Perovskite Tandem Photovoltaics

Established A perovskite–silicon tandem stacks a metal-halide perovskite cell, band gap tuned near 1.67–1.70 eV, on top of a silicon cell at 1.12 eV, splitting the spectrum so each junction converts its share at higher voltage than silicon manages alone. The physics is not in dispute. Single-junction silicon is pressed against a practical ceiling of 29.4% set by Auger recombination (Richter, Hermle and Glunz, 2013); the certified silicon cell record sits near 27%; a two-terminal tandem lifts the radiative ceiling into the low forties, with realistic device limits usually put near 39–40%. Perovskites won the top-cell job because they deposit from solution or vapour at low temperature, tolerate defects unusually well, and tune their band gap by composition.

Established The headline number — a certified tandem cell above 34% — is real, and it is the least informative number in the field. It was measured on roughly one square centimetre of designated illumination area under standard test conditions. Whether tandems matter is decided by different numbers: the certified efficiency of a full-size module; the loss incurred scaling from cell to module; whether an encapsulated product passes the damp-heat and thermal-cycling protocols silicon passes routinely, and what passing does and does not prove; whether measured lifetimes support a 25-to-30-year warranty; manufacturing yield against silicon lines running above 98%; what happens to the lead; and whether a lender and an insurer will put money behind any of it. This brief is organised around those numbers, in that order.

Frontier The field’s honest position in the mid-2020s: records run years ahead of products. One company has shipped commercial tandem modules to a paying customer; audited industry-wide perovskite shipments are measured in megawatts, against a silicon industry shipping on the order of six hundred gigawatts a year at spot prices that fell to roughly US$0.09–0.12 per watt in 2024–2025. A verification note this site owes its readers: the record behind this brief was checked to early 2026; entries the review could not re-verify since are flagged as such rather than silently updated.

Established Rung one of the ladder: certified small-area cells. The two-terminal perovskite–silicon record, on devices of about 1 cm² designated area, climbed from 29.15% (Helmholtz-Zentrum Berlin, January 2020) through 29.5% (Oxford PV, December 2020), 32.5% (HZB, 2022) and 33.7% (KAUST, mid-2023) to 33.9% (LONGi, November 2023) and 34.6% (LONGi, announced June 2024 with independent certification). These are certified entries of the kind carried on the NREL research-cell chart and in the Green et al. efficiency tables — measured by accredited laboratories, not vendor arithmetic. Vendor announcements during 2025 pushed the claimed record to roughly 34.9%; this review could not re-verify the 2026 state of the chart and says so rather than guessing.

Established Rung two: commercial-area cells, where the first tranche of scaling loss appears. Oxford PV reported 28.6% on a 258 cm² commercial-wafer cell certified at Fraunhofer ISE in May 2023 (vendor announcement, independently certified). Hanwha Qcells announced a certified 28.6% on a full-area M10-format cell in April 2025, stated to use mass-production-compatible processes (vendor). LONGi reported a 30%-class full-wafer device in 2024 (vendor). Read the gap plainly: about six absolute points separate the 1 cm² record from the best certified commercial-area cell. That is not fraud; small designated-area measurements are how every PV technology’s limits are probed. But it is the first of three discounts between headline and product.

Established Rung three: modules. The best certified full-format tandem module this review can document is Oxford PV’s 24.5% residential-size 60-cell module, certified at Fraunhofer ISE CalLab and announced in January 2024 (vendor announcement, independently certified). For comparison, premium mass-market silicon modules ship at roughly 22–23.5%, and the best silicon module records sit near 25%. So the demonstrated tandem module advantage over the best shipping silicon was, as of early 2026, on the order of one to two and a half points — not the eleven points a naive reading of cell records suggests. Trinasolar announced an 808 W industrial-format tandem module in late 2024 (vendor figure; aperture-area convention not consistently stated in coverage), which if borne out puts industrial formats in the mid-to-high twenties.

