Tongwei Visit in Chengdu: A Polysilicon Giant, the Meishan Lighthouse Factory and a Solar Industry at the Bottom of Its Cycle

A Tongwei visit puts a delegation inside the most consequential cost position in the global solar supply chain, at the least flattering moment in its history. Tongwei (600438.SH) sold 384,800 tonnes of high-purity polysilicon in 2025 and 103.03 GW of solar cells, ranking first in the world in both, and still lost 9.55 billion yuan. Revenue for 2025 was 84.13 billion yuan, down 8.55%, and the first half of 2026 came in at 34.36 billion yuan, down 15.19%, with a loss attributable to shareholders of 5.12 billion yuan. The programme itself runs on two sites roughly 60 kilometres apart in Sichuan: the Tongwei Experience Centre at the group's Chengdu headquarters, which tells the story from a channel-cage fish-farming experiment in 1982 to a Fortune Global 500 company, and the Tongwei Solar base in Meishan, which the World Economic Forum named in September 2025 as the first lighthouse factory in photovoltaic cell manufacturing anywhere. That combination is the reason to come. Any Chinese solar maker can describe a down cycle; very few will walk you through a plant that has cut conversion cost by 37% inside it, and explain why that is not enough on its own.

Key takeaways

The standard programme is a full day built around two sites: the Tongwei Experience Centre on the fourth floor of Tongwei International Center in southern Chengdu, and the Tongwei Solar cell base in the Ganmei Industrial Park at Meishan, about an hour south by road. A half-day version drops the headquarters and keeps Meishan, which is where the manufacturing content sits. Both run by appointment, on working days, and are arranged for organised groups rather than individual visitors.

Tongwei is not a pure solar company, and the visit is honest about that. Two businesses sit under one listed entity: agriculture — feed, food and aquaculture — which brought in 29.26 billion yuan in 2025 at a gross margin of 9.47% and sold a record 6.53 million tonnes of feed, and photovoltaics, which brought in 54.14 billion yuan and posted its first negative gross margin, at minus 1.12%. Delegations that arrive expecting a solar pure-play usually leave with a more useful idea: one profitable, slow-growing, cash-generative business sitting underneath one enormous, currently loss-making one.

The manufacturing content concentrates in Meishan. The site runs what the company calls the largest single-site crystalline silicon cell base in the world following its fourth phase, and it is the plant the World Economic Forum recognised in September 2025. Across more than fifty deployed fourth-industrial-revolution use cases, with AI coverage the company puts above 64%, the WEF's own assessment recorded photoelectric conversion efficiency up 12%, defect rates down 41%, conversion cost down 37% and carbon dioxide emissions down 33% against the plant's pre-transformation baseline. Seeing that while the company is losing money is the point.

The session is where a delegation converts the tour into something it can use. Standard themes are the cost curve — what actually determines whether a polysilicon plant survives the next four quarters — the technology choice between TOPCon, HJT, back-contact and perovskite tandems, and exporting into tariffs and origin rules. Tongwei sold 9.52 GW of modules overseas in 2025, up 164%, and has said it is targeting roughly 40% overseas module shipments in 2026, so this is a live operational question rather than a slide.

We arrange the visit end to end: the introduction to the host, the internal scheduling on the Chinese side, the security registration, the session theme, transport between the two sites and English interpretation on the day. Access to production areas is confirmed with the host during planning rather than assumed, and it is the item most likely to change the shape of the day.

Key visit details

Where the visit is held

Two addresses. The Tongwei Experience Centre sits on the fourth floor of Tongwei International Center on Tianfu Avenue in southern Chengdu, the 39-storey tower completed in 2014 that houses the group's management headquarters. The exhibition covers more than 3,000 square metres across ten themed zones, including photovoltaic manufacturing, the fishery-photovoltaic mode, feed industry, green aquaculture and food processing, with a digital sand table, a 3D cinema and a virtual-reality section built around the fishery-PV parks. This is the narrative half of the visit.

