Chapter 1　Industry Overview: 2026 Global and Chinese Polysilicon Supply-Demand Map

Polysilicon sits at the very top of the photovoltaic value chain, and it is also where the most violent price collapse and capacity reshuffle of 2024–2026 has played out. Pulling the lens back, global polysilicon production in 2025 reached approximately 1.9 million tonnes, while nameplate capacity has now broken past 3.2 million tonnes — an overall utilization rate below 60 percent. China accounts for nearly 90 percent of output, with the remaining slice split among Germany's Wacker, the United States' Hemlock, South Korea's OCI, Malaysia's OCIM, and the newly commissioned Oman polysilicon facility. According to Bernreuter Research's November 2025 capacity audit, nine of the world's top ten producers are now located in China, with Germany's Wacker holding the sole overseas seat. This level of concentration is rare even within heavy chemicals, more closely resembling potash or indium-tin oxide.

On the demand side, BloombergNEF in December 2025 cut its forecast for 2026 global solar PV additions to 649 GW, slightly below the 690 GW seen in 2025. This represents the first annual contraction in continuous records dating back to 2000. BNEF cited three drivers: China's 15th Five-Year Plan signals a slowdown, European and US subsidies have begun rolling off, and grid integration constraints have tightened. Converting GW to polysilicon demand, 2026 global silicon consumption falls around 1.1 million tonnes, aligned with SunSirs' year-end 2025 forecast range of 1.02–1.18 million tonnes.

With 1.1 million tonnes of demand against 3.2 million tonnes of nameplate capacity, the polysilicon industry on paper is in 2.9× oversupply. Even after stripping out idled legacy capacity and overseas capacity not engaged in China's spot market, domestic effective capacity to domestic demand still runs at more than 1.8× oversupply. This is the fundamental reason that Chinese spot prices have stayed pinned near RMB 40,000 per tonne through 2025–2026, and the direct backdrop for the December 2025 announcement that six leading Chinese polysilicon makers will jointly raise on the order of USD 5 billion to absorb idle capacity.

Charting the price curve from early 2022 through mid-2026 is striking. In April 2022, N-type dense polysilicon spot prices briefly touched RMB 300,000 per tonne; that third quarter, the Xinjiang incident layered with silane tightness pushed P-type dense to a historical ceiling of RMB 330,000. Entering 2023, a flood of new capacity slammed prices down to RMB 60,000; an early-2024 rebound to RMB 72,000 proved short-lived, and by late 2024 N-type spot had settled at RMB 40,000. Through 2025 the price traded sideways in a RMB 38,000–45,000 band, corresponding to roughly USD 5–6 per kilogram.

Across the four-year span the cumulative decline exceeds 85 percent — the steepest move in any major chemical or new material over the last decade, outpacing the 2011–2013 polysilicon cycle. That earlier cycle dragged Wuxi Suntech and Jiangxi LDK into bankruptcy and toppled Poly GCL from its earlier global throne. This cycle is rewriting the map differently: Tongwei has expanded polysilicon capacity roughly tenfold from 2020 to 2025 to about 910,000 tonnes; GCL Technology has completed a full pivot from rod-shaped silicon to granular silicon; while Daqo New Energy, Xinte Energy and Asia Silicon each hold roughly 300,000 tonnes.

The global supply-demand structure is being reshaped by the N-type era. Before 2022, P-type PERC cells dominated and the industry graded materials into "dense," "cauliflower," and "coral" cuts. From 2024, N-type TOPCon captured over 60 percent of module shipments; in 2025 the share climbed to 75 percent; and 2026 is expected to approach 90 percent. N-type cell requirements for metallic impurities in polysilicon plunged from the P-type-era benchmark of 5–10 ppbw to below 1 ppbw, with leading module makers tightening internal control to 0.5 ppbw. The N-type-to-P-type premium widened from about RMB 1,000 per tonne in early 2024 to RMB 3,000–5,000 in 2025, and at certain moments granular N-type silicon even traded above rod-shaped N-type dense.

Polysilicon's downstream is not only solar. The 2025 global semiconductor-grade polysilicon market sat at about 50,000 tonnes, dominated by an oligopoly of Wacker, Hemlock, OCI, Tokuyama and Mitsubishi Materials. Semi-grade polysilicon requires 11N purity or higher, prices range from USD 70–100 per kilogram (more than 10× solar-grade), and gross margins consistently hover around 30 percent. This is the foothold that allows Wacker and Hemlock to maintain healthy books even amid the solar-grade collapse.

Major Chinese polysilicon producers also visibly stepped up semi-grade and electronic-grade R&D and expansion across 2024–2025. Xinte Energy in Inner Mongolia has planned kilotonne-scale electronic-grade trichlorosilane and electronic-grade polysilicon projects; Asia Silicon in Xining is pushing float-zone-grade trial production; and producers in Ordos, Leshan in Sichuan, and Zhongwei in Ningxia have begun writing electronic-grade silicon into their strategic plans. This is not mere "diversification" but a deliberate upward shift of the capacity matrix into higher-margin tiers under expectations that solar-grade margins will stay thin.

On the national policy front, China issued a polysilicon carbon footprint label and a revised unit energy consumption ceiling in 2025. The EU's CBAM enters its final transition phase to full collection in 2026, shifting Europe's carbon footprint requirement on polysilicon from "report-only" to "priced." The US in 2024–2026 has rolled out a 232 investigation on Chinese modules and anti-circumvention probes on Indonesia and Malaysia, with polysilicon implicitly swept into scope as the chain's origin point. Polysilicon is no longer just a chemical feedstock — it has become the carrier of US-China-EU energy security and carbon trade contests.

Stitched together, the 2026 polysilicon landscape is clear but harsh: 3.2 million tonnes of nameplate capacity built, only 1.1 million tonnes of global demand, and a downstream cell side that has already entered the N-type plus perovskite-tandem era. Every tonne of polysilicon must find its way through a four-jaw vise of oversupply, low price, low carbon, and N-type. The remaining thirteen chapters unpack that path.

Chapter 2　Value Chain: Material and Energy Flow from Industrial Silicon to Polysilicon

Polysilicon is not refined directly from sand. It goes through two stages of purification, two phase changes, and two clusters of electricity investment. Straightened out, the chain starts with silica (silicon dioxide), is smelted in an electric arc furnace into industrial silicon (metallurgical-grade silicon), reacts with hydrogen chloride to form trichlorosilane (TCS), and is reduced in a Siemens reactor under hydrogen to yield polysilicon. This route is the Modified Siemens Process, used by more than 85 percent of global polysilicon capacity.

