Investment Rating - The report suggests a positive outlook for cloud service providers due to Deepseek's reduction in hardware computing power demand, indicating a new growth momentum in the short term for local deployment in large enterprises and specific industries [1] Core Insights - Deepseek aims to enhance unit computing efficiency by tenfold through algorithm optimization, significantly lowering model training and inference costs. The training cost for the 671B DeepSeek V3 is $5.576 million (approximately 40.7 million RMB), which is only 7% of Llama 3's cost, while OpenAI's ChatGPT-4o training costs range from $78 million to $100 million, requiring thousands of NVIDIA H100 chips. In contrast, DeepSeek V3 utilizes the NVIDIA H800, a specialized AI chip with reduced performance [1] - The technological revolution of Deepseek introduces a new paradigm in large model development, employing innovative architectures like MoE and MLA for efficient inference and cost-effective training. The dynamic sparse expert network design allows the model to utilize less than 4% of neural network parameters during inference, and the FP8 low-precision training framework reduces energy consumption by 80% while maintaining model stability [2] - Open-source models like DeepSeek are expected to play a crucial role in the AI era, similar to how Android transformed the mobile internet. This will reshape the industry ecosystem, accelerating the development of upper-layer applications and unifying lower-layer systems, thereby enhancing collaboration across software, hardware, and supply chains [3] Summary by Sections - **Investment Outlook**: The report highlights the potential for local domestic inference computing to explode, alongside the expansion of new foundational software like vector databases [1] - **Technological Innovations**: Deepseek's introduction of reinforcement learning-driven paradigms and self-evolving training mechanisms significantly reduces the data annotation requirements for efficient training, showcasing a systematic disruption of the "computing power arms race" [2] - **Market Impact**: By reducing reliance on high-end imported chips, Deepseek provides a viable technological path for domestic enterprises, boosting confidence in the development of self-researched computing power chips [3]

Investment Rating - The report maintains an "Overweight" rating for the industry [1] Core Insights - A new perspective on assessing physical volume rhythm through the industry chain balance sheet has been established, which helps explain and predict the rhythm of physical volume demand, indicating a potential stabilization point in year-on-year physical volume by the second half of 2025 [1][2] - The construction industry has not yet shown signs of a right-side balance sheet contraction point as of Q3 2024, but it is reasonable to infer that the debt levels are near stabilization, with major enterprises' debt ratios recovering to around 75%, suggesting that key behavioral change points in the industry are approaching [2] - The demand is highly sensitive to funding, with a focus on downstream capital injection being crucial for future demand elasticity, particularly in infrastructure financing and real estate sales [3] Summary by Sections Industry Overview - The industry is transitioning to a balance sheet contraction model, with real estate enterprises leading this trend since 2021, which has been a primary factor in the decline of physical volume in 2022-2023 [1] - Infrastructure entities are expected to start their balance sheet contraction in 2024, which will significantly impact physical volume [1] Financial Performance - The construction materials sector is anticipated to show a favorable base effect in revenue starting from Q3 2024, with leading companies in the sector expected to demonstrate improved financial performance post-2025H2 due to industry optimization and competitive advantages [3] Future Outlook - The key inflection point for the current cycle of physical volume decline is projected to occur in the second half of 2025, with potential early signs observable from the first half of 2025 [2]

核能周报|欧洲商业核能联盟成立 摘要: 产业政策及动态 产业研究中心 | ર્ | 陈磊(分析师) | | --- | --- | | ల | 021-38038037 | | ಡಿ | chenlei022459@gtjas.com | | 登记编号 | S0880522060001 | | ર્ | 王浩(分析师) | | 8 | 0755-23976068 | | ಶಿ | wanghao013539@gtjas.com | | 登记编号 | S0880513090004 | 具备颁发建造许可证的条件。(来源:生态环境部) 鲍雁辛(分析师) 0755-23976830 baoyanxin@gtjas.com 登记编号 S0880513070005 在期回顾 欣界能源获海目星、亿航智能增资,三方持续深 化产业链合作 -- 低空产业跟踪 03 期 2025.02.16 【AI产业跟踪】李飞飞团队创新的极简测试拓展 方法,极低成本构建比肩 ol-preview 的推理模型 s1-32B 2025.02.13 【智能车产业跟踪】小鹏 MONA M03 Max 版将 于二李度交付,文远知行发布无人物流车 2025 ...

