Xinyuan Co., Ltd. (688521.SH) is a company that relies on its proprietary semiconductor IP to provide customers with platform-based, comprehensive, and one-stop chip customization services and semiconductor IP licensing services. The company’s business revenue mainly comes from IP licensing and one-stop chip customization, including mass production. The company’s IP licensing business holds the largest market share in China and ranks eighth globally, with intellectual property licensing revenue ranking sixth worldwide, and its IP types ranking second among the top ten IP companies globally. The company significantly benefits from the trend of AI development, driven by market demand for AI computing power, with rapid growth in its data processing, computer and peripheral, and automotive electronics sectors. In response to the industry trend where major internet companies and automotive manufacturers are increasingly developing their own chips, the company continues to strengthen its core competitiveness in semiconductor IP and chip customization services while actively embracing emerging technologies and market demands, with the potential to grow alongside internet and automotive clients.
1. Chip Customization: Continuous Increase in Customization Demand, Unceasing Innovation Drive The global demand for AI computing power is growing at an unprecedented rate. According to Marvell data, the capital expenditure of the top four hyperscale cloud service providers in the U.S. is expected to reach $327 billion by 2025, with total data center capital expenditure expected to reach $593 billion, expanding at a compound annual growth rate (CAGR) of 20%. This is expected to drive the data center market size to exceed $1 trillion by 2028, with custom computing chips (Custom Compute) and additional components (XPU Attach) becoming the core growth engines. Customized chips will become the decisive factor for enterprises in the AI era. Traditional general-purpose computing chips can no longer meet the diverse demands of AI workloads, leading to more market opportunities for custom computing. AI algorithms (such as the Transformer architecture) are updated every few months, making it difficult for general-purpose chips to dynamically adapt to new algorithm requirements. AI applications cover scenarios such as cloud training, edge inference, and real-time analysis, where general-purpose chips cannot balance flexibility and performance. From training to inference, from large language models to real-time data analysis, AI tasks demand higher efficiency, energy efficiency, and flexibility, especially when facing complex neural networks and large-scale data processing, where the bottlenecks of traditional hardware are increasingly evident. Custom chips can achieve improved training efficiency, reduced heterogeneous computing latency, and optimized performance ratios through hardware optimization. Marvell points out that the custom XPU market (such as dedicated AI acceleration chips) will grow at a CAGR of 47% from 2023 to 2028, while the XPU additional components market (such as high-speed interfaces and memory pooling chips) will grow at a staggering CAGR of 90%, together contributing over $55 billion in total addressable market (TAM), and the company is expected to continue benefiting in this field. The downstream market comprehensively covers data center cloud computing, automotive electronics, smart homes, and the Internet of Things, with tech giants accelerating their entry. Broadcom began developing custom AI chips for clients like Google several years ago, initially with limited business scale. As the wave of AI applications sweeps through the tech industry, especially in the field of generative AI, the demand for custom chips has rapidly increased, becoming a significant driver of Broadcom’s performance. According to the company’s latest financial report, revenue from AI processing chips and related network chips is expected to reach $5.1 billion in FY25Q2, a year-on-year increase of about 60%, accounting for about one-third of the company’s total revenue. U.S. chip manufacturers Marvell and Taiwan’s MediaTek are also accelerating their layout in the custom AI chip field, with some businesses already putting pressure on Broadcom. Domestic internet giants like Baidu, Alibaba, and Tencent primarily use ASIC architecture for their self-developed AI chips, mainly applied in their own business scenarios. Leading integrated circuit manufacturers such as AMD, Intel, and TSMC have successively released related Chiplet solutions, interface protocols, or packaging technologies. Xinyuan has become one of the first companies in mainland China to join the UCIe alliance, advancing its Chiplet technology in areas such as interface IP, Chiplet chip architecture, advanced packaging technology, and solutions for AIGC and smart mobility, with its Chiplet technology leading in the fields of generative artificial intelligence big data processing and high-end intelligent driving. 2. IP Licensing: Market Scale Significantly Rising, Company Approaching Volume Release Point Compared to traditional chip R&D models, the semiconductor IP licensing model can help manufacturers save over 50% in time and cost. The global semiconductor IP market is expected to reach $8.49 billion in 2024, with an average annual compound growth rate of 16.78% over the past five years. According to the China Economic Information Agency, the global semiconductor IP market is projected to rise to $14.35 billion by 2029. Currently, the global semiconductor IP market is still dominated by international giants like ARM and Synopsys. Although the demand for semiconductor IP in the Chinese market accounts for nearly 30%, the domestic IP self-sufficiency rate is only 8.52%, indicating a low self-sufficiency level. The company’s market share is 1.3% in 2024, indicating significant growth potential. The company has IP product layouts in areas such as Bluetooth headsets, smartwatches/bands, virtual reality-based smart glasses, and automotive-grade products.