The first gate is industrial silicon. China's industrial silicon capacity is concentrated in Yunnan, Sichuan, Xinjiang, and Inner Mongolia, with 2025 national output around 4.1 million tonnes against overseas capacity under 1 million tonnes. Industrial silicon is the bridge between "black metallurgy" and "white chemistry": it feeds aluminum alloys, silicones, polysilicon, solar-grade silicon, and semiconductor-grade silicon. Polysilicon takes about 30 percent of industrial-silicon downstream — the largest single category aside from silicones. Industrial silicon prices fell from RMB 15,000 to RMB 8,500 per tonne across 2024–2025, a 40 percent drop, trimming the industrial-silicon share of polysilicon total cost from 12 percent to 8 percent.

The second gate is trichlorosilane. Industrial silicon is milled to dozens of microns and reacted with HCl in a fluidized chlorination reactor, producing primarily TCS with by-products including silicon tetrachloride, dichlorosilane, and hydrogen. TCS is the core intermediate of the Modified Siemens Process, and its purity directly determines downstream metallic impurity levels in polysilicon. Chinese TCS capacity is concentrated at Xinjiang Daqo, Xinjiang GCL, Xinte Energy, Sichuan Silicon Treasure, and Hubei Yihua. Overseas capacity sits inside closed-loop captive systems at Wacker, Hemlock, and Tokuyama, with thin merchant markets.

The third gate is reduction deposition. High-purity TCS is fed with hydrogen into the Siemens CVD reactor, where at temperatures above roughly 1,100 °C silicon is deposited layer by layer onto heated silicon core rods, forming rod-shaped polysilicon. This is the heaviest power-consumption step, accounting for more than 50 percent of total electricity use. Modified Siemens line power consumption fell from an industry average of 58 kWh/kg in 2022–2025 to around 48 kWh/kg, with leading producers like Xinte and GCL's Xuzhou base reporting reduction-furnace consumption below 40 kWh/kg. This has been the most visible operational gain of the past three years.

By-product recycling is the hidden difficulty of the process. One tonne of polysilicon produces three to four tonnes of silicon tetrachloride, which must be converted back to TCS through cold hydrogenation. Cold hydrogenation requires high pressure, high temperature, and noble-metal catalysts, with unit investment approaching the reduction reactor itself. A producer without proprietary cold hydrogenation must either buy TCS on the open market or sell silicon tetrachloride downstream to silicone makers at a loss. Cold hydrogenation has therefore become the implicit process moat — Xinte, Daqo, GCL, Asia Silicon, and East Hope all run captive cold-hydrogenation lines.

Fluidized bed reactor (FBR) technology is the parallel route, starting not from TCS but from silane (SiH4). Silane is heated in a fluidized bed reactor to 600–700 °C and decomposes onto continuously tumbling seed particles, ultimately forming millimeter-scale granular polysilicon. The FBR route uses about 60 percent of the Modified Siemens Process's power and roughly 60 percent of the capital intensity; the trade-offs are granular product form, historically weaker minority carrier lifetime and metal impurity control versus rod silicon, and high difficulty in scaling. In China, only GCL Technology has achieved million-tonne-scale FBR commercial production. Overseas, REC Silicon and SunPower attempted it in the early stages, but REC Silicon shuttered its Ohio silane and FBR plant in 2024.

From an energy-flow view, polysilicon is intensely electricity-bound. Producing one tonne of rod-shape polysilicon takes 48,000–55,000 kWh of total electricity. At China's western industrial rates of RMB 0.33–0.38 per kWh, power alone is 35–40 percent of total cost. This is why every new Chinese polysilicon project clusters in Xinjiang, Inner Mongolia, Qinghai, Ningxia, and Leshan in Sichuan, where hydro, wind, and coal power are abundant.

Chapter 3　Process Routes: Modified Siemens vs. Fluidized Bed

The contest between Modified Siemens and FBR has been the most enduring debate in China's polysilicon industry for the past fifteen years. From 2008, when Poly GCL licensed MEMC's FBR patents, granular silicon has carried the labels of "low energy, low capex, future winner." But rod silicon, leveraging N-type-era metal impurity control, mono-pull efficiency, and market inertia, has repeatedly pushed back. Only in 2025 has granular silicon truly turned the tide on N-type quality.

Lining up the key metrics side by side reveals each route's edges and weak spots. Rod silicon: 48 kWh/kg of total power, 8 tonnes of water per tonne, 2.3 tonnes of steam per tonne, RMB 4 billion per 10,000-tonne nameplate capacity. N-type dense rod silicon comes out as 120 mm diameter rods, then must be crushed to 30–100 mm chunks before entering downstream mono pullers — crushing introduces re-contamination, metal impurities, and fill-density variability.

Granular silicon: 20 kWh/kg of total power, 5 tonnes of water per tonne, 1.8 tonnes of steam per tonne, RMB 2.5 billion per 10,000-tonne nameplate capacity. Product form is spherical particles 1–3 mm in diameter, recharge-friendly directly into the mono puller, no crushing, high bulk density. But granular silicon has carried three historical burdens: high hydrogen content (causing bubbles during pulling), historically elevated metal impurities, and lower density than rod-shape dense.

The 2025 inflection came from GCL Technology. According to its 2025 interim report, the share of FBR granular product meeting N-type quality (901A and above) exceeded 96 percent, and the share with 5-element total metallic impurities below 1 ppbw reached 95 percent, with select batches even falling below 0.5 ppbw. "In continuous N-type rod pulling, the drop in minority carrier lifetime at the rod head was notably better than contemporaneous rod-shape N-type dense" — this was the first time in over a decade that granular silicon received customer validation as quality-superior to rod silicon in N-type scenarios.

Infolink's July 2025 weekly index showed for the first time the price of granular silicon trading above N-type dense chunks. This local inversion appeared repeatedly through the second half of 2025; in the October–December 2025 average, granular N-type matched rod-shape N-type dense at the index level. This is a watershed for granular silicon's market positioning.