Frontier Rung four: delivered product — the number the editorial line of this brief turns on. In September 2024 Oxford PV announced the first commercial shipment of perovskite–silicon tandem modules, from its roughly 100 MW-capacity Brandenburg an der Havel line, to an undisclosed United States customer, in an undisclosed volume (vendor announcement; no third-party audit of volume or field performance is public). That shipment is, as far as this review can establish, the delivered record of the entire tandem industry as of early 2026. Reports during 2025 described Oxford PV licensing its technology to Trinasolar for manufacture in China; this review treats that as reported rather than confirmed. Meanwhile China’s single-junction perovskite firms — Microquanta, UtmoLight, GCL’s perovskite arm, Renshine — operate lines in the 100 MW class, have announced gigawatt-scale plants, and have fielded demonstration installations; cumulative deployed perovskite capacity worldwide remains in the tens of megawatts, overwhelmingly single-junction rather than tandem.

Established The stability protocol state, stated precisely. The qualification gate for terrestrial modules is IEC 61215 (design qualification and type approval; the 2021 edition is current) plus IEC 61730 (safety). The suite includes 1,000 hours of damp heat at 85 °C and 85% relative humidity, 200 thermal cycles between −40 and +85 °C, humidity-freeze cycles and UV preconditioning, with a pass requiring no more than 5% power loss plus intact safety. Several manufacturers state that perovskite or tandem products have passed — Microquanta was among the first to claim IEC 61215/61730 passes for single-junction perovskite modules around 2020, and Oxford PV states its tandem module passed the suite (vendor statements; certificates exist but the underlying test data are not published).

Established What passing IEC 61215 does not mean. IEC 61215 is an infant-mortality screen calibrated on silicon failure modes, not a lifetime rating. Silicon’s 25-to-30-year warranties rest on the correlation, built over four decades of fielded fleets, between passing that screen and median field degradation around 0.5–0.6% per year (the Jordan and Kurtz reviews). No such correlation exists for perovskites, and some of their known degradation modes — halide segregation under light at operating temperature, ion-migration-mediated damage under reverse bias — are not what the suite was designed to exercise. A perovskite pass is necessary and genuinely hard-won; it is not 25 years.

Frontier The T80/T90 literature, read honestly. T80 and T90 are the times to 80% and 90% of initial (or post-burn-in) performance. The best measured — not extrapolated — operational results for encapsulated cells are in the low thousands of hours at elevated temperature: damp-heat endurance beyond 1,000 hours at 85 °C/85% (Azmi et al., Science 2022, and successors), and maximum-power-point operation for 1,000–5,000 hours at 40–65 °C with under 5–10% loss in the strongest reports. Claims of tens of thousands of hours are extrapolations through Arrhenius acceleration factors that are themselves contested, because several mechanisms with different activation energies overlap. The ISOS consensus protocols (Khenkin et al., 2020) exist precisely because shelf-life tests in nitrogen were being tabulated beside operational tests in air; even now, many champion papers report T90 under mild conditions where T80 under realistic ones would look worse. Outdoor data are sparse: small arrays, one to five years, with degradation rates spanning roughly 1% per year for the best to above 10% per year for the worst.

Established The cost bar every tandem must clear. Unsubsidised United States utility-scale solar cleared at $38–78 per MWh in Lazard’s June 2025 assessment, and IRENA’s global benchmark for 2025 puts solar photovoltaics near $44 per MWh. With silicon modules near US$0.09–0.12 per watt, a tandem’s extra efficiency is worth only what it saves in area-related balance-of-system cost — commonly estimated at a few US cents per watt per absolute efficiency point in BOS-heavy markets. That is the entire budget available to pay for every extra deposition step, every point of yield loss, and every hour of accelerated testing.

Frontier Is the wide-gap top cell photostable at product level? Mixed iodide–bromide perovskites at 1.67–1.70 eV are prone to light-induced halide segregation — the Hoke effect, measured in 2015 — which splits the band gap and bleeds voltage. Compositional fixes (caesium–formamidium alloys, reduced bromide, 2D capping layers, additive passivation) have suppressed it dramatically in champion cells; whether it stays suppressed over 30 °C-to-85 °C operating years in a laminated module is the single most consequential open materials question.