The manufacturing half is at Tongwei Solar's base in the Ganmei Industrial Park in Meishan, roughly 60 kilometres south of central Chengdu and about an hour by road on the Chengdu-Leshan corridor. The park is a flywheel development between Meishan city and Garze prefecture, built deliberately on cheap hydropower from the Sichuan mountains — which is the single most important locational fact about it. The drive is arranged as part of the programme rather than left to the day.

Format and duration

A full day, and the day is genuinely full because of the transfer. The morning opens at the Chengdu headquarters with the experience centre, followed by the drive south, security registration at the Meishan gate, a plant exhibition hall, the production route where access is confirmed, lunch, and then a themed session and closing discussion either on site or back at the headquarters. A half-day alternative stays at Meishan for four to five hours and skips the Chengdu exhibition entirely; teams with limited time usually prefer it, because the plant is what they came for.

Who the programme suits

Five groups get proportionally more from it. Solar developers, independent power producers and utility investment teams, because the module and cell economics are inputs to their own procurement decisions. Manufacturers in any process industry, because Meishan is an unusually well-documented example of what AI-driven quality and yield control looks like when it actually runs. Supply-chain and procurement functions, because seeing the concentrated upstream positions changes how a sourcing strategy should be built. Sustainability and ESG teams, because the company can show both the carbon arithmetic of its own plant and the land-use argument behind the fishery-PV model. And investors or analysts, who can test whether a cost advantage survives a cycle that has already lasted two years.

Booking, group size and site rules

Around 30 days is the practical lead time, and the binding constraint is rarely the calendar. It is the internal approval to bring visitors into a working semiconductor-grade plant, plus the security registration. The programme is best suited to groups of 10–20 participants; larger delegations are split into smaller groups for the production route at the host's discretion, both for safety and because a line-side briefing does not survive twenty people talking. Visitor names and passport details are submitted in advance so that passes can be issued.

Why visit a solar manufacturer at the bottom of the cycle

The honest answer to why anybody should visit Tongwei right now begins with a number most corporate websites would rather bury. In 2022 the company earned 25.73 billion yuan on revenue of 142.42 billion. In 2023 it earned 13.57 billion on 139.10 billion. In 2024 it lost 7.04 billion on 91.99 billion, and in 2025 it lost 9.55 billion on 84.13 billion. To get there took also around 5.02 billion yuan of asset impairments in 2025 alone, roughly 2.98 billion of it against inventory and 1.67 billion against fixed assets — money written off plant that will probably never run again at an economic utilisation rate.

That is precisely why the visit is worth a day. A tour of a solar plant taken in 2022 would have shown you the same equipment and told you almost nothing, because everything looked profitable and none of the differences mattered. A tour taken now separates the things that are structural from the things that are cyclical. Cost position, product quality, technology optionality and balance-sheet depth are structural. Price is not. You can only see which is which when the price is below everyone's cost.

From fish feed to polysilicon: the story the exhibitions tell

The Tongwei Experience Centre opens not with a solar panel but with a fish. In 1982 a young man named Liu Hanyuan, now chairman of the Tongwei Group, began experimenting in a Sichuan irrigation channel with metal cages set into fast-flowing water — the channel metal-cage flow-through fish farming technique — and used the proceeds to start a fish feed plant in 1986. That plant grew into Tongwei Feed, listed on the Shanghai Stock Exchange in 2004. The enterprise moved its headquarters from Meishan to Chengdu in 1994 and began looking seriously at polysilicon around 2006.

The thread that ties the two halves together is also the one visitors remember. In 2014 Liu Hanyuan proposed putting photovoltaic arrays above fish ponds — panels generating above, aquaculture continuing below — and the first such project connected to the grid at Rudong in Jiangsu in 2015. The experience centre has a virtual-reality section and a digital sand table built around these parks, and for any delegation working on land-use conflict, rural livelihoods or dual-use infrastructure, this is the segment to slow down in. It is also worth noting what the halls do not dwell on: there is very little here about the current price environment. That conversation happens later, in the session.

