Compounding rewards patience — nothing else. That is the sentence I keep coming back to when I look at the solar industry’s first half of 2026, because on the surface the numbers look like anything but a compounding story. New domestic PV installations came in at 72.02 GW, down 66.04% year-over-year. The main production chain has been grinding through output cuts across the board. And the industry has responded the way distressed industries do: with a pact.
The prudent question is not “what wins”, but “what survives”. I want to use that lens on this industry, because a 66% installation drop in a headline is a shock, but a 66% drop inside a deliberate, industry-wide consolidation is a different thing entirely. The distinction is where the durable value hides — or fails to hide.
The numbers in order: what the first half actually shows
Let me lay the numbers down in the order a long-horizon investor would read them. First-half installations of 72.02 GW, down 66.04% year-over-year — a brutal reset in demand timing. Output across the main production chain has fallen industry-wide; the word being used internally is “deep adjustment.” This is not a single company’s problem. It is a supply-demand structure in which capacity overshot the deployment schedule, and the market is doing what markets do: pricing the surplus down until someone leaves.
The second number is the policy anchor, and it matters more than the quarterly figure. On August 3, the “15th Five-Year Plan for Building a New Power System” was issued, setting a target of 50% non-fossil power generation by 2030 and keeping national renewable utilization around 90%. That is the twenty-year-horizon number. It says the long-term demand curve for solar in China is not in question. What is in question is which capacity gets to serve it — and at what price.
Now the pact, which is where the industry is trying to manufacture the bottom. On August 6, eight major polysilicon producers — including Tongwei and GCL — signed the “Anti-Involution Initiative,” committing to price their product at no lower than the cost calculated under the Photovoltaic Industry Cost Accounting Model, a standard that formally took effect in August and runs through the whole chain from polysilicon to wafers, cells, and modules. In plain terms: the eight largest suppliers in the segment agreed to stop selling below modeled cost.
I have seen enough industry pacts to know what they are and are not. They are a declaration of the floor the sellers want to defend. They are not a guarantee the floor holds, because a pact is only as strong as the discipline of its weakest member when orders get scarce. But the existence of the pact, and the existence of a formal cost-model standard behind it, is itself a signal: the sellers have agreed on where they think the bottom is, and they have written it down.
What survives this, and what does not
This is where the long-horizon lens earns its keep. In any oversupply-driven consolidation, what survives is not the largest capacity — it is the lowest-cost capacity with a balance sheet that can wait. The cost-model standard matters precisely because it forces everyone to price against the same accounting, which exposes who is genuinely low-cost and who was only surviving on cash. When the floor is enforced, marginal, high-cost capacity bleeds first. That is the mechanism by which a painful reset becomes durable value for the survivors.
Let me think about how this reads on a balance sheet. The industry’s problem has been that every producer was selling at cash cost or below, so nobody earned a return and nobody could fund the next round of capacity. A cost-model floor, if it holds, changes the arithmetic: it lets the efficient producers earn again, it starves the inefficient ones, and it rebuilds the pricing power of the segment as a whole. The losers are the ones who only competed on price because they had nothing else to compete on. The survivors are the ones who had technology, efficiency, or a balance sheet. That sorting is exactly what a consolidation is for.
Over a twenty-year horizon, the story is usually boring. For solar, the boring version goes like this: demand is structurally rising, anchored by policy targets that put non-fossil generation at 50% by 2030; the current cycle is a supply correction, not a demand collapse; and the consolidation — however painful in 2026 — concentrates the market in the hands of capacity that can price at real cost. The asset that compounds steadily through that process is the one that survives it, not the one that flashes brightest during it.
That is the difference between a decision and a hope in this industry right now. A decision is to recognize the first-half number for what it is — a pricing-in of surplus, not a verdict on demand — and to let the consolidation do its work. A hope is to buy the narrative that the pact instantly restores pricing power. The pact defends a floor; it does not restore demand. The second half of 2026, and the recovery in installations, is the real test, and it will take quarters, not weeks.