Frontier Does any accelerated test predict the field? Silicon earned its acceleration factors empirically over decades. For perovskites the activation-energy spectrum is unresolved, stressors interact nonlinearly (light plus heat plus bias does more than the sum), and reversible metastability contaminates short measurements. Until a test protocol is validated against outdoor outcomes on the same module population, every lifetime number in the literature is an argument, not a measurement.

Frontier Does the STC advantage survive outdoors? Two-terminal tandems are current-matched at standard test spectrum; real spectra drift red and blue daily and seasonally, clipping the advantage, while perovskites’ gentler temperature coefficient (around −0.25%/°C against roughly −0.30%/°C for good silicon) pushes the other way. Early energy-yield studies suggest tandems keep most but not all of their nameplate edge; fleet-scale kWh-per-kWp data do not yet exist in public.

Frontier Reverse bias is an unsolved product hazard. Shaded perovskite junctions break down destructively at a few volts reverse bias, against roughly −15 V for silicon, so a partially shaded string can permanently damage tandem cells. Cell-level mitigation, bypass architecture and safe breakdown engineering are active research, not settled practice.

Frontier Is the tandem partner even settled? All-perovskite tandems reached roughly 29–30% certified on small areas by 2024–2025, promising lighter, cheaper stacks, but their tin-containing narrow-gap cell oxidises readily and trails badly on stability. Alternative wide-gap partners (CIGS, III–V at falling cost) remain live. The silicon-bottom architecture leads because it borrows a terawatt industry, not because physics prefers it.

Established Scaling loss has an anatomy, and each part is measurable. Going from 1 cm² to a module costs: film-uniformity loss, because solution- or vapour-coated perovskite must hold thickness and composition across square metres where a few percent variation shifts band gap and current; defect statistics, because a single pinhole shunt that a lab cell would never contain appears once per square decimetre on an imperfect line; interconnection dead area; and the resistive and optical compromises of full-area transparent electrodes. The certified ladder in section 2 — 34.6% at 1 cm², about 28.6% at commercial cell area, 24.5% at module level — is the integrated measurement of all four.

Established Nobody publishes tandem line yield, and the absence is informative. Incumbent silicon lines run above 98–99% yield after decades of learning. A monolithic tandem adds roughly ten or more process steps — each multiplying yield — on a substrate that cannot be reworked. No manufacturer had, as of early 2026, published line yield, binning distributions or cost curves for tandem production. A field that publishes efficiency records weekly and yield never is telling you where the problem lives.

Established Encapsulation is a genuine materials constraint, not packaging detail. Perovskites tolerate lamination only in a narrow thermal window — standard crosslinking encapsulants near 150 °C sit at the edge of what the absorber survives — and demand glass–glass builds with low-permeation edge seals such as polyisobutylene. The moisture-ingress budget of a 30-year product is orders of magnitude tighter than organic-electronics packaging routinely delivers at PV cost.

Frontier The bifaciality forfeit. Utility-scale silicon is now largely bifacial, harvesting 5–15% extra energy from rear-side light. Most tandem designs are monofacial — the perovskite stack and its electrodes occlude the architecture — so part of the nameplate advantage is silently returned in the field. Bifacial tandem concepts exist in the laboratory; none is a certified product.

Frontier Inputs: high-purity lead iodide and organohalide salts at gigawatt volume, and indium. Champion results depend on precursor purity that today’s speciality-chemical volumes supply in kilograms, not the thousands of tonnes a terawatt industry would draw; and common tandem stacks lean on sputtered indium-based transparent oxides in places silicon mostly does not, adding a supply exposure the incumbent avoided. Neither is a hard wall; both are un-built supply chains — and they would be built in a minerals landscape where, by the IEA’s 2025 accounting, more than half of a broad group of energy-related minerals already sit under some form of export control.

