Silicon solar efficiency is closing on a wall, and the wall is nearer than the number usually quoted. The wall I am talking about is a limit on efficiency, i.e., the fraction of incoming sunlight a cell turns into electricity rather than losing as heat or reflecting away, and the question is how high that fraction can physically go. The best silicon cells in production today convert about a quarter of the light that strikes them. The intrinsic ceiling for a single-junction silicon cell, set by Auger recombination, an unavoidable process in which charge carriers crowd together at high density and shed their energy as heat instead of leaving the cell as current, sits at about 29.4%.^1,2 But this is the limit of the material. The design that actually dominates the world’s factories, the tunnel-oxide passivated contact cell, or TOPCon, exhibits a lower upper limit, set by losses built into its layout rather than into the silicon itself. TOPCon best cells are already pressing hard against it: JinkoSolar’s certified champion TOPCon cell reached 27.79% at the end of 2025 ³, and the company’s own roadmap does not expect to cross 28% until 2028. The material itself still has room for improvement, but the chassis most factories are built around has much less left to give.

This distinction, i.e., the limit of the material against the limit of the design, explains why the best results are achieved through a different architecture, that is, the way the cell layers are stacked and physically connected, rather than through a better TOPCon manufacturing process. You cannot push a design past its own ceiling by refining it; you have to change the layout. Most of the contenders are still single-junction, built around one light-absorbing layer of silicon, with the gains coming from rearranging everything around it. Back-contact cells relocate the metal current-collecting lines from the front of the cell, which is exposed to sunlight, to the back, so that they no longer cast a shadow on the energy-generating surface; this configuration holds the efficiency record for single-junction cells, with a certified efficiency of 28.13%. ⁴.

Silicon heterojunction cells, in which the silicon is coated with ultrathin films that seal the surfaces where charge leakage would otherwise occur, achieve an efficiency of around 27%.⁵ Above them all is the tandem route, which abandons the single-material approach: it involves placing a second absorber on top of the silicon, designed to capture the high-energy light that silicon struggles to absorb and wastes as heat, while the underlying silicon handles the rest.

Against that backdrop, the question stops being whether silicon can be pushed further and becomes how the industry proposes to spend its remaining potential… and whether it is really worth pursuing.

Silicon Solar Efficiency: Spending the Last Point

The answer, at least from the industry, is an emphatic yes… and it is worth seeing who is giving it. In early 2026, it arrived in a tight cluster: three papers in Nature Energy within a single month, all of them JinkoSolar collaborations, bundled by the company into one announcement and escorted by a framing commentary from Richter, Benick, and Wolf.⁶. This is worth mentioning before I quote any single efficiency number, since the two studies that set the terms of the current debate ^7,8, one with the Ningbo Institute of Materials Technology, the other with Soochow University, are not two independent groups arriving at the same crossroads from opposite directions. They are one manufacturer planting a flag on every available path at once.

The first path, incremental, a certified 26.66% on a standard industrial wafer was achieved:⁹ the company’s best peer-reviewed result on a full production-format cell, a notch below the higher 27.79% it has separately claimed for an advanced champion cell. Their work systematically optimizes the loss balance in an industrial n‑type TOPCon cell. The cell front employs a shallow, lightly doped boron emitter that minimizes recombination and optical absorption, overlaid with a grid of fine screen‑printed silver fingers that extract the carriers. Reducing the emitter doping improves open‑circuit voltage and blue response, but increases sheet resistance and impedes lateral carrier transport to the contacts. The authors compensate for it by using ultra‑fine silver fingers with widths on the order of 10 micrometres and optimized spacing, limiting the optical shading while maintaining low series resistance. On the rear, a refined tunnel‑oxide/poly‑silicon stack acts as a chemically and structurally robust barrier, inhibiting silver diffusion into the crystalline silicon and preserving passivation. At each step, they deliberately accept a small penalty in one loss channel to enable a larger reduction in another, yielding the net efficiency gain.

