Chapter 3

The Polysilicon Engine

TBEA's largest subsidiary by assets is also its only loss-maker. Xinte Energy — 66.61%-owned and Hong Kong–listed — houses the high-purity polysilicon business, whose product-level gross margin fell from +71.10% in FY2022 to −47.84% in FY2025 as output was deliberately cut to 37% of capacity [1] [2]. Two structural features cushion the group: roughly a third of the loss sits with Xinte's minority holders, and the capacity build is finished.

Sources: FY2025 Annual Report, Analysis of Major Subsidiaries [3]; Photovoltaic Product Information [4].

What TBEA consolidates

Xinte Energy Co., Ltd. (1799.HK) is a self-contained polysilicon and solar company that TBEA controls but does not wholly own. On a full-consolidation basis it is the group's heaviest entity: ¥82.1bn of total assets and ¥45.1bn of liabilities, against ¥37.0bn of net assets [5] [6]. TBEA carries all of that on its balance sheet even though it owns two-thirds of the equity. In FY2025 Xinte generated ¥15.3bn of revenue and a ¥1.33bn net loss — the loss narrowing sharply from ¥4.04bn in FY2024, when revenue was ¥21.2bn [7] [8].

Set against its siblings, the picture is stark: the entity that owns the most assets earns the least.

Source: FY2025 Annual Report, Analysis of Major Subsidiaries [9].

Tianchi's coal-and-power operation earned ¥3.39bn and the transformer group ¥1.70bn; together they more than covered Xinte's loss and delivered the group's ¥5.95bn of attributable profit [10]. Xinte is the reason group earnings swing; it is not the reason they exist.

How far the polysilicon economics fell

The segment disclosure understates the collapse. TBEA's "new energy" segment — which bundles inverters, EPC engineering, and solar-station operations with polysilicon — reported a 0.59% gross margin on ¥13.6bn of FY2025 revenue [11]. The polysilicon product on its own tells a harder story: revenue of just ¥2.92bn at a −47.84% gross margin [12]. Three years earlier the same product sold ¥25.3bn at a 71.10% margin.

Sources: FY2022 [13], FY2023 [14], FY2024 [15], FY2025 [16] Annual Reports, Photovoltaic Product Information.

The margin swing translates into roughly ¥19bn of gross profit lost off a single product line: about ¥18bn of gross profit in FY2022 became a ¥1.4bn gross loss in FY2025.

Source: derived from reported polysilicon product revenue and gross margin, FY2022–FY2025 Annual Reports [17].

Two responses run underneath the margin line. First, TBEA cut volume hard: FY2025 polysilicon output was 9.64万吨 (96,400 tonnes) against roughly 260,000 tonnes of nameplate capacity — a 37.08% utilization rate, and output down 51.51% year on year [18]. The company calls this "self-disciplined production control" (自律控产) — running plants for maintenance and technical upgrades rather than chasing loss-making tonnes. Second, this is an industry price problem, not a TBEA-specific one: the FY2024 report states plainly that the market price of polysilicon had "fallen below producers' cost" and stayed there [19]. A −47.84% gross margin at a third of capacity is a company selling well below the cost of what it makes, while trying to make as little of it as pricing allows.

The minority buffer

Here the ownership structure matters more than it first appears, and it cuts two ways.

Because Xinte is only 66.61%-owned, roughly a third of its losses accrue to Xinte's own minority shareholders, not to TBEA's. That shows up in the minority-interest line. In FY2024 minorities absorbed a ¥541m loss, which lifted TBEA's attributable profit above the group's total net profit; in FY2025, as Xinte's loss narrowed, the same line flipped to a small +¥47m [20]. Group management attributes the FY2025 recovery in profit chiefly to polysilicon losses narrowing, alongside profit growth in transmission equipment and gold [21]. The listed-subsidiary structure means TBEA's shareholders never ate the full polysilicon loss.

The other side of the same fact is the balance sheet. TBEA consolidates 100% of Xinte's ¥45.1bn of liabilities and 100% of its capital programme, while owning two-thirds of the equity. It is not retreating from that exposure — in April 2025 TBEA raised its Xinte stake from 64.52% to 66.61%, paying ¥417m to buy in shares from employee partnerships [22]. And it continues to fund the capital-heavy businesses partly with outside money: group subsidiaries absorbed ¥4.59bn of fresh minority-shareholder capital in FY2025, and minority equity on the consolidated balance sheet stands at ¥26.5bn [23] [24]. The economics are therefore split: about two-thirds of Xinte's profit and loss reaches TBEA holders, but the whole of its debt and capex sits on the group.

The impairment record confirms the pain is being taken, not deferred. Xinte's fixed-asset and inventory write-downs drove a ¥3.56bn group asset-impairment charge in FY2024 — most of that year's ¥4.14bn attributable profit — easing to ¥976m in FY2025 as the worst of the mark-downs passed [25].

Where the cycle stands

The case set out at the start of this report turns on whether the steady core is worth owning through the polysilicon down-cycle; the state of that cycle is therefore the live variable. Three points frame it.

The price has stabilised but not recovered. N-type polysilicon traded around 31.5–33.7 RMB/kg in late June 2026, a tentative floor after prices spent 2025 and early 2026 below cash cost, with Chinese producers in Xinjiang and elsewhere still cutting output to hold the line. At those levels much of the industry, TBEA included, is still not covering full cost — consistent with a product gross margin that was −47.84% in FY2025.

The capital build is essentially done. TBEA's polysilicon nameplate already stands near 260,000 tonnes a year, with no new lines under construction disclosed [26]. Group capex was still heavy at ¥22.1bn in FY2025, but that spend is now flowing to the coal, power and materials businesses rather than to new polysilicon capacity [27]. The polysilicon plant is built; the question is utilisation and price, not further investment.

And the FY2025 profit recovery did not come from a polysilicon rebound. It came from the loss narrowing — helped by lower impairments, a stronger transmission and gold business, and a mark-to-market gain on a listed equity holding — not from polysilicon returning to profit [28].

The reasonable read is that Xinte is a large, deeply cyclical, loss-making asset whose downside is genuinely shared with outside shareholders and whose capital demands are largely behind it — a drag the group can carry, rather than a hole it must keep filling. The strongest fact against that read is size: at ¥82bn of assets and ¥45bn of consolidated liabilities, Xinte is too big to be treated as a rounding error, and a renewed leg down in polysilicon prices, or a decision to restart idled capacity into a weak market, would deepen the loss and the impairment risk again. What would change the read in the other direction is straightforward to watch: polysilicon prices clearing producers' cash cost on a sustained basis would move Xinte from a roughly ¥1.3bn drag toward breakeven, and the utilisation rate — 37% in FY2025 — is the single line that will show it first.