Established Even measuring the product is harder. Perovskite metastability means certified power must be a stabilised value under maximum-power tracking, not a flash measurement; module-scale stabilised certification capacity at accredited labs is limited and slower than the flash lines silicon uses, which is one quiet reason module records lag cell records by years.

Established Tandems inherit their bottom half from the silicon roadmap. The leading architectures build on heterojunction and TOPCon silicon platforms, so every improvement and every cost decline in mainstream silicon flows into — and simultaneously raises the bar for — the tandem business case. This double dependency is the strategic signature of the field.

Established Enabling threads from outside PV. Atomic-layer-deposited buffer layers (tin oxide especially) that protect the perovskite during sputtering came from semiconductor tooling; barrier films and edge-seal chemistry come from OLED and insulated-glass industries; and the two-step hybrid vapour–solution processes that conformally coat industrially textured silicon are adaptations of thin-film deposition practice.

Frontier Degradation science is the load-bearing research dependency. Ion migration modelling, operando microscopy of halide segregation, and machine-learning screens over composition space are active; what the field lacks is not activity but a validated bridge from any of it to fielded-module lifetime.

Frontier Standards drafting is itself a dependency. IEC technical committee work on perovskite-specific test methods — stabilised power measurement, light-soak protocols beyond the silicon-era suite — was in progress as of early 2026; products will be certified against whatever this work converges on, so its pace partially gates the market.

Frontier The single result most able to change this assessment is a published, third-party, multi-year degradation dataset on production tandem modules operating in customer fields alongside silicon controls. Not champion cells, not chamber hours: fielded product, independent instruments, several years, published rates. The instrument for it exists — the US Department of Energy’s Perovskite PV Accelerator for Commercializing Technologies (PACT), led from Sandia with NREL, has fielded perovskite modules at multiple test sites since the early 2020s — but production tandem modules only began shipping in September 2024, so the decisive multi-year readout on real product cannot be complete before late this decade. If it shows fleet degradation near 1% per year, tandem bankability follows almost mechanically; if it shows 3% or worse, the field returns to materials work for years.

Frontier The correlation experiment: one module population, split across accelerated tests and the field. Run identical production modules through the ISOS and IEC matrices while their siblings sit outdoors, then derive acceleration factors from the pairing. This is PACT’s explicit design and the only known way to convert chamber hours into warranty years; partial results exist, a validated protocol did not as of early 2026.

Established The disclosure that would count as an experiment: an audited yield and cost curve from a gigawatt-class tandem line. Every other number in this brief can be excellent and the technology still fail here. A single manufacturer publishing line yield, binning spread and cost per watt — as silicon’s learning-curve literature has for decades — would move the assessment more than any cell record.

Frontier The policy-result: a 25-year performance warranty on a tandem product with a named insurer behind it. Warranties are priced bets on lifetime by parties who lose money when wrong. The first insurance-backed long-tenor tandem warranty would be the market’s first honest lifetime number, and none existed publicly as of early 2026.

Established A round-robin on module-scale certification. Stabilised-power measurement of metastable large-area devices is young; the same tandem module certified at, say, Fraunhofer ISE CalLab, NREL and JET, with results published, would anchor the module ladder the way decades of round-robins anchored cell records.

Established The production architecture is monolithic two-terminal on industrial silicon wafers, and it forces hard choices. Conformally coating perovskite over industrially textured, light-trapping silicon requires hybrid two-step processes (evaporated inorganic scaffold, solution-converted organics) or planarised front surfaces that sacrifice optics; recombination junctions and sputter-buffer layers must survive transparent-electrode deposition; and every one of the added layers must co-anneal within the absorber’s narrow thermal budget.

Established Yield discipline replaces rework. A defective perovskite layer cannot be stripped and recoated economically on a finished tandem stack, so the line needs inline photoluminescence and electroluminescence metrology able to see buried-layer defects at coating speed, with statistical process control tight enough that scrap does not eat the efficiency premium.