In the second path, the authors circumvent the front‑emitter conductivity–recombination trade‑off by implementing high‑quality tunnel‑oxide passivating contact (TOPCon) structures on both the illuminated and rear surfaces of the silicon wafer.⁸ This dual‑sided TOPCon configuration enables a certified 26.34% efficiency in a full‑size bifacial cell, which is engineered to collect photocurrent from irradiance incident on both faces, enhancing energy yield in ground‑mounted systems where albedo significantly contributes to rear‑side illumination. When this bifacial TOPCon device is deployed as the crystalline‑silicon bottom cell in a monolithic perovskite–silicon tandem, and overlaid with an optimized wide‑bandgap perovskite top absorber, the same architecture achieves a certified tandem efficiency of 32.73% under standard test conditions.

Get the Raw Science Brief in your inbox.

Energy systems, critical materials, geopolitics..

The two paths look like a contrast, a modest tune-up against a bolder change of architecture, but that contrast is thinner than advertised, and chasing it is where the economics turn against the whole exercise. Both cells are designed to be manufactured using existing production lines: standard wafer formats, conventional deposition processes, and standard screen-printed silver. Even Gao’s “alternative architecture” is a double-sided TOPCon, not a new device class.

The efficiency race has run into a problem that has not much to do with physics: each additional point is worth less than the last, and it costs more to achieve.

The contest between these two papers is a third of a percentage point. Winning a larger gain by leaving the TOPCon platform means new deposition and metallization tools, line downtime, lower yields through the ramp, and the slow climb back up a fresh learning curve: costs that could offset the value of those additional 0.3 to 0.5 percentage points for years to come. A cell that tops the certification table still has to unseat hundreds of billions of dollars of installed equipment, and dislodging an entrenched manufacturing standard has historically been a decade-scale process, not something that happens between earnings calls.

JinkoSolar’s own rhetoric gives away its intentions. Its press releases repeatedly emphasize terms like “industrial-scale,” “scalable,” or “compatible with current production”… language typical not of a company that sells efficiency, but of one that sells efficiency that doesn’t require upgrading equipment. That, and not the data itself, is where it truly competes.

Which is the last thing to keep in view: who is telling you the race matters. The argument that any improvement, no matter how small, is worth pursuing comes, within this group, from the contextual comments and the manufacturer’s own document: sources with a direct stake in keeping efficiency as the main focus. The improvements are real, and the mechanism is robust… I will return to this later. But we are selling a 0.1-point improvement as a breakthrough… and what the real value of that 0.1 point is, given that the cell is merely a small fraction of the cost of a completed, grid-connected plant, is the question I will address in the following section.

A note on the tandem route, which carries its own unresolved catch. The ceiling is real, and the records climb: LONGi reports a certified 34.85% for a silicon–perovskite tandem. ¹⁰ But the perovskite layer that makes those numbers possible is chemically fragile: think of it as a second filter stacked on the silicon, catching light the silicon wastes, but prone to wearing out in a way silicon never does. The figures from this cluster make the point more honestly than any record can. Gao’s 32.73% tandem retained 80% of its output after 2,000 hours of stress testing,⁸ a standard durability milestone, but a modest one for a device meant to sell power for twenty-five to thirty years, where silicon routinely holds far more of its output far longer. The efficiency certificate and the stability figure appear on the same page, and only one of them is reassuring. The stability of that layer, not the headline number, is the open question… and I would not read any tandem record as a near-term factory number on the strength of a certificate alone.

Where the constraints bind: from thermodynamic limits to the infrastructure

I would classify these constraints into three different scales: the cell, the system, and the grid. At the cell level, the limit is mostly thermodynamic and, as we saw, has almost been reached; at the system level, it is economic; at the grid level, it is the physical infrastructure. The more one analyzes where the limiting constraint actually lies today, the further it moves away from the cell… which is precisely where the records, headlines, and most of the capital continue to point.

