StarPower: AI Expansion Cannot Offset EV Slowdown; Profit Revision Maintains Neutral

Overview

Goldman Sachs released an earnings review report on StarPower (603290.SS), maintaining a neutral rating and a 12-month target price of 121.2 CNY. The report considers that the company is actively expanding its product lines from IGBT to SiC MOSFET, GaN HEMT and MCU, with end markets extending from EVs to AI data centers, industry, robotics and home appliances, with a clear long-term growth logic. However, given the slower-than-expected recovery in China EV market demand, increased R&D spending for new products, rising raw material costs, and new capacity depreciation burdens, the firm has significantly downgraded profit forecasts for the next three years, causing growth and margin pressure in the short term, hence maintaining a wait-and-see stance.

Core views

Significant downgrade in profit forecasts is the core change in this report. Based on disappointing performance in Q4 2025 and Q1 2026, Goldman Sachs downgraded 2026-2028 net profit forecasts by 38%, 22% and 28% respectively, and revenue forecasts by 18%, 14% and 20%. Management attributed the Q1 revenue gap to weak recovery in the China EV market, which currently accounts for about 50% of total revenue. Additionally, gross margin forecasts were also downgraded, mainly due to rising raw material costs and depreciation during the initial phase of new capacity commissioning, but as shipment ramp-up and capacity utilization improve, gross margins are expected to trend upwards in future years. R&D expense forecasts were adjusted upwards to match the investment pace for product portfolio upgrades and production line expansion. Although short-term performance is under pressure, the company's layout in AI-related application areas is viewed as an important medium-to-long-term highlight. StarPower is driving the evolution of power semiconductor devices towards high power density and high energy efficiency to capture emerging demands such as AI data center power supply units (PSU), IBC, hot-swap and VRM. At the same time, the company's transition from Fabless to Fab-lite mode continues to advance. It now has the capability to produce 60,000 pieces (6-inch) of automotive-grade SiC MOSFET chips and 300,000 pieces (6-inch) of 3300V and above high-voltage special power chips annually, laying the foundation for subsequent product structure optimization. Valuation levels show a hedging effect between "profit downgrades" and "valuation multiple upgrades." Goldman Sachs rolled the valuation benchmark year to 2027 and raised the target P/E ratio from 24.6x to 31.4x. This adjustment is based on peer forward P/E regression analysis against EPS growth rates, reflecting the valuation revaluation obtained by the power semiconductor sector due to the rise of AI data center applications. However, due to the large magnitude of profit forecast downgrades, the final 12-month target price remains unchanged at 121.2 CNY. The current target P/E ratio is still below one standard deviation below the company's historical mean, reflecting prudent considerations regarding the high proportion of EV business and slowed growth.

Analysis framework

The report adopts a 'fundamental driven + relative valuation' analysis framework. In terms of fundamentals, analysts first revise revenue and profit assumptions based on the latest quarterly performance and management guidance, paying special attention to the direct impact of changes in downstream EV market sentiment on the company's short-term revenue; secondly, they evaluate the stage compression of profit margins due to product expansion by splitting the cost structure (raw materials, depreciation, R&D), and judge the repair elasticity after capacity utilization improves. In terms of valuation, analysts did not use a fixed historical PE multiple, but established a peer 'forward PE-EPS growth' regression model, using the pricing rules of comparable companies in the industry to derive reasonable target multiples, thereby quantifying the valuation premium brought by new AI application scenarios, so that the target price reflects both long-term track value and the reality constraints of short-term performance revisions.

Methodology notes

• Valuation MethodPE/PEG valuation

  PE valuation method based on correlation between peer forward P/E and EPS growth rate

  The report does not simply apply the company's own historical PE or fixed industry average, but selects a set of comparable companies to establish a linear regression relationship between their forward PE and expected EPS growth, and then substitutes StarPower's own expected growth into the regression equation to derive a reasonable target PE. This method can quantify the growth premium brought by AI applications into the valuation multiple, reflecting dynamic changes in track revaluation better than static PE.