Frontier Warranty engineering is product engineering. A product whose power may drift needs conservative nameplate binning, replacement-stock strategy, monitoring-friendly electronics, and reverse-bias protection sized to perovskite breakdown — module-level electronics or denser bypass diodes — all of which cost watts or cents against the very premium the tandem exists to earn.

Established Lamination and glass–glass builds set the factory. Low-temperature encapsulants, polyisobutylene edge seals and dual-glass handling change cycle times and capital equipment relative to a silicon line; tandem factories are not silicon factories with one extra coater, and the capital-expenditure delta is part of the unpublished economics flagged in section 4.

Established Cheap high-efficiency PV is the benchmark that disciplines solar fuels. The artificial photosynthesis brief on this site records PV-plus-electrolysis at 30% solar-to-hydrogen averaged over 48 hours in a III–V/PEM demonstration and about 20% at commercial scale — the standing bar every integrated solar-fuel device fails to clear. Every point tandems add, and every cent they remove, raises that bar further.

Frontier Space power wants what tandems may become, not what they are. Thin-film perovskite tandems promise specific power above 1 kW/kg and have shown encouraging radiation tolerance in early flight and proton-beam experiments, which is why they recur in space-based solar power concepts; space qualification is a separate, unstarted certification ladder, and nothing in this brief’s terrestrial record transfers automatically.

Established Storage siblings set the value of a marginal point. As energy storage revolutions and advanced battery technologies document, systems increasingly pay for firm delivery rather than peak watts. The measured 2025 record makes the point concretely: France, Germany, the Netherlands and Spain each saw roughly 6% of hours at negative wholesale prices; Germany curtailed about 9.6 TWh, around 4% of renewable output on a total-curtailment definition; and on a single Californian day, 29 March 2025, over 58,000 MWh of solar was curtailed while batteries absorbed nearly 39,000 MWh. In hours like those, an extra efficiency point earns little; where area is scarce — rooftops, vehicle skins, dense grids — it earns most. Tandems’ first real markets are the area-scarce ones.

Frontier Spillover runs both directions. Perovskite optoelectronics feed LEDs, detectors and indoor PV; conversely, the OLED industry’s barrier films and the display industry’s large-area coating know-how are the quiet suppliers of any future tandem gigafactory.

Established The certification institutions exist and function. Accredited laboratories — NREL, Fraunhofer ISE CalLab, JET, ESTI among them — certify records; the Green et al. efficiency tables and the NREL charts curate them; IEC TC82 owns the module standards. This machinery is why the cell-record ladder in this brief can be trusted at all, and it is silicon’s gift to its challenger.

Established Bankability is a chain of named parties, and the chain is unassembled. A financed project needs IEC certificates, extended testing of the kind Kiwa PVEL’s scorecard reports (protocols at two to three times IEC durations), an independent engineer’s sign-off, a warranty, and an insurer or investment-grade balance sheet behind the warranty. As of early 2026 no perovskite tandem product had publicly assembled the full chain; the instruments all exist, exercised for decades on silicon and once before on thin-film entrants.

Established Public programmes are explicitly aimed at the gap this brief describes. The US DOE’s PACT centre exists to field-test and validate perovskite modules toward bankability; DOE’s solar office has run repeated perovskite funding rounds; European and Chinese programmes fund pilot lines. The notable institutional fact is the alignment: public money is pointed at lifetime validation and scale-up, not at efficiency records, which markets already fund.

Frontier The vendor-claim economy needs naming. Most numbers a reader encounters about this field originate in company announcements — certified records announced by press release, IEC passes asserted without published data, factories announced without shipment follow-up. This brief marks every such number (vendor); an institutional reform as simple as an industry norm of publishing certificates and yield data would move the field’s credibility more than the next record.