By Q2 2025, GCL's granular silicon average cash cost (excluding tax) had fallen to RMB 25.31 per kilogram, down another 6.5 percent from Q1 2025. After the 60,000-tonne Xuzhou project ramped in the second half of 2025, unit power consumption was projected to fall to 10–12 kWh/kg, less than half the Modified Siemens process worldwide. Over the same period, leading rod-silicon makers Tongwei, Xinte, and Daqo held cash costs in the RMB 30–35 per kilogram range. Granular silicon now leads the rod-silicon top tier by about RMB 5 per kilogram on cash cost and tier-two rod producers by over RMB 10 per kilogram.

But cost is not everything. The rod-silicon moat is built from process maturity, customer qualification inertia, mono puller compatibility, and industry voice. Before 2024, over 80 percent of N-type TOPCon cell makers used a "dense plus recharge" rod-silicon structure; switching to granular silicon means retuning pulling parameters, recertifying lines, and re-evaluating yield. This switching friction is why granular silicon, despite its cost advantage, cannot displace rod silicon overnight.

Chinese market share confirms this. As of the first half of 2025, granular silicon held about 24 percent of national polysilicon output, versus less than 10 percent five years earlier. Granular silicon's share expansion has been a smooth rising line, without sudden surges or cliffs. The neutral 2025–2030 forecast is that granular silicon's share will steadily expand to 35 percent — not "wholesale replacement."

Chapter 4　Major Producers: Tongwei, GCL, Daqo, Xinte, Asia Silicon, and Overseas Giants

China's polysilicon narrative always centers on the five-house leadership of Tongwei, GCL, Daqo, Xinte, and Asia Silicon, plus four overseas houses — Wacker, Hemlock, OCI, and Tokuyama. These nine companies together hold over 90 percent of global capacity, the structural core of the industry.

Tongwei (Yongxiang Polysilicon). Total 2025 capacity reached 910,000 tonnes across three bases — Leshan in Sichuan, Baotou in Inner Mongolia, and Baoshan in Yunnan. Leshan is the oldest base. Baotou anchors the western-electricity dividend, with unit power consumption below 42 kWh/kg. Baoshan is the hydro-rich "green silicon" base, eligible for a "full-hydro silicon" carbon label. Tongwei's 2025 polysilicon sales reached about 800,000 tonnes, with the 862,400-tonne multi-year supply contract signed with Longi covering 2024–2026. But Tongwei expects full-year 2025 net loss of RMB 9 billion to RMB 10 billion — its largest single-year loss since IPO.

GCL Technology. In 2025 the company fully completed its rod-to-granular pivot, with 480,000 tonnes of nameplate capacity, all FBR granular silicon, spread across Xuzhou, Leshan, Hohhot, and Baotou. Xuzhou is the granular-silicon birthplace, with 210,000 tonnes of 2025 capacity and 10–12 kWh/kg power consumption; Leshan is tied to hydropower with 120,000 tonnes; Baotou takes Inner Mongolia's green power with 100,000 tonnes; Hohhot was commissioned in the second half of 2025 with green-power allocation. GCL's 2024 FBR granular output was 269,200 tonnes, 17 percent market share; in H1 2025 share rose to 24 percent.

Daqo New Energy. Two bases at Shihezi in Xinjiang and Baotou in Inner Mongolia, 2025 total capacity 350,000 tonnes, all rod silicon. Daqo's actual 2025 production was 123,700 tonnes, down 40 percent from the 205,100 tonnes of 2024, as the company voluntarily throttled production and inventory. Q4 2025 production cost fell to USD 5.83 per kilogram (about RMB 42 per kg); cash cost fell to USD 4.46 per kilogram (about RMB 32). 2025 full-year revenue was USD 665.4 million, down 35 percent from 2024's USD 1,029.1 million. 2026 production guidance is 140,000–170,000 tonnes.

Xinte Energy. Three bases in Changji (Xinjiang), Junggar (Inner Mongolia), and Ordos (Inner Mongolia), 2025 capacity 300,000 tonnes, all rod silicon. The differentiator is captive power — Changji is paired with a 700 MW captive coal plant, Junggar relies on Inner Mongolia green power direct supply, and Ordos is a 200,000-tonne base commissioned in 2024. Xinte's cash cost has consistently ranked top-three, reaching RMB 26 per kg in 2025. Public announcements in 2025 signal green power allocations and electronic-grade silicon as the two strategic priorities for 2026–2028.

Asia Silicon. Xining in Qinghai, 2025 capacity 100,000 tonnes, rod silicon. Asia Silicon is the smallest in scale but most stable in process, with N-type share above 95 percent and select batches feeding semi-grade customers directly. The company is one of the few never publicly tied to a polysilicon incident, owing to Qinghai's high altitude, clean industrial water, and the salt-lake chemistry feedstock chain. Asia Silicon disclosed no expansion plans in 2024–2025, focusing instead on float-zone-grade silicon and electronic-grade trichlorosilane.

East Hope. Two bases in Junggar East (Xinjiang) and Zhongwei (Ningxia), 2025 capacity 350,000 tonnes, rod silicon. East Hope is unusual as a non-traditional chemicals player originating from feed and aluminum, with cash costs in line with Daqo but weaker downstream brand. Bernreuter Research groups East Hope alongside Tongwei, GCL, Daqo, Xinte, and TBEA in the "China five-house" standard.

Overseas, Germany's Wacker is the world's earliest industrial polysilicon producer, with 2025 capacity at 80,000 tonnes (split between solar and semi-grade), of which 20,000 tonnes of semi-grade hold 35 percent of the global semi-grade share. The new semi-grade line at Burghausen was completed in 2025, expanding capacity by more than 50 percent. Wacker reports semi-grade margins above 30 percent in 2025, nearly fully offsetting the thin margins of solar-grade.

US Hemlock — owned jointly by Corning and Japan's Shin-Etsu — holds 30,000 tonnes of 2025 capacity, focused on semi-grade. Hemlock joined Wacker in 2024 lobbying for the US 232 investigation that brought polysilicon under national security review; in 2025 Hemlock secured Inflation Reduction Act advanced-manufacturing tax credits that left its unit cost roughly USD 3 per kilogram below OCI's.

Korea's OCI has shifted weight to its Bintulu base in Malaysia after closing its Gunsan plant in 2020. OCIM's 2025 capacity is 35,000 tonnes, expanding to 56,600 tonnes in 2026–2027. OCI is one of the few producers with UFLPA exemption status, giving its Malaysian product a policy moat in the US market.