• Company Fundamentals and Financial FrameworkProfit Quality Analysis

  Analysis of profit margin phase pressure and repair path during capacity expansion period

  The report decomposes gross margin changes into three driving factors: raw material costs, new capacity depreciation and capacity utilization. It points out that in the stage where product lines and capacity expand simultaneously, depreciation and R&D investment will temporarily depress profit margins, but there is inherent repair momentum for margins as shipments ramp up and utilization rises. Such analysis helps investors distinguish between "structural deterioration" and "cost of cyclical investment", avoiding excessive pessimism during expansion troughs.

• Industry/Industrial Analysis FrameworkUpstream and Downstream Transmission in Supply Chain

  Performance transmission mechanism from downstream terminal demand changes to upstream semiconductor manufacturers

  The report directly anchors the adjustment of company revenue forecasts to the changes in the sentiment of the China EV market, reflecting the high sensitivity of power semiconductors as an intermediate link to downstream vehicle sales. When EV terminal demand accounting for about 50% of revenue falls below expectations, the income and capacity utilization of upstream chip manufacturers will be quickly impacted, which is the key transmission chain to understand short-term performance volatility.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• StarPower (603290.SS)

  Report covered target, product lines expanded to AI applications but dragged by EVs in short term

  Strengths

  Product lines successfully expanded from IGBT to SiC/GaN/MCU; AI data center power applications open new space; Fab-lite transition enhances supply chain autonomy; gross margin has repair elasticity after capacity utilization improves

  Weaknesses

  EV business accounts for about 50% of revenue and demand recovery slower than expected; new capacity depreciation and R&D expenses suppress short-term margins; rising raw material costs

  Comparison

  Target PE 31.4x is on the peer regression line, but below one standard deviation below own historical mean, reflecting EV business weight higher than pure AI targets

  Risks

  IGBT market growth weaker or stronger than expected; new product design introduction and market share acquisition slower or faster than expected; new product development progress delayed; intensified competition

Key data

• 2026E Net Profit Forecast Adjustment5.77 hundred million CNYDowngraded 38% from prior forecast; original forecast 9.36 hundred million CNY
• 2027E Net Profit Forecast Adjustment9.18 hundred million CNYDowngraded 22% from prior forecast; original forecast 11.78 hundred million CNY
• 2026E Revenue Forecast Adjustment46.15 hundred million CNYDowngraded 18% from prior forecast; original forecast 56.24 hundred million CNY
• 2027E Target P/E Ratio31.4xAdjusted up from prior implied 24.6x, reflecting valuation revaluation from AI applications
• 12-Month Target Price121.2 CNYMaintained unchanged; profit downgrade offsets valuation multiple upgrade
• 2026E Gross Margin Forecast26.7%Downgraded by 5.3 percentage points from prior forecast, affected by raw material costs and depreciation
• Automotive SiC MOSFET Annual Capacity6 ten-thousand pieces (6-inch)New capacity added in Fab-lite transition
• High-Voltage Special Power Chip Annual Capacity30 ten-thousand pieces (6-inch)3300V and above specifications, supporting industrial and AI applications

Impact & implications

The report believes that StarPower is in a stage where "long-term track upgrade" coexists with "short-term pain". The expansion of AI data center power supply applications has opened a second growth curve for the company and is the fundamental reason why the valuation multiple could be raised; but the cyclical slowdown of EV business and the digestion of new capacity take time, resulting in decreased visibility of profitability over the next one to two years. For the market, this means there may be a lack of strong upward catalysts for stock prices in the short term, and earnings projections need to wait for EV demand recovery or a significant improvement in the proportion of AI business revenue before there is a basis for upgrades. The fact that the target price is maintained unchanged itself also conveys a signal that "valuation has partially reflected AI expectations, but performance has not yet been realized".

Risks

• IGBT market growth stronger or weaker than expected
• New customer design adoption and market share acquisition speed faster or slower than expected
• New product development progress faster or slower than expected
• Industry competition level higher or lower than expected