Tesla killed the Solar Roof after a decade and 3,000 roofs. It lost to the cheapest hardware in the industry.

Tesla is done building the Solar Roof, and the reason is a number, not a disappointment. This week the company told its installers that the solar tiles are no longer available to order, and that it will supply only conventional panels from here. Internally, the verdict was that the product was not financially viable. Strip the sentence of its Tesla-ness and it describes something the solar industry has known for years: the beautiful thing lost to the cheap thing, on cost per watt, the way the beautiful thing almost always does.

The Solar Roof was building-integrated photovoltaics — BIPV, in the trade — which means the solar cell and the roof are the same object. Instead of bolting panels onto shingles, you replace the shingles with glass tiles that generate. It is a genuinely appealing idea, and it has been an appealing idea for thirty years, because it solves an aesthetic problem: the roof stops looking like it has machinery on it. Unveiled in 2016 at a rooftop reveal in California — the tiles at that event, it was later confirmed, were not yet functional — it promised roughly $22 a square foot and a roof that paid for itself. The promise did not survive the invoice.

The number that decides it

Everything about why this failed lives in one unit: dollars per watt. A solar system is, at bottom, a machine for turning capital into electricity, and the only question that matters is how much capital per watt of capacity. On that number, the last decade has been the most brutal cost collapse in the history of energy hardware. A crystalline-silicon module that cost around a dollar a watt a decade ago now trades, as a commodity, for somewhere between a dime and thirty cents a watt depending on the market. The panel is no longer the expensive part of a rooftop system. It is close to the cheapest part.

That collapse changed where the money goes, and this is the part the Solar Roof got exactly backwards. In a typical American residential install, the hardware — panels, inverter, racking — is a minority of the bill. The majority is soft cost: labor, permitting, inspection, sales, the crew on the roof for two days. The commodity module won by making the hardware nearly free and letting installers standardize the boring rectangle so the labor got cheaper too. The Solar Roof did the opposite. Every tile is a roofing component and an electrical component at once, so it needs a crew trained to do both, and there are very few of those. A product that makes the most expensive part of the job — skilled labor — rarer and slower is a product fighting its own cost structure.

You can watch the economics in the deployment numbers, which are the only numbers I fully trust, because they record what people actually paid for rather than what they were quoted. Across roughly seven years, Tesla installed about 3,000 Solar Roof systems in the United States, by the estimate of the research firm Wood Mackenzie. At its best the line ran 21 to 32 roofs a week — against a stated goal of 1,000 a week. That is not a slow start. That is a market declining to buy, quietly, one household at a time. One customer's contracted price reportedly climbed from about $72,000 to roughly $146,000 before the roof was installed; the company settled a class action over its pricing in 2023 for $6 million. When the invoice doubles between signature and shingle, the product has a cost problem, not a demand problem.

The Solar Roof made the most expensive part of the job — skilled labor — rarer and slower. That is a product fighting its own cost structure.

Boring won

What replaces it tells you the whole story in one swap. Tesla will now sell a conventional panel: a Tesla-designed 420-watt module, made at its New York factory, about 20.5 percent efficient, all black, low-profile. In other words, a good version of the boring rectangle. It is not prettier than the Solar Roof. It is cheaper per watt, faster to install, and serviceable by any competent crew, and those three things beat pretty every time the buyer is paying the bill themselves. This is the same lesson the grid keeps teaching in every market I cover: the technology that wins is rarely the one with the best rendering. It is the one with the lowest cost per watt that any ordinary installer can put up without a specialist.

Durability is the other number people forget, and it cuts the same way. A rooftop system has to survive twenty-five years of sun and weather, and when something fails, someone has to fix it. A failed panel is unbolted and swapped in an afternoon by anyone with a ladder. A failed tile is a roofing repair and an electrical repair braided together, on a surface where every unit is load-bearing. The commodity module isn't just cheaper to install; it is cheaper to own, because it is cheaper to service. Over the life of the system, serviceability is a cost per watt too, just spread across the years, and it points in the same direction as all the others.

The rectangle is a product for the planet

There is a global point buried in this, and it is the one the American coverage will miss. The reason solar modules got so cheap is that they became a true commodity — manufactured at enormous scale, mostly in China, standardized to the point where a panel is a panel. That commodity is what is electrifying the places I write about: rooftops in Lagos, mini-grids in the Sahel, irrigation pumps that used to run on diesel. None of that runs on integrated glass tiles that need a certified roofer and a six-figure quote. It runs on the cheapest panel that clears the bar, bolted to whatever is already there. The Solar Roof was, from the first reveal, a product for a particular kind of expensive American house. The commodity module is a product for the planet. Only one of those scales.

None of this means integrated solar is dead as an idea, and I want to be fair to it. BIPV can still make sense where you are already buying the expensive surface — a commercial facade, a new-build roof going on regardless, a market where labor is cheap enough that the install penalty shrinks. The idea fails as a mass consumer product for one reason: it has to beat a commodity that fell ninety percent and is still falling, using bespoke manufacturing and scarce labor. That is the wrong side of every cost curve at once. To win, a solar tile would have to get as cheap per watt as a commodity module and as fast to install as a rectangle, and nothing about making the roof and the generator the same object points in that direction.

So here is the arithmetic, which is the only eulogy this product needs. About 3,000 roofs in a decade, against a goal of 1,000 a week. A tile that had to beat a panel costing a dime a watt, using labor that costs more than the panel. A price that could double before the install. Tesla did not kill the Solar Roof out of a failure of nerve or design; it killed it because the number never worked, and the company that keeps every other number in view finally read this one out loud. The cheap, abundant, boring rectangle won. It usually does. On a rooftop, as on a grid, the invoice is the only critic that counts.