Solar-grade polysilicon is sorted into metallurgical-grade (MG-Si, ~98 to 99% purity), solar-grade (SoG-Si, 6N to 7N) and electronic-grade (9N+), with mono wafer lines effectively requiring 7N+ material because impurity-driven lifetime loss directly caps cell open-circuit voltage [S1].
Spec discipline at this tier means fixing the resistivity band (typically 1 to 3 Ω·cm for p-type mono Czochralski), the oxygen and carbon target (≤10 ppma each for high-lifetime wafers), and the format (chunk, rod or granular) before any price talk begins, since each variable reshapes the cost stack differently [S1]. Granular polysilicon feeds fluidised-bed reactor (FBR) mono lines and avoids the crushing step that chunk material requires, which is why Siemens-process chunk still dominates butcher-bar mono lines and granular dominates FBR [S1].
Purity Bands, Format Split, and the Siemens-versus-FBR Decision
The Siemens process and the fluidised-bed reactor (FBR) are the two principal production routes for polycrystalline silicon, with Siemens-process output typically delivering higher purity grades for advanced semiconductor and high-efficiency PV applications, while FBR is associated with reduced energy consumption per kilogram [S3]. Buyers evaluating 2026 offtake should track three decision criteria: purity grade (N-level), manufacturing route (Siemens versus FBR), and contractual indexation to solar module ASPs [S3].
GB/T 12963 (Chinese national standard for solar-grade polysilicon) and SEMI PV-grade specifications are the two named quality anchors for 2026 procurement, though any specific revision date or purity threshold should be verified against the actual GB/T or SEMI document [S3]. The cost stack of a c-Si module still tracks silicon content quoted in polysilicon (USD/kg), wafer (USD/pc), cell (USD/W) and laminate (USD/m²), three of which carry different units, which is why buyers normalise everything to USD/W before comparing suppliers [S1].
Wafer Format, Thickness, and Lifetime Anchors That Constrain the Buy
The 2026 wafer field is a duopoly of M6 (166 mm) legacy stock, M10 (182 mm) mainstream volume, and G12 (210 mm) large-format mono, with diamond-wire sawing universal at wire diameter 55 to 65 µm; wafer thickness sits at 150 µm for M10 and 160 µm for G12, trending toward 130 µm and 140 µm respectively as kerf-loss economics tighten [S1].
Procurement specs at the wafer step should pin diameter tolerance (±0.25 mm), thickness tolerance (±10 µm), TTV (total thickness variation, ≤5 µm for premium mono) and bow/warp (≤50 µm); those four numbers written into the PO are what separate a Tier-1 wafer from a usable-but-suspect lot [S1]. Minority-carrier lifetime is set at the wafer and cell step, so procurement discipline upstream pays for itself downstream: a controlled-cooling 150 µm M10 mono PERC wafer routinely shows bulk lifetime above 1 ms, while a poorly gettered multicrystalline wafer lands at 30 to 100 µs, a 10x to 30x spread that buyers are effectively paying for [S1]. This is why silicon-content cost modelling pairs directly with wafer process cost modelling, as the Solar Cell Manufacturing Cost Breakdown reference walks through in detail.
The 630 Rush, Anti-Involution, and the Stockpile Window
As of March 2026, the photovoltaic industry is facing the convergence of two deadlines: the April 1st photovoltaic export tax rebate policy window, and the domestic grid connection deadline of June 30th, driving a concentrated installation rush across the chain [S4]. SMM data cited for March 2026 shows global cell production schedules for sampled Chinese enterprises rose 26% month-on-month, indicating a significant recovery in production enthusiasm, with cell inventories continuing to decline and overseas orders plus downstream trader orders becoming the primary demand sources [S4].
On the supply side, planned maintenance and production cuts by leading enterprises, combined with reduced operating rates at certain facilities, drove polysilicon production significantly down from a cyclical peak in October 2025, and 2025 overall supply dropped substantially compared to peak levels in previous years [S4]. Polysilicon prices saw a significant quarterly increase in Q3 2025, helping some producers reduce losses, and with continued production cuts and suspensions through 2026, prices and profitability are positioned for stabilisation and recovery, making the March to April 2026 window a critical stockpile gate for any PV offtake tied to Q2 delivery [S4]. For module-line planners running this same window, the Solar Cell Industry 4.0 module-line spec map lays out the downstream equipment gates that absorb the silicon pull-through.
