By: Luis Reyes
Published: Aug 27, at 12:00pm ET
The glass on the front of a solar panel is the one component that does nothing. It stops hail, keeps water off the cells and lets light through, and that is the entire job description. The black rectangles underneath are where the electricity actually comes from.
A company in Baldwin Park, California has spent a decade arguing that this is a waste of a perfectly good sheet of glass.
Caelux coats that cover glass with a thin perovskite film, so the sheet itself becomes the first layer that generates and the silicon cells below become the second. The product is called Active Glass. The panel it produces is what the company calls a hybrid tandem.
Since April, Caelux has signed 13 gigawatts of that glass across three module manufacturers on two continents. And in an interview published this week, CEO Scott Graybeal told PV Tech that “every module in the world in the 2030s will have perovskite.”
That is a very large claim built on a material whose longest stretch of outdoor evidence, offered by Graybeal in the same interview, runs to six weeks.
Almost everyone else chasing perovskite builds it into the cell. Oxford PV, Qcells and LONGi all lay perovskite films directly on top of a silicon base cell and seal the sandwich inside glass. It is a genuine tandem cell, and it requires a genuinely new production line.
Caelux does not touch the cell at all. The silicon underneath stays exactly what it was, and the perovskite rides on the piece of glass that was going on top anyway.
Graybeal’s stated reason is return on invested capital. Building out module manufacturing to go with the glass would have pushed revenue further into the future and raised the cost of getting there, to the point where the returns stopped looking like returns. Selling one component into lines that already exist gets paid sooner.
No spam. Unsubscribe anytime. Privacy policy (opens in new window)
He is also open about what the approach costs on the physics side. In a conventional tandem cell, light passes through the perovskite and reaches the silicon directly. In Caelux’s version, it has to travel through an extra transparent conductive layer on the way down, and traditional architectures can therefore reach higher total efficiencies.
The compensation, by his argument, is failure mode. Separate the perovskite from the cell and a degrading top layer does not take the silicon with it. You lose the bonus, not the panel.
The US buyer for all of this is Solx, a module manufacturer building inside a former Hewlett-Packard plant in Aguadilla, Puerto Rico, on a site the company first announced in 2024 with construction starting in April 2025.
On April 21 the two companies announced a five-year, 3-gigawatt partnership. Caelux’s glass goes on top of Solx’s Aurora module, which the companies say produces 28 percent conversion efficiency and up to 30 percent more power density than a silicon-only panel of the same size.
“U.S.-made hybrid tandem is no longer a theoretical, it’s now in commercialized production,” said James Holmes, co-founder and CEO of Solx, in the announcement. Broader commercial availability is targeted for 2027.
The rest of the panel is domestic too, which is the entire commercial point. The silicon cells come from Suniva, headquartered in metro Atlanta, and the steel frames from Origami Solar.
Suniva is worth a moment on its own. The company announced in April that it is putting over $350 million into a 620,000-square-foot cell plant in Laurens, South Carolina, opening in the second quarter of 2027 with 4.5 gigawatts of capacity and 564 jobs. Added to its existing Georgia line, that takes it past 5.5 gigawatts a year.
American module assembly has never been the bottleneck. Cells have. Panels stamped as domestic have leaned on imported internals for years, and a merchant cell supplier at that scale is what lets a mid-sized assembler clear federal domestic content thresholds. Caelux is selling into exactly that gap: a perovskite layer that arrives as a piece of American glass rather than as a redesign of somebody’s factory.
Perovskite has always had the same problem. It converts light beautifully and it falls apart when exposed to light, air, moisture and heat, which are the four things a solar panel spends 25 years sitting in.
Graybeal says Caelux has worked the chemistry hard, through additives, structural changes and process adjustments, and that its current technology can support 25-year lifetimes. He also says the company is in the middle of qualification and certification.
The supporting anecdote he gave PV Tech was that Caelux has had its perovskite material outdoors and has seen no degradation over six weeks.
Six weeks is real data. It is not 25 years, and nobody in the industry has produced 25 years for perovskite yet.
That is why the certification paperwork matters more than any efficiency number right now. Qcells spent this summer collecting exactly that kind of document, and put a tandem panel through 42 days of damp heat and 200 thermal cycles to get it. No world record attached, just the file that makes a bank willing to finance a field.
The lab-to-field gap is the recurring story here. We covered it again this month with a see-through perovskite cell whose headline efficiency came from a smaller device under a lamp than the pane that went outside.
On July 14, LONGi announced a crystalline silicon-perovskite tandem cell at 35.5 percent, certified by the European Solar Test Installation in Italy. It is a world record, and it beats every commercial module on the planet by a wide margin.
It is also a laboratory device. LONGi did not disclose the active area, and as pv magazine noted, the company said in June that it has no active mass-production plan for the technology. Its certified tandem modules, the ones built at something like commercial size, sit at 31.4 and 29.4 percent.
So the record holder has a 35.5 percent cell and no factory schedule. Caelux has a 28 percent module and 13 gigawatts of five-year commitments. Those are two completely different bets, and only one of them has to survive a warranty claim.
Caelux’s first manufacturing site stays in the United States, with India as the likely second. The two Indian deals, with Rayzon Solar and Navitas Solar, are each 5 gigawatts over five years and are aimed at production starting around 2028 to 2029, which is the runway Graybeal says the company needs to build or convert a plant there.
Between now and then, the number that decides everything is not efficiency. It is whether TÜV Rheinland or an equivalent body signs off on the full reliability and safety standards, and whether an insurer will stand behind the resulting panel for the life of a project.
There is a smaller consequence worth noting for anyone who operates solar hardware. Once the front sheet is the first power layer, the glass stops being the disposable part of the panel. It is already the surface that costs you output when it gets dirty, which is why desert farms now run robots across it every night. Coating it with an active semiconductor raises the stakes on every hailstone, every dust storm and every cleaning contract.
The engineering logic here is hard to argue with. A manufacturer can drop Caelux’s glass into an existing line without redesigning a single cell, which is precisely why three of them signed inside four months. Whether that logic holds in a field is a different question, and answering it takes roughly 200 times longer than the longest run anybody has watched so far.
Agree or laugh out loud?
Luis Reyes · Aug 14, 2026
Luis Reyes · Aug 10, 2026
Luis Reyes · Aug 23, 2026
Luis Reyes · Aug 18, 2026
Luis Reyes · Aug 8, 2026
Luis Reyes · Aug 9, 2026
Luis Reyes · Aug 27, 2026
Chema Bonilla Díaz · Aug 27, 2026
Luis Reyes · Aug 27, 2026
Chema Bonilla Díaz · Aug 27, 2026
Olivia Richman · Aug 27, 2026
Autonotion is the English-language automotive editorial by Autonocion.com — car news, reviews, and industry analysis for American readers.
Other links
Company
Subscribe
Get the latest car news in your inbox:
By submitting your email you allow autonocion.com to send you news or promotions. More info

source