Morgan Stanley: China's wafer foundry capacity and chip supply will meet core sovereign demand around 2028

Morgan Stanley published a report on China's artificial intelligence (AI) image processors (GPUs), indicating that China has made substantial progress in alleviating equipment and wafer foundry bottlenecks over the past 12 months. It is expected that with policy support, China's wafer foundry capacity and chip supply will meet core sovereign demands around 2028.

The report states that policy support can accelerate early development, but long-term value depends on commercial competitiveness. Chinese AI GPU suppliers must demonstrate compelling economic benefits to sustain growth beyond 2028. The analysis shows that with lower prices, cheaper electricity costs, and increasingly improved infrastructure, the total cost of ownership (TCO) for Chinese AI data centers is competitive. In the inference workload domain, the cost per code unit (token) is more critical than peak performance, further enhancing the competitive advantage of Chinese solutions.

The firm believes that China's localized strategy to compensate for process technology disadvantages by expanding chip, wafer, and equipment scales is continuing to yield results. The optimistic scenario assumes that domestic GPUs will expand into training workloads and may gain overseas adoption; the pessimistic scenario assumes that the differentiation advantage fades, leading to commoditization and industry consolidation. The firm also predicts that the market size for China's AI chip sector will reach $67 billion by 2030, implying a compound annual growth rate of 23% from 2024 to 2030. It is expected that China's self-sufficiency rate for AI chips will reach 76% by 2030.

The firm remains optimistic about China's AI semiconductor supply chain, including SMIC (00981.HK), NAURA (002371.SZ), and ASMPT (00522.HK), as well as AI chip investments that help strengthen the strategic layout of Chinese internet platforms