Japan's Tokuyama and Mitsubishi Materials have consolidated their silicon business into semi- and electronic-grade. In 2025 Tokuyama and OCI launched a next-generation electronic-grade silicon joint venture in Sarawak, Malaysia.

Chapter 5　The N-Type Era: Dense, Recharge, Electronic-Grade Metal Silicon, and Mono Premiums

The rise of N-type cells has rewritten the structure of polysilicon downstream demand. From the three-tier "dense / cauliflower / coral" of the P-type PERC era, the N-type TOPCon era converged onto a two-tier "dense / recharge" structure; the N-type HJT era is converging onto a two-tier "dense / granular" structure, with metallic impurity, hydrogen content, density, and bulk density specifications tightening across the board.

N-type dense is the current benchmark. Physical form is 30–100 mm chunks from crushed rod silicon, density above 1.6 g/cm³, 5-element total metallic impurities under 1 ppbw, carbon under 0.2 ppmw, hydrogen under 1 ppmw. N-type dense is the main charge for first-fill in mono pullers, and is the reference price for N-type silicon in general.

N-type recharge takes physical form of 3–30 mm small chunks or particles. Density is less critical for first-fill, but bulk density and metallic impurity requirements match dense. Recharge is the consumable for continuous feeding during mono pulling and indispensable for large-format rods (G12, M10, N10).

Granular silicon is naturally a recharge product, but GCL's granular silicon's N-type quality since 2025 gives it potential to serve directly as dense. On N-type HJT lines, a 50/50 mix of granular and rod-shape dense has become standard operating practice. HJT cell makers including Anhui Huasun, Jinneng Technology, and Akcome publicly stated in 2025 that granular silicon's metallic impurity control under HJT actually exceeds rod-shape dense.

The N-type premium has widened from about RMB 1,000 per tonne in early 2024 to RMB 3,000–5,000 in 2025, with the absolute premium running 8–10 percent of total silicon price. But the relative growth of the premium matters more than the absolute level: when overall silicon prices fell 80 percent, the N-type premium kept widening, meaning N-type pricing power has decoupled from module-cycle dynamics — a rare growing niche even in a downturn.

Electronic-grade metallic silicon is the upstream extension of the N-type era. Semiconductor wafer manufacturing requires float-zone-grade silicon (FZ-grade) of 11N or higher purity. The feedstock for FZ-grade is highly purified electronic-grade metallic silicon. Domestically only Xinjiang Daqo, Xinte, and Asia Silicon are running hundred-tonne-scale pilots, while globally Tokuyama, Mitsubishi Materials, and JFE remain dominant. China's next ambition in N-type cell silicon is to capture this upstream link.

Mono pull premium is, in essence, a pull yield premium. Using Tongwei's N-type dense at a Longi or Jinko line, mono head minority carrier lifetime can sit stably above 2,000 µs; using second-tier silicon under the same conditions, the metric may be only 1,500 µs, corresponding to 5 percentage points of difference at the mono head yield. Each percentage point of pull yield translates to roughly RMB 1.2 per kW at the module end. This is why N-type-era silicon procurement negotiations are starting to resemble semi-grade negotiations — looking less at the spot index, more at process-fit metrics.

Chapter 6　Granular Silicon Breakthrough: GCL's Yield, Energy, and N-Type Compatibility

Granular silicon's 2025 breakthrough was not sudden — it was the inevitable release of fifteen years of technology accumulation in the N-type era. Tracing the path from Poly GCL's 2008 MEMC FBR licensing, through the 2017 Xuzhou pilot, the 2020 Jiangsu Xuzhou 60,000-tonne scale-up, and the 2023 Hohhot and Baotou bases, granular silicon's key technical metrics — hydrogen content, metallic impurities, density, bulk density — have iterated every three to five years. 2025 marked the inflection on that accumulation curve.

The yield inflection arrived in the second half of 2024. GCL's announcement: from Q3 2024, FBR granular silicon total yield (qualified output to charged input) for the first time stably broke 95 percent, and by H1 2025 it sat steadily above 97 percent. This let granular silicon's per-tonne investment ROI match the rod-silicon top tier for the first time.

The energy-consumption inflection arrived in the first half of 2025. The 60,000-tonne Xuzhou granular project commissioned in May 2025 measured power consumption at 11 kWh/kg — 1 kWh below design. Granular silicon's specific power consumption is now only 23 percent of the Modified Siemens average, the world's lowest polysilicon route in production. Xuzhou also runs the lowest water consumption (3 tonnes per tonne of silicon) and steam consumption (1.2 tonnes per tonne of silicon) in the industry.

The N-type fit inflection arrived in Q2 2025. Infolink's July report first showed granular silicon trading above N-type dense chunks. The inversion reappeared repeatedly through the second half and continued into 2026. HJT cell makers — Huasun, Jinneng, Akcome — publicly stated at technical forums that granular silicon's metallic impurity performance under HJT is "substantively superior" to rod-shape N-type dense. This was the decisive evidence that granular silicon had earned its quality vote in the N-type market.

Layered together, granular silicon's cost curve, quality curve, and scale curve all moved up simultaneously in 2025, producing a market-share jump from 17 percent in 2024 to 24 percent in H1 2025. The share data may sound incremental, but layered against shrinking industry output, the absolute capacity increment for granular silicon in 2025 expanded 30 percent over 2024.

But granular silicon is not winning everywhere. The N-type market still favors rod-shape dense for density and first-pull yield. In leading domestic module-maker process data, when first-fill granular share exceeds 50 percent, minority carrier lifetime drops slightly. This is the last mile granular silicon has yet to fully conquer, and the holdout where leading rod producers Tongwei, Xinte, and Daqo defend their N-type position.

Chapter 7　Downstream Through the Platform Lens: Capability Atlas of the Polysilicon Cut

Shifting the lens from polysilicon itself to downstream, you find a fact often overlooked — the solar polysilicon chain is not just "silicon to wafer to cell to module." It rings around a circle of supporting factories: quartz crucible plants, quartz sand plants, hot-zone graphite component plants, silicon core rod plants, reduction-furnace electrode plants, high-purity gas suppliers, silicon crusher plants, silver paste makers, EVA and POE encapsulant plants, aluminum frame plants, and glass plants. This ring of supporting factories underpins the entire competitiveness of the polysilicon chain, yet rarely appears in mainstream analysis.