Inside the Meishan lighthouse factory

The hall before the line

Meishan has its own exhibition space — the intelligent manufacturing hall that visiting delegations are shown before anything else — and it is worth taking seriously rather than treating as a waiting room. It frames the plant in terms of throughput, automation and how few people are on the floor, and it establishes the vocabulary the rest of the visit uses: conversion efficiency, yield, defect rate, cost per watt. Delegations who skip it asking questions usually find the next hour harder to read.

What the lighthouse designation actually measured

On 16 September 2025 the World Economic Forum published a batch of twelve new lighthouse factories. Tongwei Solar Meishan was the only photovoltaic company in it, the first lighthouse factory in PV cell manufacturing anywhere and the first for a Sichuan-headquartered enterprise. Lighthouses are assessed by the WEF with McKinsey and require the integration of at least five world-class technology applications; the ones cited here include billion-scale big-data analytics driving cell efficiency optimisation, a deep-learning model that self-adjusts screen-printing parameters per cell, CNN and DNN models running closed-loop control of coating, generative AI with retrieval augmentation for equipment maintenance and overall equipment effectiveness, and closed-loop quality management built on ResNet and XGBoost models.

The recorded outcomes — efficiency up 12%, defects down 41%, conversion cost down 37%, carbon emissions down 33% — are the numbers a professional delegation should interrogate rather than accept. Ask what the baseline was and over what period. Ask how much of the improvement survives a change in product mix. Ask what it cost in capital and what the payback was. These are questions the plant managers answer readily, and the answers are more informative than the headline percentages, because they tell you whether this is a replicable method or a very expensive one-off.

Silver, TOPCon and the road not taken

Nothing explains the current state of solar manufacturing better than the paste used to print contacts onto a cell. Silver prices rose sharply through 2025 and into 2026, and the company names this explicitly among the causes of its loss. The response is visible at Meishan in reduced silver consumption: Tongwei says it has brought silver usage below 4 milligrams per watt, and it pioneered a three-cut cell design and a 780-watt heterojunction production route. If your team buys cells, this is the part of the route that affects your own bill of materials.

What is off the route

Two things are worth being clear about before booking. First, this is a cell plant, not a polysilicon plant. The polysilicon bases under Yongxiang are at Leshan and Baotou, with a third at Baoshan in Yunnan, and they are a separate journey into higher-security chemical sites that is rarely available to general delegations; do not build the day around seeing a reduction furnace. Second, even at Meishan, which areas of the production floor are open on the day is decided by the host. Some delegations see the full route; others see designated viewing positions. Plan for the second and treat the first as a bonus.

What the numbers actually say now

Go in with the figures in hand, because they shape every conversation in the room. Revenue for 2025 was 84.13 billion yuan, down 8.55%. Net loss attributable to shareholders was 9.55 billion yuan against 7.04 billion in 2024; excluding non-recurring items the loss was 9.88 billion. Gross margin came in at 2.70%, down 3.69 percentage points, and the group's weighted return on equity was minus 21.95%. Total assets at the end of 2025 stood at 187.78 billion yuan with net assets attributable to shareholders of 38.88 billion, putting the debt-to-asset ratio at 72.63%. Headcount was 44,798, down 19.61% on the year, with revenue per employee of about 1.88 million yuan.

Operating cash flow is the number that surprises people. Despite the loss, it came in positive in 2025 at around 1.38 billion yuan, up 20.59% year on year, and the company reported a small positive figure for the first half of 2026 as well, at about 109 million yuan against an outflow of roughly 1.95 billion in the same period a year earlier. A business can lose money on paper while generating cash if depreciation exceeds capital expenditure and working capital is released, which is largely what is happening here. It does not make the company profitable. It does explain why the conversation is about duration rather than survival.