The cell and the thermodynamic issue

At the single-cell scale, the remaining work has turned microscopic in the literal sense. A good example is a recent study on the contacts used in TOPCon silicon solar cells, ¹¹ which focuses on a layer approximately one nanometer thick, essentially a few atoms, located between the silicon and one of its metallic contacts. This ultrathin layer has to solve a tricky problem: it must protect the surface so that charge carriers are not lost, while still allowing the useful electrical current to flow through it. It does this mainly because it is so thin that electrons can “tunnel” through, and because the materials around it are designed to guide the right kind of charge carriers and block the others. Tiny defects or openings in this layer can sometimes help reduce resistance if they are very small and well-controlled, but if they are larger or too frequent, they act as faults that let carriers recombine and waste energy. Researchers can now measure and model these features down to nanometer sizes and extremely high densities, and they use that information to squeeze out the last fractions of a percent in efficiency. Imagine a screen door designed to keep insects out without blocking the flow of air, with engineers fine-tuning the mesh thread by thread. So, what more can actually be gained from this kind of work?

In the case of standard silicon cells, the physical efficiency limit is already so close that further improvements typically come in increments of just one or two percentage points, at the cost of increasingly complex designs and manufacturing processes.

However, much of the visible research effort and industry rhetoric continues to focus on these final fractions of a percentage point of cell efficiency, rather than on issues such as durability, recycling, or reducing the total cost of installed solar power.

The system and the economic issue

If we look at the system as a whole, the trends are inverted. The U.S. National Renewable Energy Laboratory reports that the price of the panels themselves fell by approximately 90% between 2010 and 2024; however, the installation cost of a completed solar project remained stable between 2021 and 2024, and increased in some years, as the rush to build drove up the price of land, the studies required to connect to the grid, and on-site construction. ¹²

Costs unrelated to the panels are divided into hardware (mounting racks, wiring, inverters) and what the industry refers to as soft costs: permits, inspections, grid connection procedures, financing, and labor. Soft costs alone run something like 30 to 35% of a large utility-scale project, and reach a share of around 50% in residencial systems.

You can compare it to building a house: while the price of wood has dropped significantly, the cost of land, permits, and the contractor’s labor hours has not, so using cheaper lumber barely affects the final price. As the cost of the cell shrinks to become a minor expense and the surrounding costs refuse to fall with it, a marginal improvement in cell efficiency has less and less impact on the final cost of electricity.

You might reasonably argue that even a 0.1-point improvement in efficiency still reduces energy costs, and that’s true: a more efficient panel produces the same amount of power in a smaller footprint, which reduces area-related costs (mounting structures, wiring, land, labor). That advantage is real, especially in the case of large ground-mounted plants, where the footprint is a determining factor. But it is being reduced on two fronts at once. On the one hand, the area-related costs it affects represent an increasingly smaller portion of a bill, increasingly dominated by indirect costs that efficiency does not affect. On the other hand, the additional electricity it generates is worth less and less precisely where it would have to be sold: on the power grid.

The grid and the infrastructure issue

And at the grid, the constraint is once again a physical one, and has nothing to do with the cell. A grid operator must balance supply and demand on a second-by-second basis; any generated power that has no demand must be shut down, a measure known in the industry as a curtailment. In 2024, California’s grid operator curtailed 3.4 TWh of utility-scale wind and solar output, a 29% jump over 2023, with solar making up 93% of it. Most was spilled in spring, when the sun is strong but mild weather keeps demand low. In Texas, ERCOT, which runs roughly 90% of the state’s grid, curtailed more than 8 TWh of wind and solar in 2024.¹³ In Europe, Germany’s curtailment is dominated by a transmission bottleneck rather than by oversupply: of roughly 8 TWh curtailed in 2025, only ~1.7 TWh was price-sensitive, the rest was grid-congestion redispatch.¹⁴ Spain’s curtailment also reached records in 2025 with roughly 3.6 TWh of renewable generation curtailed because of grid constraints (a 110% jump year on year) and summer curtailment rates peaking near 11%.¹⁵ Table 1 reports price-sensitive (commercial) curtailment (not full curtailment) for some European countries in 2025, per Montel Energy Reports. So what do these figures tell me?

A more efficient solar array that generates even more power at midday in April (when the grid is already overloaded) doesn’t make use of any of that surplus. And the solutions being implemented should focus more on power lines and storage, not on solar cells.

For an overview of the state-of-the-art of electrochemical storage, the reader can have a look at this Raw Science article.

