1. Why this race is hard to read now

On July 14, 2026, LONGi announced a new world record: 35.5% efficiency for a crystalline silicon-perovskite tandem cell, certified by the European Solar Test Installation.

The number is important. It is also easy to overread.

One month earlier, Trinasolar announced a 907W, 29.2%-efficient tandem module on a standard industrial area, verified by TÜV SÜD. Oxford PV, meanwhile, had already shipped the first commercial perovskite-silicon tandem modules and converted part of its patent position into licensing agreements.

These announcements appear in the same race, but they sit at different stages between laboratory performance and a repeatable, financeable product.

That is the reason to compare them. Not to crown a winner, but to ask what each milestone actually proves.

This is not a complete market ranking. It is an evidence comparison built around three visible milestones and the engineering questions underneath them.

2. Evidence

2.1 Three milestones, three questions

LONGi answers the cell-performance question. The 35.5% result shows how far the tandem architecture can push certified efficiency. LONGi also reported 34.3% on a 261 cm² device, 32.2% on a 274 cm² device, and tandem-module efficiencies of 31.4% and 29.4%.

Those larger-area results matter because area, texture, interconnection and process variability become harder to control as devices scale. They move the evidence beyond a single small champion cell.

They do not disclose production yield, sustained throughput, manufacturing cost or shipment volume.

The defensible conclusion is therefore narrow: LONGi leads the published cell-performance frontier. The public evidence does not yet establish repeatable high-volume output.

(Source: Patsnap Eureka and LONGi public disclosure.)

Trinasolar answers the industrial-module question. The company says its 907W module uses a standard industrial area and complies with mainstream module specifications. TÜV SÜD verified the reported 907W output and 29.2% full-area efficiency.

That is a different scale-up signal from a champion cell. Film uniformity, passivation, spectral matching, interconnection and module architecture are all part of the result.

A record module is still a selected result rather than a production distribution. It does not reveal the median efficiency, line yield, cycle time or cost across sustained manufacturing.

(Source: Patsnap Eureka and Trinasolar public disclosure.)

Oxford PV answers the early-commercialization question. In September 2024, Oxford PV shipped the first commercial perovskite-silicon tandem modules to a US customer. The delivered 72-cell modules had 24.5% efficiency. Its 26.9% figure was a separate record module, not the shipped product.

Oxford PV later signed an exclusive China licence with sublicensing rights for Trinasolar and a non-exclusive US licence for First Solar. The US agreement explicitly excludes crystalline-silicon semiconductors.

These transactions show that Oxford PV’s portfolio has become a commercial asset. They do not establish that every tandem route requires an Oxford PV licence, or that the company has already proven manufacturing scale and bankability.

(Source: Patsnap Eureka and Oxford PV public disclosures.)

The comparison is not one number against another. It is three milestones answering three different questions.

2.2 The engineering routes behind the headlines

Patent intelligence is most useful here when it explains technical direction, not when it counts documents and turns the total into a leaderboard.

LONGi’s representative families point toward interface and deposition control. EP4376569A1 addresses electrical-function matching at the tandem interconnection. US12133398B2 uses an ordered induction layer to improve perovskite growth on textured surfaces. US20250008830A1 combines solution and vacuum deposition to form perovskite layers on textured silicon.

Trinasolar’s representative filings point toward silicon-platform integration. CN113206123A describes a perovskite-TOPCon tandem structure. CN113013277A also addresses construction of a tandem architecture on a TOPCon silicon bottom cell. Together with the 907W module disclosure, the route is oriented toward integrating tandem technology with an industrial silicon platform and module format.

Oxford PV’s visible signature combines portfolio protection and commercialization. Its company-reported portfolio of more than 400 granted patents, the 2024 shipment and the subsequent licences show a strategy that connects device and manufacturing IP with market access.

The evidence types must remain separate. Patent families indicate technical and strategic direction. Shipments and licences indicate commercial action. Neither category independently proves yield, freedom to operate, essentiality or market leadership.

(Source: Patsnap Eureka.)

3. Bottlenecks: Public evidence gets thinner closer to bankability

The US Department of Energy groups the commercialization problem into four connected challenges: stability and durability, efficiency at scale, manufacturability, and technology validation and bankability.

1. Repeatability, not another decimal place. Champion results prove that an architecture can work. Manufacturing requires a stable distribution of results across runs, tools and material lots. Public announcements rarely include median efficiency, yield, downtime or process capability.
2. Module integration at industrial area. Larger devices add inactive area, interconnection loss, encapsulation effects and uniformity challenges. Trinasolar’s module is therefore meaningful, but the missing evidence is sustained production rather than a higher selected output.
3. Reliability must connect accelerated testing to field life. A laboratory test lasting 1,000 or 2,000 hours cannot be divided directly into a 25-year warranty. Heat, humidity, UV exposure, electrical bias and thermal cycling require validated degradation models across climates and module designs.
4. Bankability is a system property. A shipped product is not automatically financeable at scale. Warranty backing, supplier strength, field history, production statistics and replacement risk all affect whether capital providers can underwrite deployment.

A 2024 reliability review involving NREL, industry and multiple research institutions concluded that perovskite-silicon tandem devices were not yet sufficiently reliable to establish confident commercial viability.

The efficiency frontier has moved since then. Public evidence on repeatable manufacturing and field reliability has not moved at the same speed.

4. Solutions: Match the evidence to the engineering constraint

The table is not a recommendation. It is a way to avoid asking one metric to answer a question it was never designed to answer.

Engineering teams can build their own comparison scope in Patsnap Eureka, filtering by assignee, technology classification, jurisdiction and filing date, then separating patent evidence from performance and commercial disclosures.

5. Trends

Signal 1: LONGi manufacturing commitment (MONITORING, watch: H2 2026).

LONGi’s module results show that the work has moved beyond a single champion cell. The next stronger signal would be a named pilot or production line with capacity, process configuration and start date. A capacity announcement without yield or throughput evidence would still leave the repeatability question open.

Signal 2: Trinasolar production evidence (MONITORING, watch: 2026-2027).

The 907W result establishes an industrial-format milestone. The next signal is sustained production evidence, including capacity allocation, median output, yield and cost relative to conventional silicon modules.

Signal 3: Oxford PV scale and licence boundaries (MONITORING, watch: 2026-2027).

Oxford PV has already shown shipment and licensing activity. The next questions are manufacturing expansion, licence implementation and the boundary between crystalline-silicon tandem routes and other perovskite applications.

Signal 4: Independent reliability evidence (OPEN, watch: 2026-2028).

The most important reliability announcement will not be a longer isolated test alone. It will connect accelerated testing, outdoor exposure and a credible lifetime model to a commercially meaningful warranty.

Signal 5: Repeatable full-size output (OPEN, watch: 2027).

A disclosed distribution of production results would change the evidence hierarchy more than another selected record. Watch for median module efficiency, yield, throughput and failure-mode data from a named line.

Conclusion

The tandem race does not lack breakthroughs.

It lacks equivalent evidence.

LONGi’s cell record, Trinasolar’s industrial module and Oxford PV’s shipment and licences all matter. They matter for different reasons.

The sharper question is not who owns the highest number.

It is what the milestone proves, which engineering decision it supports, and what evidence is still missing before the technology becomes a repeatable, bankable product.