Established The lead question, quantified rather than waved. A perovskite layer carries on the order of 0.4–0.8 grams of lead per square metre — roughly an order of magnitude less than the several grams per square metre of lead in the solder of a conventional silicon module. The honest difference is speciation: solder lead is metallic and largely immobile, while degraded perovskite yields soluble lead halides that rain can leach from a shattered module. Broken-module leaching experiments measure real release; on-device sequestration coatings captured above 96% of leached lead in laboratory tests (Li et al., 2020), and are not yet standard product practice.

Established The regulatory state is a carve-out, not an assessment. The EU’s RoHS directive excludes photovoltaic panels from its lead restriction by explicit scope exception, so lead-based perovskite modules are lawful to sell in Europe — a decision made for silicon-era solder, inherited by perovskites without fresh evaluation. The WEEE regime obliges take-back and recycling in the EU; recycling of glass–glass perovskite laminates at scale is undemonstrated anywhere.

Frontier Lead-free alternatives are real chemistry and poor products. Tin-based perovskites reach roughly 14–15% certified with markedly worse stability as tin(II) oxidises; as of early 2026 no lead-free perovskite approaches tandem-relevant performance. Choosing lead is currently choosing the technology existing at all — which is an argument for sequestration and end-of-life engineering, not for pretending the trade-off away.

Frontier A publication-culture ethics item specific to this field. Stability reporting has repeatedly tabulated incomparable numbers — shelf life in nitrogen beside operation in air, T90 under mild light beside T80 under heat — a practice the ISOS consensus and journal checklists were created to end. Overclaiming here is not victimless: it feeds investment cycles and policy expectations that the delivered record, one shipment as of early 2026, cannot support.

Established The stakes are multiplicative, which is why the field is worth its hype discipline. Photovoltaics passed roughly two terawatts cumulative in the mid-2020s with annual additions above half a terawatt, inside a renewables fleet IRENA counted at 5,149 GW at end-2025 after a single-year addition of 692 GW; against a base that large, a durable 20% relative efficiency gain compounds into continent-scale savings of land, glass, steel, transport and installation labour. Efficiency is the one PV variable that cheapens everything downstream of it at once.

Frontier The forecasting record argues for taking upside seriously. An ex-post analysis of every IEA World Energy Outlook from 1993 to 2022 found solar deployment persistently underestimated — the direction and persistence are defensible even where the magnitude is argued by interested parties. Institutions planning grids, trade and industrial policy have systematically under-planned for solar; a working tandem industry would repeat the surprise one level up.

Established The deflationary counterweight: tandems gate nothing. If perovskite tandems fail commercially, terawatt-scale solar still arrives on single-junction silicon, which improves on its own learning curve of roughly 20% cost decline per doubling. Tandems are an accelerant on an already-won trajectory, not a hinge of the energy transition — which is exactly why their claims deserve unsentimental audit rather than indulgence.

These horizons track the distance from today’s one delivered shipment to a bankable, warranted, terawatt-relevant product, dated from the mid-2020s record this brief could verify.

• 10 yr: Frontier The decade of decision. Either third-party field data (PACT-class, plus the first customer fleets) supports degradation near 1% per year, a named insurer backs a 25-year tandem warranty, and at least one gigawatt-class line publishes yield — or tandems settle into a premium niche while single-junction silicon keeps the terawatt market. Certified production modules in the 26–28% band are the plausible engineering outcome; both commercial verdicts remain open.
• 25 yr: Speculative If the bankability chain assembles, tandems become the default premium format and module efficiencies near 30% are credible, with all-perovskite or alternative top cells contesting the silicon bottom. If it does not, this period belongs to whatever survived: the record to date licenses no confident forecast either way.
• 50 yr: Speculative Multi-terawatt PV is the safe projection; which absorber pair delivers it is not. Practical module ceilings near 30–35% imply the tandem question — this one or a successor chemistry — gets answered well before this horizon, making 50-year perovskite-specific claims mostly moot.
• 100 / 250+ yr: Handwave Any named-material claim at this range is decoration. The durable facts are thermodynamic: single-junction limits near 29%, tandem limits near 40%, and sunlight’s delivery rate — everything else is institutional history not yet written.