ESG, Energy Consumption Standards, and the Exit of High-Carbon Capacity
The Standardization Administration of China completed the drafting of three national standards, including the "Energy Consumption Limits per Unit Product for Polysilicon and Germanium," with stricter energy-consumption limits expected to drive the gradual exit of existing high-carbon capacity [S4]. Producers that can leverage renewable electricity, improve furnace efficiency, and use biomass reducing agents are gaining a distinct competitive advantage, and for PV companies the supplier choice is now an ESG-strategy element, not just a cost line [S4].
Capacity expansion in wafers and modules is projected to stagnate in 2026 with some segments potentially experiencing production cuts, and the alleviation of oversupply is expected to drive industry consolidation toward leading producers [S4]. Anti-involution remains the central theme of the 2026 photovoltaic industry, and the 630 rush combined with the April 1st export rebate window is the most concentrated demand event on the calendar, which is why buyers who have not pre-locked offtake by March face a steepening cost curve into Q2 [S4].
Procurement Decision Matrix: Solar-Grade versus Electronic-Grade
Solar-grade polysilicon (6N to 7N) is consumed by multi- and monocrystalline PV ingot growers, while electronic-grade (9N+) feeds 300 mm wafer fabs for logic and memory, and the historical electronic-industry focus is the source of the high-purity process know-how that underpins the Siemens-route advantage [S3]. On a per-kilogram basis, electronic-grade commands a premium that scales with N-level, but the spread between 6N and 9N is also driven by lot size, contract length, and trace-metal certification rather than purity alone [S3].
For PV offtake, three options line up against the standard procurement criteria: (1) Siemens-process chunk at 7N+ for mono Czochralski lines, the legacy default with the broadest supplier base and the most forgiving crushing logistics; (2) FBR granular at 6N to 7N for fluidised-bed mono lines, with lower energy intensity but tighter format compatibility, the route gaining share as ESG pressure rises; (3) off-spec or recompound feedstock for multicrystalline and merchant-ingot lines, the lowest cost per kg but increasingly throttled by the new energy-consumption standards drafted in March 2026 [S1][S3][S4]. Buyers specifying 7N+ mono feedstock with ≤10 ppma oxygen and carbon, granular format for FBR lines or chunk for Czochralski, and a renewable-power attestation in the QA packet, are positioning for both cost stability and ESG-aligned offtake through the 2026 cycle.
Risks, Failure Modes, and What Wrecks a Polysilicon PO
The most common procurement failures are upstream of the price line: a chunk-format order placed against an FBR-only producer, a 6N lot accepted into a 7N+ mono line, or a 25-ton MOQ contract signed by a buyer whose real consumption is 8 tons per quarter, leaving them holding expensive inventory through the 630 rush and then again into Q4 destock [S2]. Spot-listing screens for "polysilicon" return a wide quality spread, from USD 1,800 to 2,000 per ton for bulk polysilicon offers with 25-ton MOQs, to USD 40 to 50 per kilogram for 99.9999% optical-grade granules with 1-kilogram MOQs, and that 25x to 40x gap is not arbitrage, it is grade separation [S2].
Downstream, the practical failure is lifetime loss: a polysilicon lot that passes ICP-MS trace-metal certification but slips on resistivity uniformity will produce a 1 to 3 Ω·cm target on average while individual ingot sections drift to 0.5 or 5 Ω·cm, which in turn caps cell Voc and quietly pulls 0.3 to 0.5 absolute efficiency points off the module line [S1]. The fix is to write the resistivity band, the oxygen and carbon ceiling, and the format into the PO before price talk, to require a renewable-power attestation in the QA packet, and to align contract tonnage with a realistic quarterly drawdown rather than the headline MOQ [S1][S4]. Trackable signals into Q3 2026 include the SMM monthly cell-production schedule print, any GB/T 12963 revision activity, and the next quarterly disclosure from the leading polysilicon producers on operating rate and inventory days.
Spec-level background on the components involved: linear guide, crossed roller guide, and pressure transmitter.