China Securities Co., Ltd.: The computer sector is currently experiencing a turning point of fundamental repair and AI paradigm shift resonance.

China Securities Co.,Ltd. released a research report stating that in 2026, the computer sector is experiencing a resonant turning point in fundamental repair and AI paradigm transfer. The model's iterative capabilities have no upper limit, the penetration rate of Claw applications is rapidly increasing, domestic computing power demand is soaring, and AI infrastructure is entering a new stage. In the second half of the year, it is recommended to focus on the following: 1) For the AI industry, it is recommended to focus on the direction of computing power that is experiencing price increases and shortages, infrastructure and cloud industries that improve efficiency, and application directions with high business potential. 2) For non-AI industries, investment recommendations include focusing on opportunities catalyzed by policies such as digital currency 2.0, autonomous driving, and commercial space under policy catalysis. Key points from China Securities Co., Ltd. are as follows: Review of the computer sector. As of the end of April, the computer index was down by -4.11% year-to-date, underperforming the Shanghai and Shenzhen 300 by about 6 percentage points, with computing power and Agent applications performing better while the industrial end faces pressure. Revenue in 2025 increased by 8%, net profit attributable to shareholders increased by 57%, and the first quarter of 2026 saw a significant increase in non-GAAP net profit, with continued profit recovery and a decrease in expense ratio offsetting pressure on gross profit margin. Institutions have been underweight for eight consecutive quarters, with a PE ratio of 81x, historically high at the 90th percentile, but PS ratio still cost-effective within TMT. In the second half of the year, focus on three main themes: computing power price increases and shortages, cloud and infrastructure price increases, AI applications with calculable ROI, and other strong policy catalysis directions. Large-scale model iteration has no limit, continuously leading global AI development. Overseas, OpenAI, Anthropic, and Google are leading the way, focusing on B-side Agent ecosystems, breakthroughs in intelligent boundaries, and full matrix coverage. Domestic flagships such as DeepSeek-V4 and Kimi K2.6 are on par with or even surpass top overseas levels in programming, long-form text, and multimodal fields. Model capabilities double every 3.5 months, and the rapid decline in inference costs has led to a surge in token calls, with the daily average number of tokens in China increasing from one trillion times in early 2024 to over 200 trillion times in March 2026, highlighting the competitive advantage of domestic models due to cost efficiency. Sudden increase in domestic computing power demand, AI infrastructure enters a new stage. The landing of Agents is driving the comprehensive reconstruction of infrastructure from an inference platform to a tool chain, with MCP agreements, security sandboxes, and observability becoming the most certain directions. Inference optimization accounts for 60%-80% of the TCO of AI systems and is the highest ROI priority investment. CPUs have regained value due to the surge in Agentic AI demand, with CPU time consuming over 80% of Agent tasks, global price increases and shortages continuing, domestic CPUs benefiting first under the dual catalysts of innovation and AI, and driving the shipment of memory interface chips and servers. CLAW leads to a new form of Agents, with B2B focusing on product capabilities. OpenClaw's emergence marks AI's transition from "assisting tool" to "digital employee", with token consumption driving price increases throughout the industry chain, and major domestic factories accelerating the layout of their own Claw products to compete for ecological access. AI application directions such as manufacturing, scientific research, power trading, and content creation are expected to be the first to enter the value realization period. DIGIHUMAN RMB 2.0 officially launched, creating a billion-dollar financial infrastructure market. The increasing uncertainty of the international situation, the strategic window for the internationalization of the RMB, DIGIHUMAN RMB entering the 2.0 era, introducing the M1/M2 measurement framework and implementing interest payment, and operational institutions are expected to rapidly expand. This will stimulate the demand for core banking system transformation and end-to-end upgrade, with potential market space exceeding one billion. Risk warning: Bad debt risk of accounts receivable; intensification of industry competition; impact of changes in the international environment.