The common thread linking these three scales is likely a shift. At the cell level, the decisive constraint is thermodynamic: the carrier crowding limit and the design ceiling. At the system level, it becomes economic: fixed indirect costs versus a panel price that is trending toward zero. At the grid level, it becomes an infrastructure issue: whether power lines, storage, and market reach allow electricity to be utilized. My interpretation is that the constraint governing the economics of solar power has shifted beyond the cell, and that the efficiency record is now approaching the least significant figure in the chain.

Editorial note

Curtailment figures are drawn from each market’s system operator or specialist analysts and use differing definitions: California and ERCOT report total curtailment, the European table reports price-sensitive (commercial) curtailment only, and Germany’s total is separated into price-sensitive and grid-congestion components. The figures are therefore indicative of scale and cause, not directly comparable line for line.

This article does not aim to address: an in-depth review of the literature on perovskite stability; the polysilicon and silver/materials supply chains; detailed levelized cost models; nor the capital costs involved in converting factories to heterojunction, back-contact, or tandem lines, on which the argument regarding factory inertia is based, but which it does not quantify.

How to cite

Caniglia, G. (2026). The Last Point: Why Silicon’s Efficiency Record Has Stopped Being the Interesting Number. Raw Science.

References

1. Richter, A., et al. (2013). Reassessment of the limiting efficiency for crystalline silicon solar cells. IEEE Journal of Photovoltaics, 3(4), 1184–1191.doi:10.1109/JPHOTOV.2013.2270351
2. Niewelt, T. et al. (2022), Reassessment of the intrinsic bulk recombination in crystalline silicon, Solar Energy Materials and Solar Cells 235, 111467. doi:10.1016/j.solmat.2021.111467
3. JinkoSolar Press Release. 2025. JinkoSolar Breaks World Record Again with 27.79% TOPCon Cell Efficiency.
4. LONGi Press Release. April 2026. 28.13%, 26.4%! LONGi Sets New World Records for Crystalline Silicon Solar Cell and Module Efficiency
5. Xie, Z., et al. (2025). 27%-efficiency silicon heterojunction cell with 98.6% cell-to-module ratio driving new momentum towards the 29.4% limit. Nature Communications, 16, 9421. doi: 10.1038/s41467-025-64465-0.
6. Richter, A., Benick, J. & Wolf, A. (2026). Pushing tunnel oxide passivating contact technology. Nature Energy, 11, 653–655. doi: 10.1038/s41560-026-02051-4.
7. Yang, Z., et al. (2026). Dual-side electrical refinement enables efficient industrial tunnel oxide passivating contact silicon solar cells. Nature Energy, 11, 699–709. doi: 10.1038/s41560-026-01982-2.
8. Gao, K., et al. (2026). Bifacial tunnel oxide passivating contacts for silicon and perovskite/silicon tandem solar cells with improved efficiency. Nature Energy, 11, 710–719. doi: 10.1038/s41560-026-02007-8.
9. Yang, Z., et al. (2026). Dual-sided electrical refinement enables efficient industrial tunnel oxide passivating contact silicon solar cells. Nature Energy, 11, 699–709. doi: 10.1038/s41560-026-01982-2.
10. LONGi. Press Release. 2025. 34.85%! LONGi Breaks World Record for Crystalline Silicon–Perovskite Tandem Solar Cell Efficiency Again.
11. Zhang, W., et al. (2026). Passivating pinholes for large-area and high-efficiency silicon solar cells with tunnel oxide passivated contact. Nature Communications, 17, 2490. doi: 10.1038/s41467-026-70511-2.
12. Ramasamy, V., Zuboy, J., Feldman, D., Narayanaswami, M., Woodhouse, M. & Margolis, R. (2025, January). Documenting 15 Years of Reductions in U.S. Solar Photovoltaic System Costs (NREL/TP-7A40-92536). National Renewable Energy Laboratory.
13. ModoEnergy. 2025. The Curtailment Crisis: Saving wind and solar investments in ERCOT.
14. SMARD. Netzengpassmanagement je Energieträger. Mehr Erhöhung und Reduzierung je Energieträger seit Juli 2022.
15. Aurora Energy Research. 2026. Spain’s Latest Curtailment Surge.
16. Montel Energy. 2026. European price sensitive curtailment report 2025.