Tianxia Gongchang is a B2B platform covering 4.8 million producing factories — distinct from databases like Qichacha or Tianyancha that list all registered enterprises. The platform records only real producing factories, filtered through multi-source verification (satellite surface signatures, electricity usage, pollution permits, tax data, public ERP records) to weed out shell companies, trading entities, and non-production enterprises. On the polysilicon chain, the platform covers more than a dozen supporting categories — silica, industrial silicon, quartz sand, quartz crucibles, graphite components, hot-zone materials, reduction-furnace electrodes, mono pullers, wafer-cutting machines, silver paste, aluminum frames, glass, encapsulants, ties, trusses, transport vehicles, and testing equipment.

From the capability map of these factories, the true moat of the polysilicon industry is not just Tongwei, GCL, or Daqo, but the western Chinese ecosystem of quartz sand, crucible, graphite, gas, and electrode plants forming a "chemicals support cluster." Xinjiang Changji's quartz sand plants, Inner Mongolia Ordos's graphite plants, and Sichuan Leshan's high-purity gas plants all sit within a 500-kilometer procurement radius of the polysilicon makers. The corresponding logistics cost and inventory turnover advantages cannot be replicated overseas.

A specific detail: reduction-furnace electrodes are the critical consumable of the Modified Siemens process, with each furnace needing 24–36 electrodes priced around RMB 50,000 each and lasting 7–10 days. A 50,000-tonne polysilicon plant consumes 200,000 electrodes a year — a steady RMB 10-million-class order. Electrode plants concentrate in Sichuan Leshan, Xinjiang Shihezi, and Inner Mongolia Baotou, forming tight binding with polysilicon plants.

Quartz crucibles are the core consumable for mono pulling, with each crucible lasting about 300 hours, each mono puller consuming 30–40 crucibles a year at roughly RMB 10,000 each. China's quartz crucible production concentrates in Xuzhou (Jiangsu), Bengbu (Anhui), and Nanyang (Henan). Overseas quartz sand has long come from Spruce Pine in the US, and that channel has been repeatedly disrupted since 2024, pushing Chinese domestic quartz-sand plants in Xinjiang, Inner Mongolia, and Qinghai to accelerate expansion.

Silver paste is the critical non-silicon consumable for N-type TOPCon and HJT cells, with HJT silver use roughly double TOPCon's. Silver paste makers cluster in Changzhou (Jiangsu), Tongxiang (Zhejiang), and Shenzhen (Guangdong). The total capacity of China's top five silver-paste makers in 2025 is 5,000 tonnes, supporting around 500 GW of HJT modules at a 1 g/W silver use rate.

Encapsulants (EVA and POE) are critical for module lamination. Foster, Hiuv, and Sveck together had 15 billion m² of 2025 capacity, supporting 750 GW of modules. EVA encapsulant plants cluster in Shaoxing (Zhejiang), Suzhou (Jiangsu), and Jinshan (Shanghai). POE encapsulants long depended on imported POE resin from Dow and Mitsui; from 2025 Wanhua and Sinopec Zhenhai have broken through domestically, with 2026–2028 self-sufficiency expected to rise from 20 to 50 percent.

Chapter 8　Price Crash and Capacity Clear-Out: From RMB 300,000 to RMB 40,000

The trajectory of polysilicon prices from early 2022 to 2026 is the most dramatic chemical-raw-material episode of the past decade. Plotting each quarter's spot average, corresponding capacity, demand, and inventory yields a clear causal chain at every step.

Early 2022: global polysilicon capacity 1.4 million tonnes against demand of 1.4 million tonnes — tight balance, with N-type dense spot starting at RMB 240,000 per tonne. The April 2022 Xinjiang incident sparked short-term supply panic, pushing N-type dense spot to a historical ceiling of RMB 300,000 per tonne.

By end-2022, capacity announcements totaling 5 million tonnes had been disclosed from Tongwei, GCL, Daqo, Xinte, Asia Silicon, East Hope, Baofeng, Qinghai Lihao, and Ordos for 2022–2024 commissioning. Solar demand expanded from 240 GW to 350 GW in one year — but nameplate silicon capacity also expanded from 1.4 million to 2.6 million tonnes in the same year. The supply-demand balance flipped in Q1 2023.

2023 brought the avalanche. N-type dense fell from RMB 210,000 per tonne in January to RMB 62,000 in December — a 70 percent drop in the steepest single year in Chinese polysilicon history. Markets expected at least a 2024 recovery to RMB 80,000–100,000, but the early-2024 rebound from RMB 72,000 was ephemeral, with full-year 2024 average at just RMB 45,000.

2025 ran flatter. N-type dense full-year average RMB 40,000 per tonne, granular N-type average RMB 39,000, overall range RMB 38,000–45,000. This was the first year in polysilicon history that spot prices did not touch RMB 70,000 at any point. Infolink's year-end review showed the maximum monthly increase was only RMB 1,500 per tonne, well below the industry historical average of RMB 5,000.

2026 H1 has been even flatter. N-type granular traded at RMB 44,000 per tonne in March, slipped to RMB 42,000 in April, and stabilized at RMB 41,000 in May. BNEF and Bernreuter neutral forecasts call for full-year 2026 prices in the RMB 39,000–45,000 range, with annual average likely below RMB 45,000.

Capacity clear-out formally began in the second half of 2025. At least five tier-two polysilicon plants announced production cuts or shutdowns: two in Ordos, one in Changji (Xinjiang), one in Xining (Qinghai), and one in Leshan (Sichuan), totaling 300,000 tonnes of nameplate capacity. In December 2025, the six leading polysilicon makers (Tongwei, GCL, Daqo, Xinte, East Hope, Asia Silicon) jointly announced a "polysilicon integration fund" to raise USD 5 billion to acquire and shutter 1 million tonnes of low-end legacy capacity — roughly one-third of national total.

Chapter 9　Integration Counterattack: Vertical Integration vs. Independent Silicon Producers

In the course of polysilicon prices falling from RMB 300,000 to RMB 40,000, the parallel storyline of the industry is the contest between "integration" and "specialization." Module giants Longi, Jinko, Trina, and JA Solar have moved upstream into silicon; silicon giant Tongwei has moved downstream into modules; wafer giant Zhonghuan has built joint ventures with silicon producers. Each path is seeking a new equilibrium in the N-type market.