The segment split is where the story is sharpest. In 2025 agriculture brought 29.26 billion yuan at a gross margin of 9.47%, marginally better than the year before, on record feed volumes of 6.53 million tonnes. Photovoltaics brought 54.14 billion yuan at a gross margin of minus 1.12% — its first negative year — and within that, high-purity polysilicon and chemicals ran at minus 5.38%. Meanwhile the photovoltaic power generation business, small at 2.02 billion yuan, kept a gross margin of 50.46%. The first half of 2026 moved the mix further: photovoltaics contributed 19.65 billion yuan, or 57.2% of revenue, and feed and food 14.35 billion yuan, 41.76%. Almost half of this company's revenue now comes from fish feed.

The cost curve: electricity, scale and vertical integration

Three things determine whether a polysilicon producer gets through this cycle, and all three are visible from a Meishan meeting room even though the furnaces are elsewhere. The first is power. Polysilicon is essentially electricity refined into chemistry, and Sichuan's hydroelectricity together with Inner Mongolia's coal-based grid power are the two poles around which Chinese capacity is built. The second is unit capital cost: Tongwei has said it has compressed capital expenditure to around 500 million yuan per 10,000 tonnes of annual capacity, which changes both the depreciation line and the price at which a new plant can be justified. The third is scale itself — the company's high-purity polysilicon capacity stands at roughly 900,000 tonnes a year, ahead of any competitor, with N-type cell capacity above 150 GW and module capacity around 90 GW.

Teams comparing Chinese solar materials companies will find the comparison useful rather than competitive. A GCL New Energy visit in Suzhou shows the other main approach — fluidised-bed granular silicon rather than the modified Siemens route Tongwei built its position on — alongside a diversified energy group rather than an integrated manufacturer. Nobody at Tongwei will argue theirs is the only way to do this, and seeing both makes either one clearer.

Going abroad while the trade walls go up

The interesting question a visit can address is how a company exports when tariffs, origin rules and local-content requirements are the constraint rather than cost. The answer has three visible parts. The first is market selection away from the most protected destinations towards regions where Chinese modules are still competitive on economics. The second is going down-market in structure: rather than pure component supply, large project agreements such as the 1.86 GW contract with ACWA Power, and dozens of reference projects at the hundred-megawatt scale in Romania, Ireland, Spain and Australia. The third is standardisation — publishing system-level technical documentation that lets overseas engineers design around a product rather than simply buying it.

None of these removes trade exposure, and it would be a mistake for a delegation to leave believing it does. Domestic Asian coverage such as the Korean market, Eastern Europe including Poland and Romania, and the Gulf are all subject to policy change. What is transferable is the method: pick markets by returns rather than volume, build local technical credibility before volume follows, and keep enough product flexibility to redirect when a particular border closes. Teams whose interest runs further downstream into storage rather than generation often pair this with a Sunwoda Battery visit, because the combination is what the company itself points to when it talks about powering data centres.

The anti-involution campaign, and what it means for a visitor

No serious conversation about Chinese solar in 2026 avoids this subject, and it is one a visit is unusually well placed to explain. Through July and August 2026 a sequence of measures landed together: on 2 July three mandatory national standards covering energy-consumption limits for polysilicon, wafer and module manufacturing were issued jointly, setting a threshold below which existing capacity must either upgrade or close; on 27 July the China Photovoltaic Industry Association published a group standard establishing a cost-accounting model; and on 6 August eight leading polysilicon producers, Tongwei among them, signed a pledge not to sell below their own calculated full cost. A joint notice on cost accounting for low-price competition followed in September.

What this means on the ground is easier to describe than to evaluate. Chinese industry reporting put polysilicon inventories at around 522,000 tonnes at the end of July 2026, roughly five months of global demand held in warehouses, with prices having fallen to about 31,750 yuan a tonne at the mid-year low before recovering to the low-to-mid 40,000s by late September. Reported industry average full cost sits near 48,000 yuan a tonne, with the leaders able to compress towards 38,000 to 43,000. Operating loads were reported falling towards roughly 35% in October. In other words, prices recovered to roughly the leader's cash cost and not much further.

