Scaling America’s domestic solar supply chain

Start with the shape of the chain, because most people only see its last step. Crystalline silicon solar, which is essentially the whole market apart from First Solar, runs polysilicon to wafer to cell to module. The US has enough module capacity to serve US demand, which Moskowitz puts at 40 to 50 gigawatts a year and probably more, and it has had a decent polysilicon industry for twenty years, on the order of 10 to 20 gigawatts, which he calls a moving target. The middle is the hole. Before the Inflation Reduction Act there were no US cell or wafer factories at all, and roughly 99% of wafers came from China before trade cases pushed that production to Southeast Asia. There is now more cell capacity than wafer capacity, and by his count only two companies are making wafers in the US or about to start. That is why he frames reshoring in four stages: an industry merely surviving before the IRA, an industry that has made the investments and is on a path, an industry self-sufficient for the US market, and, aspirationally, an industry competitive globally and exporting. He places the US at stage two.

The interesting question is why the two ends domesticated first, and the obvious answer turns out to be wrong. Kann offers the natural guess, that polysilicon and modules require the least labor. Moskowitz rejects it for modules. What made module assembly viable here was that panels are bulky, so shipping costs matter, and the process is relatively quick and simple, needing far less capital, chemical infrastructure and localized supplier networks than cells and wafers do. Polysilicon domesticated for a different reason again: its dominant input cost is electricity, which is why plants sit next to cheap hydropower in Washington and cheap generation in Michigan, and it has semiconductor customers alongside solar. The chain did not reshore by labor intensity. It reshored wherever a factory had the fewest interlocking dependencies, which is exactly why the middle is hard.

Then there is the price. Kann puts US modules at around 30 cents a watt against a global average near 10, and Moskowitz does not dispute the comparison, noting separately that panels can often be bought globally at seven, eight or nine cents. Asked to break the gap into its parts, he declines to name a single driver. Building the structure itself costs dramatically more here, for reasons that are not electricity but steel, contracted labor and permitting. The production equipment, by contrast, is commoditized and bought on world markets by everyone. So the difference “becomes an aggregation of factors,” and the manufacturing tax credit designed to close it, which he refers to only by its section number, has narrowed the gap without closing it. He is explicit that the credit was never meant to run forever: its job is to buy enough certainty to reach scale, and scale is the thing the US has not had, a scattering of few-hundred-megawatt plants against multi-gigawatt facilities abroad. Supplier clustering has the same chicken-and-egg shape, and he says the chicken is the factories; suppliers follow demand, the way the southern auto industry accumulated its supplier base over three decades of BMW, Honda, Kia, Hyundai and Volkswagen plants, and the way First Solar cultivated a local glass float supplier in Ohio. Customers, he adds, will not voluntarily pay 30 cents when 10 is available, which is why he keeps returning to industrial policy rather than to persuasion. He is just as insistent that the premium has not made solar uncompetitive against other generation: even US-made at 30 cents a watt, he says, it is still the cheapest and fastest thing to deploy in nearly every scenario, and sits at the bottom of the standard levelized cost charts.

What stopped stage three is two things at once. One is policy implementation: guidance on domestic content and related rules took a long time to issue under one administration and then changed under the next. The other came from outside US policy entirely. China’s property sector crashed, and the country compensated by doubling down on export-oriented manufacturing of solar, batteries and electric vehicles, oversupplying the global market and undercutting the investment case just as US factories were being planned. Which produces his answer for the single hardest thing about the middle of the chain: the general economics. Even inside a high-price, tariff-protected market you compete globally, because nearly every manufacturer including the Chinese ones has now built a US factory, so a module with US-made cells competes against a module with imported cells on the same shelf. Financing that requires customers willing to commit as offtakers for twenty years, in an industry where, as he notes a Bloomberg analyst would immediately point out, a factory’s economics typically turn over in five. His two rules of thumb are that demand grows faster than people expect and prices fall faster than people expect, and he says the second one has changed over the last three or four years. Unlike aviation or pharmaceuticals, there is no grace period: solar panels are a mature, cheap, genuinely excellent product, so a new entrant earns no premium for novelty and has to compete from the first day.