Tongwei is the most aggressive Chinese silicon producer integrating downstream. From 2018 acquiring Hefei LDK's cell capacity, entering modules in 2020, breaking into national top 10 in 2023, ranking sixth at 20 GW in 2024, and fifth at 28 GW in 2025. Tongwei's full-chain model — from silicon to cells to modules — is one of the few willing to extend vertical integration from upstream to downstream.

But Tongwei's 2025 full-year loss tells us integration is no panacea. When every segment loses money simultaneously, integration stacks the losses. Tongwei lost about RMB 5 billion in silicon, RMB 2 billion in cells, and RMB 2 billion in modules in 2025, totaling RMB 9–10 billion. This is the direct cost of "integration" in a downturn.

GCL took a different path. After completing the rod-to-granular pivot in 2024, GCL deliberately abandoned downstream integration and concentrated all resources on silicon, electronic-grade silicon, and energy storage. The strategy is "upstream specialization plus new-energy ecosystem"; in 2025 GCL's storage revenue in Xuzhou and Pinghu (Zhejiang) reached RMB 10 billion, offsetting silicon-segment pressure.

Daqo represents "pure specialization." Daqo only does silicon — no wafers, no cells, no modules. Its 2025 full-year revenue of USD 665.4 million came entirely from silicon. Daqo's logic: silicon should not be diluted by integration temptations — specialization's cost discipline is more reliable than integration's synergy claims. Daqo's 2025 cost advantage of USD 4.46/kg cash cost is the strongest evidence of "pure specialization."

The downstream giants' upstream moves are more complex. Longi has built silicon capacity in Yunnan and Inner Mongolia since 2022, with 300,000 tonnes by 2024; Inner Mongolia phase II was delayed in 2025 due to the silicon-price collapse. Jinko built a silicon JV in Xinjiang with Xinte in 2023, currently at 100,000 tonnes. Trina, JA Solar, and Canadian Solar have not entered silicon directly, instead locking 2–3 suppliers via long-term contracts.

Vertical integration's logic was clear in the P-type era: high module prices, distinct margin tiers, and integration captures end-to-end margin. But in the N-type era, fast downstream cell-tech iteration — TOPCon, HJT, BC, perovskite tandem — means any "locked-in" segment may be obsolete within two years. This iteration uncertainty has sharply reduced integration's allure in the N-type era.

Chapter 10　Overseas Build-out: Indonesia, Oman, Malaysia, US Map

Overseas polysilicon construction formed a visible wave from 2024 to 2026. Chinese silicon producers, driven by US UFLPA circumvention, EU CBAM compliance, and Southeast Asian downstream cell-module localization, alongside overseas giants pursuing tariff moats, energy subsidies, and localization, have each placed pieces in the Middle East, Southeast Asia, North America, and India. Every overseas base is a product of geopolitical and industrial-economic interweaving.

Oman has been the most-watched 2024–2026 overseas new base. United Solar Holding (USP) is building the Middle East's first major polysilicon plant at the Sohar port industrial zone, with USD 1.6 billion in total investment and 100,000 tonnes of planned capacity. Phase I 30,000 tonnes produced first silicon in early 2026. Funding included USD 480 million in long-term IFC debt, USD 260 million from Oman's Future Fund, USD 400 million in local bank debt and working capital, and a USD 30 million strategic investment from Waaree USA. USP signed an offtake agreement with India's Waaree Energies in January 2026, securing Oman silicon for Waaree's US and Indian module operations.

Oman's core differentiator is "natural gas + solar + low labor cost." Sohar natural gas supply at about USD 8 per kcal is far below China western industrial gas at RMB 1.9 per cubic meter (about USD 20 per kcal). Oman power comes in under USD 0.03 per kWh — less than half China's RMB 0.33 per kWh. USP also planned a 1 GW captive solar plant, projected to compress silicon carbon footprint below 20 kg CO₂ per kg silicon, qualifying for the EU CBAM low-carbon silicon exempt list.

Indonesia has a larger-scale polysilicon plan. In 2024 Indonesia's Ministry of Energy announced building Indonesia's first polysilicon plant at Batang Central Java industrial park, USD 800 million investment, 30,000 tonnes capacity; and a second plant at North Kalimantan KIPI industrial park, USD 3.2 billion investment, 160,000 tonnes capacity. The two projects total USD 4 billion and 190,000 tonnes. Chinese silicon producers — names officially undisclosed but rumored to include GCL and JA Solar — are likely JV partners.

Indonesia's logic is "coal power plus nickel integration." Indonesia is the world's largest nickel exporter; the new-energy chain's nickel demand has led Indonesia to write battery cathode, anode, and silicon into a green industrial-park strategy. Indonesia's abundant coal power offers about USD 0.05 per kWh — attractive for power-intensive silicon. But the weakness is coal's high carbon emissions, which may face extra CBAM costs.

Malaysia OCI Bintulu is the "special seat" in overseas silicon. OCIM in Bintulu leverages Sarawak hydropower, with a polysilicon footprint of 7 kg CO₂ per kg silicon — among the lowest globally. OCI announced in 2025 the Bintulu expansion from 35,000 tonnes to 56,600 tonnes, completing in 2026–2027. OCI–Tokuyama next-generation electronic-grade silicon JV also lands in Bintulu, starting construction in 2025 and commissioning in 2027.

The US is another key node. Wacker Charleston (Tennessee) completed expansion in 2025 to 20,000 tonnes solar plus 10,000 tonnes semi-grade. Hemlock (Michigan) holds 20,000 tonnes semi-grade. US policy — IRA advanced manufacturing tax credits, Section 232, UFLPA anti-circumvention — has been deployed densely through 2024–2026, providing real policy dividends to Wacker and Hemlock.

But the US polysilicon-to-module gap remains huge. 2025 US module demand of 40 GW implies 120,000 tonnes of silicon need, but Wacker plus Hemlock total solar-grade capacity is just 33,000 tonnes. The remaining 80,000-tonne silicon shortfall must come from Malaysia OCI, Oman USP, India Waaree, the Indonesia new bases, and Korea OCI.

Chapter 11　Energy and Carbon: EU CBAM, Energy Limits, and Green Silicon

Polysilicon is power-intensive: total electricity per tonne runs 48,000–55,000 kWh, corresponding to indirect emissions of 30–80 kg CO₂ per kg silicon depending on the power source (hydro, wind, solar, nuclear vs. coal, gas). Polysilicon's carbon footprint moved from "industry self-discipline" to "hard international trade constraint" across 2025–2026.