Sample visit programme

This is the shape of a standard full day with both sites, the version most first-time delegations take. Pick-up is normally around 08:15 from a Chengdu hotel, and the day ends in the mid-afternoon with the option of a second meeting afterwards.

Every Chinese solar manufacturer is running the same cost curve. The visit is about seeing how far down it one company has pushed itself, and what that costs while the industry waits for capacity to leave.

Practical notes for the day

Book the interpretation early and give the interpreter the glossary. Solar vocabulary does not translate cleanly — back-contact, heterojunction, tandem, conversion efficiency — and above all the distinction between full cost and cash cost, which is doing real work in this industry right now. Professional English interpretation is part of how we run the programme, and a glossary goes out with the pre-visit material.

Build the itinerary around the transfer, not in spite of it. Chengdu to Meishan is comfortable but it is real time, and the site is a chemical-adjacent plant with its own gate procedures. Delegations that try to compress this into a morning alongside a second company in the same city consistently report that both halves suffer. For context on how Chinese clean-energy manufacturers at this scale run their own visitor programmes, a CATL visit in Ningde is a useful comparison. Teams building a two-company Chengdu day usually pair this with a Gree washing machine visit in Chengdu, which keeps the second programme in the city and sets an energy manufacturer against a consumer-goods one.

Questions before a Tongwei visit

Where is the visit held? At two sites: the Tongwei Experience Centre at Tongwei International Center in southern Chengdu, and the Tongwei Solar base at Ganmei Industrial Park in Meishan, about 60 kilometres or one hour south. Transport between them is included in the programme.

Is this really a factory visit? Depends on what you mean, and it is better to be precise in advance. Meishan is a working cell manufacturing plant and delegations have historically been taken onto the production floor, but which areas are open is confirmed with the host during planning and decided on the day. The Chengdu headquarters is an exhibition and meeting facility, not a factory.

Can we see a polysilicon plant instead? Generally no. The reduction furnaces are at Leshan, Baotou and Baoshan, all remote chemical sites with far tighter access, and they are not part of this programme. The polysilicon business is discussed in the session rather than visited.

How much of the day is travelling? Roughly two hours in total, split either side of the Meishan visit, plus site registration at both gates. That is why the standard programme is a full day, and why the half-day version drops the Chengdu exhibition rather than trying to fit both.

Can our own engineers bring specific process questions? Yes, and they should. Submit them with the visitor list so the right specialists are pulled into the session. Questions about yield management, silver reduction, closed-loop inspection and maintenance scheduling are answered in concrete terms; anything touching a supplier's proprietary position will be declined, politely and quickly.

Who should be in the room? Alongside commercial decision-makers, someone who reads a plant. Procurement teams, process engineers, quality leads and anyone who owns a manufacturing footprint get far more from the production route than a room full of general managers does.

Can the session be tailored to our team? Yes. The standard themes are the cost curve and capacity discipline, the cell technology route map, and exporting into trade barriers; a fourth — running a dual agriculture-and-energy business through a commodity down cycle — is available by arrangement. It must be confirmed before booking because it determines who attends.

What are the photography and security rules? Allowed in the exhibition areas as a rule, restricted on the production floor, where equipment specifications and process parameters generally may not be photographed and phones or recording equipment may have to be left outside. A confidentiality acknowledgement is sometimes required before entry, and visitor passports are registered in advance.

How far in advance should we book? Around 30 days. The binding constraint is the internal approval to bring visitors into an operating plant and the security clearance attached to it rather than the size of the group.

What group size works best? the programme is best suited to groups of 10–20 participants. Larger delegations are workable and are split into smaller groups for the production route at the host's discretion.

Arranging your Tongwei visit

We arrange the visit end to end: the introduction to the host, the internal approvals and scheduling, the visitor registration and site security, the session theme, transport between Chengdu and Meishan, and English interpretation on the day. Tell us your dates, your delegation size and the one question your team needs answered, and we will confirm what can be arranged. Start at chinatechvisits.com/company-visits.