Bank of China International: Two Sessions boost market confidence, technology assets being revalued.

The Two Sessions boost market confidence, and the revaluation of pan-technology assets is in progress, still at the "may" window for Siasun Robot&Automation and technology growth sectors. 01. Trends and Styles Fiscal stimulus and monetary easing boost market confidence. This week, the Two Sessions were successfully held, and this year's government work report on macroeconomic policy is more positive compared to last year, especially in terms of fiscal policy. The 4% fiscal deficit rate for 2025 is significantly higher than the level of the past 5 years, indicating an increased willingness of the government to leverage debt. In terms of monetary policy, this year it is explicitly stated to implement a "moderately loose monetary policy", more positive compared to last year. Additionally, both the government work report and the post-meeting statement from the central bank officials clearly stated the intention to "timely cut reserve requirements and interest rates", which helps boost market confidence. The focus of this year's economic policy has shifted towards benefiting people's livelihoods and promoting consumption. The logic of weak earnings recovery but strong valuation support in the A-share market is expected to continue after the Two Sessions. Uncertainties in external demand have increased. Overseas markets remained weak this week, with the Trump administration's tariff policies towards Mexico affecting market risk appetite. Marginal weakening in February PMI and employment data were reported this week, indicating insufficient momentum in the U.S. economic fundamentals. The slow pace of U.S. inflation and the ongoing impact of Trump's tariff policies could make the process of U.S. inflation more challenging, raising uncertainties about the future of the U.S. economy. The latest export data from China also shows increased uncertainties in external demand. The market has entered a phase of fundamental validation, with the revaluation of pan-technology assets in progress. Next week, the release of domestic social financing data will mark a crucial phase for validating market expectations of fundamentals. While short-term volatility in the market may increase, the logic for the mid-term revaluation of pan-technology assets continues: 1) The U.S. dollar is expected to remain weak in the short term, with potential pressure from the relative economic advantages and potential interest rate hikes in Europe and Japan. This could create a window of opportunity for undervalued emerging market assets. 2) Expectations for weak economic recovery in China are set to continue, with the macroeconomic policies from the Two Sessions likely boosting market profit expectations. 3) From a valuation perspective, A-shares remain attractive. The current valuation percentile for the entire A-share market is around 52%, compared to approximately 79% for U.S. stocks (S&P). Given the current fundamentals of China and the trend of technological development, the revaluation of Chinese assets is expected to continue. 4) The impact of overseas risk factors on the short-term A-share market remains limited, and concerns about short-term "stagflation" in foreign markets have limited implications for domestic policies and demand. The sector outlook and industry prospects Strong recovery in technology sector, growth momentum continues. This week, driven by the annual growth rate of national defense budgets and the geopolitical situation between Russia and Ukraine, the defense and military industry saw the highest increase. The computer, media, machinery, and automotive sectors showed significant recovery compared to last week, with AI intelligence and other applications becoming new market hotspots. Recently, small metals like antimony, cobalt, and strontium have seen frequent price increases, boosting the industrial metals and non-ferrous industries. Copper and gold also made significant contributions. Dividend-related industries showed the weakest performance this week, but saw some recovery on Friday. The petroleum and petrochemical industry was significantly affected by the downward trend in international oil prices this week, while the real estate industry entered a period of policy expectations, showing some adjustment compared to the strong performance of the previous week. This week, the Siasun Robot&Automation sector of the A-share market showed significant recovery. Looking at the Siasun Robot&Automation (equal-weighted) index composed of 185 companies, the average turnover ratio for the sector (8.5%) this week has significantly decreased from the high value of 12.3% on February 26th, approaching the volume midpoint (8.2%) of the second round of the Siasun Robot&Automation sector's uptrend. In terms of margin financing balance, among the companies in the sector, there are 133 margin-financed companies, with a total financing balance of 67.8 billion yuan, representing a 146% increase from September 23, 2024, and matching the sector's interval increase of 144%. The Siasun Robot&Automation sector is currently seen more as a "spear" for A-share risk appetite. From the perspective of risk appetite observation, we believe that March is still in the "may" stage. In our in-depth review of the A-share market in 2024, we pointed out that the complexity of the A-share market and the underlying corporate earnings and economic cycle determine that the volatility of the A-share market is difficult to reduce. After each period of sideways volatility due to lack of clear incremental signals, a decision will be made at the level of the trend. Looking back at 2023 and 2024, the A-share market has experienced at least two instances of a "decision" phase after the sideways volatility, often occurring in April after the clarity of first-quarter economic data. Currently, from a timing perspective, the decision phase has not yet been reached, so we may still be in the "may" stage. Given the turbulent market trends in 2023 and 2024, the possibility of a downward decision phase recurring in the spring of 2025 is diminishing, making the strategic value of Siasun Robot&Automation and technology growth directions still important in the short term. General AI product Manus released, surpasses OpenAI in GAIA