EU CBAM enters its "final transition phase" in 2026, with full carbon-tariff collection on imported solar modules, silicon, aluminum, steel, and fertilizers starting in 2027. CBAM's collection method is "actual carbon footprint × EU carbon price." EU carbon price ran EUR 80–90 per tonne in 2025, expected EUR 100 per tonne in 2026. For silicon, every 30 kg CO₂ per kg gap corresponds to EUR 3 per kg of CBAM cost — about RMB 25 per kg, roughly 60 percent of the current RMB 40 per kg silicon price.

CBAM's essence is "carbon pricing internationalized." Whether Chinese silicon producers maintain European competitiveness depends on whether their carbon footprint can match the EU benchmark — Wacker Burghausen's 20 kg CO₂ per kg silicon. The Chinese national average is about 50 kg CO₂ per kg — a 1.5× gap.

Chinese silicon producers' decarbonization has three paths.

First, green silicon. Switching power from coal to wind, solar, and hydro. Xinjiang, Inner Mongolia, and Qinghai silicon plants are actively linking up to local green-power markets — Tongwei Baoshan (Yunnan hydro), Daqo Baotou (eastern Inner Mongolia wind), and Xinte Ordos (western Inner Mongolia solar + wind) hit 60 percent green-power ratios in 2025. Green silicon footprint can be compressed below 20 kg CO₂ per kg, matching Wacker.

Second, energy efficiency. Lowering total electricity, water, and steam consumption. Modified Siemens fell from 58 kWh/kg in 2022–2025 to 48 kWh/kg; granular silicon over the same period fell from 22 kWh/kg to 11 kWh/kg. Every 1 kWh/kg cut corresponds to 0.6 kg CO₂ per kg silicon. The Modified Siemens efficiency ceiling is around 35 kWh/kg, beyond which only granular silicon can deliver further gains.

Third, by-product recycling. For every 10 percentage point gain in silicon tetrachloride, hydrogen, hydrogen chloride, and dichlorosilane recycling, carbon footprint falls 1 kg CO₂ per kg silicon. Chinese silicon by-product recycling rose from 80 percent in 2020 to 95 percent in 2025 — near physical limits.

Chapter 12　Research Outlook: 2026–2030 Chinese Polysilicon Landscape

A research-institute outlook must anchor in both data and logic — not extrapolation of a single variable. Cross-referencing 2025 capacity, demand, price, technology, carbon, and trade barriers yields a neutral forecast of the 2026–2030 Chinese polysilicon landscape. This is the judgment of the Tianxia Gongchang Industry Research Institute.

Forecast one: capacity clear-out completes its main phase by 2027. If the December 2025 integration fund executes as planned, 1 million tonnes of nameplate capacity will shutter across 2026–2027, leaving Chinese effective capacity around 2 million tonnes by end-2027. Global capacity-to-demand will then compress from 2.9× to 1.8×, entering "moderate oversupply." This is enough to support a return to RMB 50,000–60,000 per tonne, though not yet enough for all silicon producers to return to normal margins.

Forecast two: a "four-plus-two" structure. By 2030 China's silicon industry will settle into a stable structure of four leaders plus two specialists. The four are Tongwei (integrated full-chain), GCL (granular silicon plus electronic-grade), Xinte (green silicon plus semi-grade), and Daqo (specialized rod silicon); the two specialists are Asia Silicon (float-zone plus semi-grade) and East Hope (aluminum-silicon synergy). Other tier-two plants — Ordos, Qinghai Lihao, Baofeng — will be consolidated, exit, or pivot to electronic-grade.

Forecast three: granular silicon market share reaches 35 percent by 2030. Driven by GCL capacity expansion and ongoing N-type qualification, granular silicon's share of total Chinese silicon output will steadily rise from 24 percent in 2025 to 35 percent in 2030. The curve will not be linear — expect visible acceleration in 2027–2028 as GCL's overseas base (Middle East or Southeast Asia) ramps.

Forecast four: N-type premium keeps widening. N-type cell penetration reaches 95 percent by 2030, with the N-type premium expanding from today's RMB 3,000–5,000 per tonne to RMB 5,000–8,000. The widening is not just price spread but a hardening of the quality threshold — producers that can stably deliver N-type will capture 90 percent of the market, those that cannot will be locked into residual P-type or semi-finished material.

Forecast five: overseas capacity breaks 500,000 tonnes by 2030. Oman USP (100,000), Indonesia Batang and KIPI (190,000), Malaysia OCI expansion (56,600), Indian Reliance and Adani planned projects (100,000-tonne class), US Wacker plus Hemlock (33,000) together reach 500,000-tonne overseas capacity scale by 2030. China's global silicon share falls from 90 percent to 75 percent, but Chinese capital stakes in overseas projects may run 30–50 percent — "Chinese capital plus overseas production" becomes the 2030 silicon-globalization theme.

Forecast six: Chinese players break through in electronic-grade silicon by 2030. Chinese producers — particularly Xinte, GCL, Asia Silicon, and Daqo — will complete process breakthroughs in electronic-grade silicon across 2025–2030, reaching kilotonne-scale Chinese electronic-grade silicon capacity by 2030 and capturing 10 percent of the global semi-grade silicon market. This breakthrough will break the Japan-Korea oligopoly in float-zone-grade silicon, electronic-grade trichlorosilane, and ultra-high-purity silane.

Chapter 13　Risks: Module Downside, Overseas Anti-Dumping, and Export Controls

Risk-mapping is not pessimism — it anchors the future judgment in real downside scenarios. The polysilicon industry faces at least four risk groups across 2026–2030: module demand downside, overseas anti-dumping and anti-circumvention, export controls and technology blockade, and feedstock-price shocks.

Risk one: module prices continue to fall, pulling silicon prices to a new bottom. Module prices fell from RMB 1.00 per W in early 2024 to RMB 0.77 per W in early 2026 — a 23 percent decline. BNEF expects full-year 2026 module average to fall further to RMB 0.70 per W, stabilizing in 2027 and recovering in 2028. This implies silicon prices may oscillate in the RMB 38,000–42,000 per tonne range across 2026–2027. If the integration fund is delayed or poorly executed, silicon prices may retest RMB 35,000 in the second half of 2026.

