A tracker published this week counts what the American solar manufacturing build-out has actually produced, stage by stage. The headline numbers are large. Cell capacity announced or under construction: 55.90GW. Modules: 41.36GW. Polysilicon and ingots: 22.1GW.
Wafers: 13.3GW. That is every stage of the pipeline added together — announced, in construction, near completion and operating.
A cell is made from a wafer. There is no route from polysilicon to a finished cell that does not pass through one.
What the gap means in practice
Read the chain in order. Polysilicon is refined into ingots. Ingots are sliced into wafers. Wafers are processed into cells. Cells are assembled into modules.
The American pipeline has substantial capacity at the top of that chain and very substantial capacity at the bottom, and a narrow waist in the middle. Fifty-six gigawatts of cell capacity fed by thirteen gigawatts of domestic wafer capacity means that most of those cells, if they are all built and all run, will be processing wafers made somewhere else.
Which is a legitimate industrial position — plenty of countries assemble things from imported intermediates and are richer for it. It is simply not the position the policy was sold as achieving. "Domestic solar manufacturing" describes a chain, and a chain with an import in the middle has the same single point of failure it had before, moved one step.
The operating numbers are smaller again
There is a second discount to apply, and the tracker supplies it. Announced capacity is not producing capacity.
Of 36GW of announced polysilicon capacity, 15.5GW is producing — about forty-three percent. Of 5GW of announced wafer capacity, 3.2GW is producing. Everything else sits at one of three earlier stages: announced and in planning, under construction, or near completion.
So the honest description of domestic wafer capability today is roughly three gigawatts operating, against a cell pipeline nearly twenty times that size. And the stage breakdown offers no relief: wafers are the smallest figure not only in total but at every individual stage. There is no large tranche of wafer capacity sitting in construction waiting to close the gap.
Why the middle is always the part nobody builds
This is a pattern rather than an accident, and it has an economic explanation.
Module assembly is the cheapest step to stand up, the fastest to build, the most labour-visible, and the one that most easily satisfies a domestic-content rule. It is where an incentive scheme aimed at jobs and announcements naturally lands. At least $3.2bn of disclosed investment backs about 19GW of new module capacity; at least $5.8bn backs about 21.46GW of cell capacity.
Wafering is capital-intensive, technically exacting, low-margin, and invisible in a ribbon-cutting. It also competes directly with an incumbent industry operating at enormous scale and low cost, which means a domestic wafer plant needs either a protected price or a subsidy sustained for longer than an election cycle.
Given a choice of where to put money that must show a result, everyone chooses the end of the chain. Nobody chooses the waist, and the waist is what determines whether the chain exists.
The same shape, three industries in a week
This desk has now written the same structural story three times in ten days with different nouns.
In artillery, the constraint is not the gun but the energetic chemistry upstream of the shell — nitrocellulose and nitroguanidine, an unglamorous continuous-process business consolidated into very few sites. In data centres, it is not the chips or even the power but the switchgear the power passes through, fabricated metal on a lead time measured in quarters. In European batteries, the plants were financed and the grid connections were not.
Every one of them is the same finding: capital and attention go to the visible end of a chain, the boring intermediate step is left to somebody else, and the boring step turns out to set the rate. Solar wafers are that step for photovoltaics, and the tracker has now put a number on how thoroughly they have been skipped.
The figure to follow
Not announced gigawatts, which is a press-release unit.
Watch operating wafer capacity as a percentage of operating cell capacity. Today that ratio is somewhere near a fifth. If it climbs toward one, a domestic chain is being built. If cell and module capacity keep growing while it falls, the build-out is producing assembly capacity that raises rather than reduces dependence on imported wafers — and the strategic argument made for the whole programme will have been quietly inverted while the jobs numbers kept rising.
The capacity figures — 55.90GW of cell, 41.36GW of module, 22.1GW of polysilicon and ingot and 13.3GW of wafer capacity announced or under construction; 36GW of announced polysilicon capacity against 15.5GW producing and 5GW of announced wafer capacity against 3.2GW producing; the stage-by-stage breakdown; and the disclosed investment figures of at least $3.2bn behind about 19GW of module capacity and at least $5.8bn behind about 21.46GW of cell capacity — are from PV Tech Research's US manufacturing tracker as published by PV Tech on 2 September 2026. The characterisation of wafer capacity as the smallest number by a wide margin is PV Tech's. The analysis is our own.
Topics businessmanufacturingsupply chainindustrial policy