The quarterly discipline and the long-horizon case
Let me think about how the discipline of quarterly review interacts with the patience of the long horizon, because that tension is the real test for anyone holding solar exposure. The quarterly discipline says: verify that the consolidation is actually progressing — watch the installation prints, watch whether the eight signatories hold the floor, watch whether high-cost capacity actually exits. The long-horizon case says: none of those quarterly details changes the 2030 anchor. The discipline is what prevents a long-horizon position from becoming a hostage to a narrative; the anchor is what prevents the discipline from becoming myopia. Both are needed, and neither is optional.
The honest truth is that this is not a comfortable position to hold in the current quarter. A 66% drop is a painful print, and the temptation is to treat it as a verdict. The defensible counter is to read it as a phase of the consolidation the industry itself chose — eight major producers signing a floor under modeled cost is not the behavior of an industry expecting to disappear; it is the behavior of an industry that intends to survive and to price accordingly. The asset class is not the quarterly print. The asset class is the accumulated installed base, the cost curve, and the policy anchor. Those are the durable numbers, and they are the ones a twenty-year horizon should be priced against.
The portfolio question, stated plainly
Let me state the portfolio question as plainly as it deserves, because that is what the client actually needs to hear. The question is not whether solar is a good industry — it is, and the 2030 anchor makes that defensible on the longest time frame available. The question is whether the current entry point is paying you for the risk you are taking. An industry in deep adjustment is cheap for a reason: the market is pricing in the possibility of more pain, more exits, and a longer bottom than the bulls expect. The disciplined response is not to refuse the industry, and not to chase it; it is to price the floor against the cost model, buy the survivors rather than the weakest balance sheets, and size the position so that a prolonged bottom is survivable rather than fatal. That is the difference between investing in the consolidation and being caught by it.
The twenty-year view, restated: solar’s installed base must grow to meet the 50% non-fossil target, the current cycle is a supply correction inside that growth, and the consolidation will leave the survivors stronger than the pre-consolidation players were. None of that requires the quarterly print to be pretty. It requires the cost floor to hold, the weak capacity to exit, and the demand anchor to remain in place. All three are in the current record, and all three are observable going forward. That is a defensible position, and the right one.
Hedged on the metrics, patient on the asset
I would also keep a hedged position on the utilization number. The 90% renewable utilization target and the 50% non-fossil generation target are not the same metric and do not move on the same clock. Utilization is about the grid absorbing what is built; generation share is about the mix of what is on the grid. Both are bullish for solar over a decade; neither tells you which quarter the current inventory clears. A long-horizon portfolio does not need the exact clearing quarter. It needs to be sure the industry is still pointed at a growing market, and that the surviving capacity is durable.
There is a second hedge worth stating plainly. A pact that ties prices to modeled cost is only as good as the model. If the cost model is generous — if it loads in costs that the most efficient producers do not actually incur — then the floor sits above the real economics, and it will break under pressure. If the model is tight, the floor is defensible. I have not seen the model’s assumptions, so the honest position is to treat the pact as a strong signal of intent and a weak guarantee of outcome. That is not skepticism about the industry; it is precision about the evidence.
Let me also put the demand drop in its proper frame. A 66% year-over-year fall is the sharpest reset the industry has seen in a long time, and it should not be explained away. Part of it is the natural arithmetic of comparison: the first half of 2025 was an unusually high base, inflated by a rush of grid-connected projects. Part of it is genuine softening in near-term deployment. A long-horizon reader does not need to choose between those two explanations; the policy anchor — 50% non-fossil generation by 2030 — tells you the installed base has to keep growing regardless of which year carries the quarterly burden. The question is never whether solar grows; it is which players are still standing when the growth reaccelerates.
The boring, defensible answer is the right one: the industry is consolidating around a cost floor, demand is anchored by policy to a rising long-term curve, and the survivors — the low-cost, balance-sheet-strong capacity — are the durable value in the story. Nothing about the first half changes that. It only changes who is left when the second half of the decade arrives.
The durable answer, in one line: price the bottom against the cost model, not against the narrative, and let the 2030 anchor do the compounding. Everything else is noise that the next consolidation will sweep away.
For the families I advise, the framework stays the same as it has been: do not chase the monthly installation print, and do not mistake a consolidation for a terminal decline. Price the bottom against the cost model, hold the survivors, and let the 2030 anchor do the compounding. That is the defensible position — and it is the right one.