benchmark test. In the early hours of March 6th, Chinese AI startup Monica launched the world's first general AI agent product, Manus, quickly gaining attention from the tech community and capital markets. In comparison to traditional AI assistants.Different from others, Manus runs on cloud virtual machines through a multi-agent architecture, allowing it to independently access tools such as browsers and code editors to complete the entire process from task decomposition to deliverables. For example, after a user uploads 10 resumes, Manus can unpack files, analyze information, and generate ranking reports like a human HR. Technologically, Manus has achieved the most advanced level of performance in the GAIA benchmark test, surpassing similar models from OpenAI. Its technical foundation integrates deep learning, reinforcement learning, and self-supervised learning, supporting task interruption recovery and user preference memory, to provide "human-like" services.AI Agent has a high degree of intelligence and is able to independently execute complex tasks. AI Agent is an intelligent "agent" with characteristics such as autonomy, adaptability, and interactivity, with a large language model (LLM) as its core engine. Its system mainly consists of planning, memory, action, and tools. AI Agent decomposes complex tasks into manageable sub-goals through planning and optimizes the decision-making process through self-reflection and learning. It has short-term and long-term memory capabilities, utilizing short-term memory for context learning and external storage for long-term information retention and retrieval. In addition, AI Agent can call external APIs to obtain additional information, compensate for lost model weights, and execute decisions through the action module, choosing appropriate action strategies to complete tasks such as memory retrieval, reasoning, learning, and programming. In the Agent model, AI independently takes on most of the work, with users only needing to set goals and provide necessary resources. In this mode, AI embodies the interactive, autonomous, and adaptive characteristics of the intelligent agent, approaching an independent actor, while humans play more of a supervisory and evaluator role. From "brain" to "hands," Manus is an important milestone in the engineering of AI applications. At a technical level, Manus may not be a disruptive technological breakthrough, but rather an integrated innovation based on existing large models for systematic engineering optimization. Its core value lies in translating "thinking" into "action," transforming AI Agent from an auxiliary tool to an autonomous productive entity. Compared to traditional AI that only provides suggestions or answers, Manus achieves a complete task loop through a multi-model collaborative architecture: from goal decomposition, tool invocation to result delivery. Its human-like workflow execution capabilities demonstrate a key breakthrough in engineering implementation. The essence of this breakthrough lies in extending AI technology from the "brain" to the "hands." Furthermore, Manus provides reusable innovation paradigms for various industries: its open source plan and technical architecture reduce the threshold for secondary development, allowing industries such as healthcare, education, and enterprise services to build custom agents based on their own scenario needs through "basic capability reuse + vertical module customization." The launch of Manus is expected to catalyze the accelerated implementation of AI vertical applications and various AI agents, with software-side "AI + application" expected to continue to benefit. Alibaba releases open-source inference model QwQ-32B, which can be locally deployed on consumer-grade graphics cards. In the early hours of March 6, Alibaba officially released the latest open-source inference model QwQ-32B. Through large-scale reinforcement learning, QwQ-32B has improved in mathematics, code, and general capabilities, achieving overall performance on par with DeepSeek-R1 while significantly reducing deployment costs and enabling local deployment on consumer-grade graphics cards. QwQ-32B adopts the Apache 2.0 open-source license, and currently, anyone can download and use the QwQ-32B model for free or experience it through the web version of Qwen Chat. The model will also be listed on the QwQ app for free. The launch of QwQ-32B is expected to further reduce the cost for various enterprises to access AI large models, especially enhancing data security guarantees for industries with higher data security requirements. The reduction in hardware requirements is also expected to catalyze the popularization of AI edge products. AI computing power orders accelerating implementation. Recently, several A-share companies have seen their computing power orders landing. On March 4, Zhejiang Haikong Nanke Huatie Digital Intelligence & Technology announced that its wholly-owned subsidiary Zhejiang Haikong Nanke Huatie Digital Intelligence & Technology Haiden Construction Machinery Equipment Co., Ltd. reached a 5-year service agreement with a company in Hangzhou for a total of 3.69 billion yuan. The next day, the New Third Board listed company Lanyun Technology announced that its wholly-owned subsidiary Beijing Suyou Technology Co., Ltd. signed a "Computing Power Cloud Service Agreement" with Company X with a total contract amount (including tax) of 3.707 billion yuan, with a term of 60 months. Recently, various links in the AI product chain have seen frequent catalyzation, with a focus on the acceleration of orders for computing power vendors, QwQ-32B can be locally deployed on consumer-grade graphics cards, catalyzing the accelerated deployment of AI large models in vertical industries, and lowering the hardware threshold to accelerate the industrialization of edge-side AI products. The launch of Manus transforms AI from "brain" to "hands," marking an important milestone in the engineering of AI applications, and is expected to catalyze the accelerated implementation of AI vertical applications and various AI agents. With multiple catalysts, a new round of opportunities in the AI industry chain can be expected. This article is reproduced from the WeChat public account "Taipingqiao Strategy Essays," authored by Zhongyin Strategy; GMTEight Editor: Xu Wenqiang.