债券研究 /[Table_Date] 2025.02.17 | | | 债 ——REITs 周报 本报告导读： 过去一周 REITs 指数整体上行，产权类 REITs 表现相对较好，其中仓储类和产业园 类 REITs 领涨，水利和保障房类 REITs 表现相对一般。 投资要点： | 板块补涨行情延续 | [Table_Authors] | 刘玉(分析师) | | --- | --- | --- | | | | 021-38038263 | | ——REITs 周报 | | liuyu024100@gtjas.com | | | 登记编号 | S0880523050002 | | 本报告导读： | | | | 过去一周 REITs 指数整体上行，产权类 REITs 表现相对较好，其中仓储类和产业园 | | 汤志宇(研究助理) | | 类 REITs 领涨，水利和保障房类 REITs 表现相对一般。 | | 021-38031036 | | 投资要点： | | tangzhiyu028695@gtjas.com | | | 登记编号 | S0880123070149 | [Table_Report] 相关报告 新 ...

Investment Rating - The report suggests a potential valuation recovery for the liquor sector, indicating that recent stock price movements have largely reflected market pessimism, with expectations of policy improvements supporting this outlook [1][2]. Core Insights - The liquor industry is experiencing a stable fundamental environment, with the "opening red" performance in 2025 aligning with expectations and a general post-festival stability in operations. The report anticipates a decline in the overall liquor market by less than 10%, with sales revenue expected to drop more than volume due to price declines [2]. - Recent macroeconomic policies have become more positive, with a focus on boosting consumption as a priority for domestic demand expansion. Initiatives such as the issuance of consumption vouchers are expected to stimulate related sectors, particularly dining [1]. - The report highlights that the liquor industry is currently facing high inventory levels, prompting companies to focus on inventory reduction and channel profitability. Major brands have begun controlling inventory and adjusting sales strategies to manage cash flow [2]. Summary by Sections - **Market Performance**: The liquor sector's performance during the "opening red" period in 2025 was in line with expectations, with a projected decline in market size of less than 10%. Mid to low-end liquor products are outperforming high-end counterparts [2]. - **Policy Environment**: Recent government policies are aimed at enhancing consumer spending, with specific measures to support income growth and consumption capabilities. This is expected to positively influence the liquor market [1]. - **Valuation and Stock Performance**: As of February 12, stock prices for major liquor brands have adjusted to reflect pessimistic market expectations, with valuations for key companies like Kweichow Moutai and Wuliangye at relatively low levels, suggesting potential for recovery [2].

Group 1: Economic Theory and Challenges - The "Balassa-Samuelson effect" indicates that differences in actual exchange rates and inflation levels between two countries are fundamentally determined by productivity differences in their tradable sectors, particularly manufacturing[1] - Higher productivity in the tradable sector leads to higher price levels and stronger currency purchasing power[1] - Despite significant productivity improvements in China, the country faces low price levels and exchange rate stability pressures, indicating a disconnect from classical theory[1] Group 2: Wage Transmission Issues - The transmission of productivity gains to wages, especially in the service sector, is crucial for the "Balassa-Samuelson effect" to hold[2] - Labor compensation in East Asian countries, including China, is generally lower than in Western countries, hindering the transmission of productivity to wages[2] - Since 2017, the growth rate of labor compensation in China has slowed, with a decline noted in 2021, despite significant productivity gains in manufacturing[2] Group 3: Fiscal Policy Implications - The allocation of fiscal funds is more important than the scale of spending; effective transmission of "productivity-wage-inflation" is essential for the "Balassa-Samuelson effect" to be realized[3] - Current market expectations for increased fiscal deficits and stimulus may reflect traditional investment-focused fiscal thinking, which may not yield the desired economic impact[3] Group 4: Risk Considerations - There are discrepancies between international and historical experiences compared to China's unique circumstances, which may affect the applicability of the findings[4] - Inconsistencies in statistical definitions across international comparative data may also pose challenges to the analysis[4]

Investment Rating - The report recommends investment in DeepSeek R1, highlighting its algorithm optimization that reduces performance requirements for computing chips, thus accelerating the adaptation of domestic computing power [1]. Core Insights - DeepSeek R1 is an open-source, low-cost solution that addresses the core bottleneck of advanced processes, facilitating the growth of domestic computing power. It offers excellent cost-performance ratio, matching the performance of o1 models and is expected to lead the domestic AI ecosystem [2]. - The demand for inference drives the expansion of total computing power, with ASIC chips taking on more inference tasks. DeepSeek R1's high inference performance is expected to promote the prosperity of inference chip clusters and AI applications, with long-term investments in AI infrastructure shifting towards inference computing [2]. - The performance of the distilled small model in DeepSeek R1 is impressive, accelerating the deployment of edge AI and significantly enhancing hardware experience, thus opening up growth opportunities for edge hardware [3].