• Depends on Nothing on this map blocks it: its true dependency is the terrestrial crystalline-silicon industry, an incumbent so established it has no brief here, whose wafers, lines and cost curve tandems both ride and must outrun. The storage build-out charted in energy storage revolutions shapes what a marginal efficiency point is worth, but gates nothing.
• Requires (not on this map) an accelerated-test protocol validated against outdoor degradation of the same modules, so chamber hours convert to warranty years; a gigawatt-class tandem line that publishes yield and cost the way silicon’s learning curve always has; a lifetime-rating standard beyond IEC 61215’s qualification screen, since passing an infant-mortality test is not a service life; insurers willing to underwrite 25-year tandem performance warranties, without whom no project finances; and solar-grade lead-halide and organohalide precursor supply at gigawatt volume, which today exists only at speciality-chemical scale.
• Enables Cheaper, denser photovoltaic electricity per unit area — which sharpens the PV-plus-electrolysis benchmark that disciplines artificial photosynthesis, and supplies the high-specific-power blanket arrays that space-based solar power concepts assume.
• Adjacent The firm-power economics of advanced battery technologies and energy storage revolutions decide where an efficiency premium is worth paying; perovskite optoelectronics share materials and tooling with displays and detectors.

Handwave “34% cells mean 34% panels are around the corner.” The certified ladder says otherwise: 34.6% at 1 cm², about 28.6% at commercial cell area, 24.5% for the best certified module, one commercial shipment. Ten points and several unbuilt institutions separate the headline from a product; the record of the field is that each rung has taken years.

Established “Perovskites dissolve in weeks” is out of date. That was true of 2012-era films. Encapsulated modern devices have passed 1,000-hour damp heat and 200 thermal cycles, and certified products exist. The live question is not weeks versus months but whether measured thousands of hours extrapolate to warranted decades — a real doubt, but a different one.

Handwave “It passed IEC 61215, so it lasts 25 years.” The inference is invalid even for silicon, where the standard is a screen whose predictive value comes from four decades of field correlation. For perovskites that correlation does not exist yet, and known degradation modes sit partly outside the suite. Vendors rarely state the fallacy outright; marketing that juxtaposes an IEC pass with silicon-style lifetime imagery invites it.

Frontier “The lead makes them undeployable” — and its mirror, “the lead is nothing.” Both fail on the numbers: less total lead than the solder in the silicon module it would replace, but in a soluble form that measured leaching experiments show escaping damaged laminates. Sequestration layers captured over 96% in lab tests and belong in products; regulation currently permits the lead by silicon-era carve-out rather than fresh assessment. Deployable, with engineering obligations.

Frontier “Announced gigafactories mean supply is imminent.” Announced perovskite and tandem capacity worldwide runs to many gigawatts; audited cumulative shipments were tens of megawatts as of early 2026, mostly single-junction. The gap between an announcement and a shipped, warranted watt is precisely the yield, lifetime and bankability distance this brief maps — treat factory announcements as intentions (vendor) until modules ship.

Established “Tiny-cell records are cherry-picked and meaningless.” Unfair in the other direction. Designated-area certification under stabilised measurement is rigorous physics, and small-area records map the approach to thermodynamic limits — silicon’s records were set the same way. The error is reading them as product specifications; the cure is the ladder, not cynicism about the top rung.

Speculative “Tandems will replace silicon.” Backwards: the leading tandem is silicon, plus a coating. Success entrenches the silicon supply chain rather than displacing it; the technologies that would actually replace silicon — all-perovskite stacks, other thin films — trail on exactly the lifetime and scale metrics this brief tracks.

Frontier “Perovskites are already space-proven.” Early flight exposures and proton tests show encouraging radiation tolerance on small samples — genuinely promising, and part of why space-based solar power studies cite them — but no perovskite array has flown an operational mission life, and space qualification is an unstarted ladder of its own.