Summary of Key Points
In the first half of 2026, the surge in demand for AI computing power has propelled the power semiconductor industry from a "backstage supporting role" to a "frontstage protagonist." However, the performance of leading companies has varied significantly: some have seen a sharp decline in profits, while others have experienced growth in revenue without a corresponding increase in profits, and yet others have seen both revenue and profit increases. The underlying factors include differences in business models (IDM full-stack production vs. fabless models), industry-wide price increases, and uneven demand across different segments (strong demand for AI data centers versus temporary pressure in the new energy sector).
1. AI Computing Power Becomes the New Engine of the Industry: Power Semiconductors Go from "Invisible" to "Essential"
Power semiconductors essentially act as "electricity managers," responsible for converting, controlling, and optimizing power usage, ensuring that servers use energy more efficiently and without waste. In the AI era, servers process large amounts of data, resulting in higher power consumption (for example, an AI server can consume 3-5 times the power of a regular server). The more servers are packed into a data center, the greater the power density required, making high-quality power semiconductors indispensable for building AI infrastructure.
2. Divergent Performance Among Leading Companies
- StarPower Semiconductor: Profits plummeted by 76%. Revenue decreased slightly by 0.4%, but net profit dropped from 276 million yuan last year to 65 million yuan, and gross margin fell from 30% to 21%. The reasons include rising raw material costs, previous price cuts on products, and increased research and development and financial expenses. Although the new energy segment (automotive, photovoltaic) saw a 16% decline, the industrial control segment (including AI data center power supply) grew by 40%.
- New Energy Semiconductor: Revenue increased by 25%, but net profit remained virtually unchanged. However, there was a improvement in the second quarter: revenue rose by 25%, profit increased by 46%, and gross margin rebounded to 31%, indicating that the business is recovering.
- Silan Microelectronics & China Resources Microelectronics: Both companies saw increases in both revenue and profit. Silan Microelectronics' revenue grew by 14%, and profit increased by 94% (partly due to stock investment gains); China Resources Microelectronics' revenue increased by 17%, and profit increased by 113%. In the second quarter, revenue reached a record high, with MOSFETs (common power devices) growing by over 20% and 8-inch IGBTs (used in AI servers) growing by 119%.
3. Business Models Are Crucial: IDM Companies Profit Easily, While Fabless Companies Suffer
- IDM Model (Full-Stack Production): Companies like China Resources Microelectronics and Silan Microelectronics produce everything from wafer fabrication to packaging and testing, possessing their own production capacity. During supply shortages, they do not rely on external foundries, allowing them to control costs and have greater profit flexibility when raising prices (for example, China Resources Microelectronics saw a direct increase in profits after raising prices).
- Fabless Model: Companies like StarPower Semiconductor design their own chips but outsource production. Rising foundry costs and depreciation pressures squeeze their profits. A UBS report also notes that IDM companies have more profit flexibility during supply shortages, while fabless companies have limited expansion options.
4. Two Waves of Price Increases in the Industry
Since the beginning of the year, the power semiconductor industry has experienced two rounds of price increases:
- Early this year: China Resources Microelectronics and Silan Microelectronics were among the first to raise prices by more than 10%.
- July: Nearly 20 companies raised prices by another 15-25%. The reason for the price increases is the surge in demand (from AI) that outpaced supply. Companies raised prices to pass on the costs and increase profits. IDM companies, with their own production capacity, benefited the most from these price hikes, but fabless companies saw their profit margins reduced due to increased foundry costs.
5. Uneven Demand Across Segments
- AI Data Centers: Demand is strong, with China Resources Microelectronics' 8-inch IGBTs growing by 119% and StarPower Semiconductor's industrial control segment (including AI power supply) growing by 40%, both driven by AI demand.
- New Energy (Automotive, Photovoltaic): These sectors are experiencing temporary pressure, with StarPower Semiconductor's new energy segment declining by 16%, possibly due to market saturation or policy changes.
In summary, AI computing power has opened new opportunities for the power semiconductor industry, but whether companies can capitalize on these opportunities depends on whether they have their own production capacity (IDM model) and whether they can meet the specific demands associated with AI. This analysis makes the financial and business trends more accessible to a general audience, highlighting how AI has boosted the industry but how different companies are affected by their respective business models and positioning.