Risk two: tightening overseas anti-dumping and anti-circumvention. The US launched anti-circumvention probes on Indonesia and Malaysia in 2024, expanded to four ASEAN countries (Cambodia, Malaysia, Thailand, Vietnam) in 2025, with 2026 final determinations expected to preserve anti-circumvention authority on "substantial transformation" of Chinese silicon, wafer, and cell in Southeast Asian processing. The EU also launched anti-circumvention investigations on Chinese modules in 2025, with potential targeted tariffs on Chinese silicon products possible in 2026–2027.

Risk three: gradual tightening of export controls and technology blockade. Chinese silicon's critical process equipment — reduction furnaces, CVD reactors, FBRs, cold hydrogenation, electronic-grade TCS distillation columns — has long depended on Germany's Borsig, US Praxair, Switzerland's Sulzer, and Japan's Asahi Kasei. From 2025, US EAR has placed parts of silicon-critical equipment on the dual-use list, creating implicit constraints on Chinese silicon equipment upgrades and overseas build-out.

Risk four: feedstock-price shocks from industrial silicon, coal, and electricity. Industrial silicon's 2024–2025 drop from RMB 15,000 to RMB 8,500 per tonne created a "cheap industrial silicon" pattern; if this reverses in 2026 — say from Yunnan or Sichuan hydro shortfalls or Xinjiang coal price rises — silicon's industrial-silicon procurement cost could rise 20 percent. Every RMB 1,000 per tonne industrial-silicon rise lifts silicon total cost by RMB 120 per tonne. The shock hits tier-two silicon plants far harder than the top tier.

Risk five: HJT-route disturbance to silicon demand structure. HJT cell silver use is double TOPCon's but silicon use is about 10 percent lower. If HJT replaces TOPCon as the mainstream cell route across 2027–2028, silicon demand growth slows by about 5 percentage points relative to expectations. This disturbance pressures silicon capacity timing.

Risk six: perovskite-tandem disruption. Perovskite-tandem cell silicon demand structure differs notably from pure silicon. If perovskite-tandem commercializes explosively across 2028–2030, silicon demand growth may slow another 10 percentage points. This is a long-run risk and not yet clear in 2026–2027.

Risk seven: tier-two silicon "debt avalanche." 2025 combined net losses across Tongwei, GCL, Daqo, and Xinte approach RMB 20 billion; tier-two silicon plants together lose around RMB 10 billion. Tier-two asset-liability ratios mostly run above 70 percent, some above 80 percent. If silicon prices retest the bottom in 2026, tier-two cash flows may not survive to 2027 integration-fund activation.

Chapter 14　Data Sources and Methodology

This report draws on five categories of sources:

First, company filings and annual reports. Tongwei 2024 annual report and 2025 quick reports; GCL Technology 2025 interim report; Daqo New Energy 2025 Q4 and full-year (draft 20-F and PR releases); Xinte Energy 2024 annual and 2025 interim; Asia Silicon 2024 annual; Wacker Chemie 2024 and 2025 annual reports; Hemlock parent (Corning and Shin-Etsu) 2024 and 2025 financials; OCI 2024 and 2025 annual reports.

Second, industry research. Bernreuter Research November 2025 global polysilicon capacity audit; BloombergNEF Q4 2025 global PV installation forecast; Infolink Consulting 2025 weekly silicon pricing and quarterly outlook; SunSirs end-2025 global silicon demand forecast; Mordor Intelligence 2025 polysilicon market report; Energy Trend 2025–2026 silicon price tracking; PV Magazine 2025–2026 coverage.

Third, government and association data. China Photovoltaic Industry Association (CPIA) 2025 silicon data weekly; China Non-ferrous Metals Industry Association Silicon Branch capacity and price statistics; China NDRC 2025 unit-energy-consumption ceiling standard; EU CBAM regulatory documents; US Section 232 investigation public documents.

Fourth, international authoritative analysis. IEA Solar PV Global Supply Chains report; Fraunhofer ISE electronic-grade polysilicon electricity consumption analysis; Reuters and Nikkei Asia coverage; Pyramids and Pagodas in-depth analysis on Chinese polysilicon; CSIS analysis on Chinese solar supply chain policy.

Fifth, platform and chain-factory data. Tianxia Gongchang, as a B2B platform covering 4.8 million producing factories, provided data collection and cross-validation on silicon downstream supporting factories (quartz sand, quartz crucible, hot-zone graphite, reduction-furnace electrodes, high-purity gases, silver paste, aluminum frames, glass, encapsulants) — the principal data basis for the "supporting-factory capability atlas" chapter.

Methodology. The report follows a three-layer "industry plus company plus product" research method: industry-layer anchored on global capacity, demand, and price structure; company-layer focused on the strategies, finances, and technology of Tongwei, GCL, Daqo, Xinte, Asia Silicon, Wacker, Hemlock, and OCI; product-layer drilling into the pricing, supply-demand, and technical metrics of N-type dense, recharge, granular silicon, electronic-grade silicon, semi-grade silicon, float-zone-grade silicon, electronic-grade TCS, and silane gas.

Forecast boundaries. The forecast sections use a "neutral scenario" baseline, assuming: global PV new installations grow at a 5 percent CAGR across 2026–2030; module prices stabilize in 2027 and recover modestly from 2028; the silicon integration fund completes its main phase in 2026–2027; CBAM is formally collected in 2027 with execution details largely matching current regulatory drafts; geopolitical tail risks do not materialize. Substantive deviation from any of these assumptions would alter the 2030 industry landscape.

Acknowledgements and disclaimer. This report is independently authored by the Industry Research Institute and was not funded or commissioned by any silicon producer, module maker, or government body. All specific numbers, prices, market shares, and corporate strategies referenced are based on public sources; views expressed are the independent judgment of the research team and do not constitute investment advice. The platform of 4.8 million producing factories at the heart of the supporting-factory chapter is a research data source for industry studies, supply-chain due diligence, and academic research; readers seeking specific supporting-factory information may query the platform for the latest factory directories and capability lists.

Research-institute judgment: 2026–2030 is the key five-year window for Chinese polysilicon's transition from "oversupply, low price, expansion" to "balance, pricing power, specialization." Silicon producers that survive the cycle will be the leaders combining cost discipline, N-type quality, decarbonization capability, and customer-service orientation. The fate of other tier-two silicon producers will be decided largely by the integration fund, debt-avalanche risk, and overseas build-out. This is a reshuffling more violent than 2012–2013, and also Chinese polysilicon's second opportunity to stand at the center of the world stage.