photovoltaic projects underway where before there was ZERO activity – which is a step change, not a growth rate.
But what about 2024 and 2025? Is this the last year of such rapid growth? The current high polysilicon price would suggest otherwise.
Since 2022 the price of polysilicon has halved, indicating the expected catchup of supply to demand. But then the price leaped back up by 50%, so to 75% of the peak 2022 price, a volatility caused in part by China’s Spring Festival. We expect the price to fall once more from March through to October, but this resurgence indicated that demand is not overawed even when monthly production of polysilicon is enough to serve 45 GW a month. Perhaps that’s an overestimate – some other elements of the supply chain within China are “only” at 36 GW a month. Those two numbers annualize to 540 GW or 432 GW – and since production capacity is rapidly expanding, let’s take the larger figure.
So 330 GW will be built this year and 540 GW manufactured, to be deployed in 2024 – a second concurrent 50%+ annual growth figure. Then – what about the manufacturing figure for 2024, which is also the deployment figure for 2025 – can it reach 700 GW?
As an aside – there’s some question about the length of the delay between making polysilicon and commissioning a completed module in a solar plants, many of which are far away on the planet, away from China. A delay of 12 months seems a bit long, but it’s what the historical figures appear to indicate. Perhaps it is a little less, but then consider that in a context of production capacity month-on-month, that only increases the figure which must be extrapolated for next year’s installations. So let’s leave it at 12 months for now, pending further examination. Some Chinese industry statistics are much better reported from now on than they were in the past.
If, as we expect, polysilicon product capacity grows from 1.2 million tons to 2 million tons by the end of 2023, then to 2.6 million tons by the end of 2024 – then with a 90% capacity factor for output at those facilities, plus the relatively small imports from the West, that’s 2.17 million tons. At a rate of 2.6 grams per Watt, which may be an overestimate, that is enough to manufacture 834 GW.
Every other part of the supply chain is quicker and cheaper to develop than polysilicon, with development times ranging for 3 months to 9 months for the various different production lines, as opposed to polysilicon 18 months (which was 24 months until it started selling for $40 a kilogram). But let’s say there are some issues there – the world (mostly China) will still be able to make 750 GW. If that’s made and sold, it will be built, in 2025.
The more prosaic outcome is that these vast numbers will not be installed in the first half of the decade but rather in the second half. Surely demand would fail to keep up, surely module prices would drop, and surely there would be overcapacity and bankruptcies among the manufacturers, before they could zoom all the way up to 750 GW a mere two years from now.
But we had to consider the – “what if,” and the flagging of demand also has to happen for a reason. Is 2023, with its over 330 GW of installations, merely catching up from delayed projects of 2021 and 2022? Hardly. But the 500 GW figure of 2024 will probably be that kind of year, with a much smaller growth rate than the current 50% heading into 2025.
The limiting factors that define the maximum possible scale of solar installations are profitability, workforce, and grid connection, with a fourth limit being accompanying energy storage. Land availability exists, but rooftop availability is not yet a limiting factor at the national level, excepting perhaps in Singapore.
If supply of modules and other components is not going to be the limiting factor – clearly it won’t be in 24 months max – then at what point do workforce or grid connection give out? Workforce is stretched more by the rooftop sector, and grid connection more by the utility-scale sector.
In the UK at least workforce has already been stretched to the limit for rooftop installers, even with only a few GW being built annually in a country of 60 million, per our recent interview with Photon Energy, but Brexit is a unique nuisance there. So let’s consider, in a casual manner pending further research, grid connection.
Vietnam in 2020 provided a fun experimental view of what happens when a country crams as much solar as possible onto its grid. The country installed 5 GW one year and then 15 GW the next, and ended up with around 5% curtailment, much higher for certain locales and projects. So 20 GW was how much it could suddenly add. Can we extrapolate from Vietnam to the world as a whole? Not really, but let’s do it anyway. A more serious analysis will be a topic for future articles.
Vietnam’s population is 100 million, and its GDP per capita (PPP) was $12,200 in 2021, while the global average was $18,600. Per capita electricity consumption is one-third of the global average, but the rate of growth is four times the global average. So Vietnam should rate above its population for capacity additions – so let’s multiply by 40, not by the 80 suggested to get from Vietnamese population to global population. If Vietnam can suddenly install 20 GW, the world can suddenly install – 800 GW. And even Vietnam, with a brief ludicrous subsidy of $93.5 per MWh, didn’t cram it quite into a single year.
In a 600 GW+ global annual installation scenario, China can install its customary one-third share – its grid, with around $80 billion invested a year, can handle 200 GW of solar. But other markets, especially the EU and US, will soon hit the buffers on their grids. Some have already, such as the Netherlands. Large-scale transmission upgrades – a couple are mentioned in our Orders section this week, one in Australia’s Victoria state, the other in New York State in the US – take years to complete and represent a much longer delay than the polysilicon bottleneck will prove to have been. Which is another reason the market will swing towards distributed solar.