Investment Rating - The report provides a positive investment outlook for the AI industry, particularly focusing on the advancements in domestic AI infrastructure and chip technology [1]. Core Insights - DeepSeek R1 optimizes algorithms to reduce the performance requirements of computing chips, facilitating the adaptation to domestic computing capabilities. This shift is driven by increasing restrictions on overseas computing resources, making self-sufficient computing a necessary choice. The development of advanced domestic processes is expected to gain momentum [1]. - The AI infrastructure is shifting towards inference computing, where ASIC chips offer a higher cost-performance ratio in inference scenarios. Efficient and low-cost ASIC chips are anticipated to benefit from this trend [1]. - The R1 distilled small model demonstrates strong performance, accelerating the implementation of edge AI, thereby opening growth opportunities for edge hardware [1].

Investment Rating - The report does not explicitly provide an investment rating for the industry Core Insights - The long-term trend of gold pricing has shifted to being influenced by the RMB exchange rate, with current Shanghai gold serving as a good tool for hedging against exchange rate expectations [2][5] - The relationship between CPI and Shanghai gold premium is expected to reconnect once the level of early mortgage repayment returns to low levels, indicating a potential for gold to maintain its cost-effectiveness [2] - The central bank's resumption of gold purchases signals a stabilization of the exchange rate, but the probability of a significant gold market rally is low, with expectations leaning towards market fluctuations [3] - The actual interest rates of US Treasury bonds are identified as the main short-term pricing factor for gold, with recent fluctuations attributed to economic policy uncertainties related to the US elections [4] Summary by Sections Latest Views - The report highlights that the Shanghai gold premium is a reflection of market expectations regarding the exchange rate and is crucial for tracking cross-border capital flows [2] - The high level of swap premiums necessitates cautious judgment regarding future Shanghai gold premiums, suggesting that holding Shanghai gold is a favorable hedging strategy [2] Central Bank Gold Purchases - The resumption of gold purchases by the central bank is interpreted as a move to stabilize the exchange rate amidst a backdrop of diverging currency trends in Asia and the US Federal Reserve entering a rate-cutting phase [3] - The future trajectory of gold prices will depend significantly on whether the economic cycles of the US and other major economies move in sync or diverge [3] Recent Currency and Gold Fluctuations - The report notes that the actual interest rates of US Treasury bonds have been a key factor in the short-term fluctuations of gold prices, with notable divergences occurring during significant geopolitical events [4] - The current low level of Shanghai gold premiums suggests a higher certainty in Shanghai gold compared to London gold [4] Underlying Logic of Recent Gold Surge - The report indicates that the monetary attributes of gold are no longer the primary pricing factors, with the central bank's gold purchases and gold ETF inflows being the main driving forces behind the recent gold market dynamics [6]

Investment Rating - The report maintains an "Overweight" rating for the industry, indicating a positive outlook on the growth potential of smart driving penetration rates [1]. Core Insights - The penetration rate of smart driving is expected to accelerate, particularly with the ongoing advancements from leading companies like BYD. The launch of BYD's "Tianshen Eye C" solution, which includes advanced hardware and features, is a significant development in this regard [1]. - The cost of smart driving technology is decreasing, and the price range for smart driving vehicles is also lowering, which is likely to sustain the growth in penetration rates. Technological advancements, such as DeepSeek, are expected to reduce the reliance on computing power and data, further lowering the entry barriers for smart driving solutions [2]. Summary by Sections - **Smart Driving Solutions**: BYD has introduced the Tianshen Eye C solution, featuring a three-camera setup and multiple radar systems, enabling high-speed NOA and intelligent parking capabilities. The solution aims for a 99% success rate in parking scenarios and is expected to be fully developed by the end of 2025 [1]. - **Vehicle Integration**: BYD plans to equip its entire lineup with the Tianshen Eye system, with 21 models set to launch, enhancing the accessibility of advanced smart driving features across various price segments [1]. - **Technological Advancements**: The integration of DeepSeek technology into BYD's architecture is anticipated to enhance AI capabilities in vehicles, promoting a dual-cycle of AI development between vehicle and cloud systems [2].