【导读】源乐晟举办2026年度策略会，曾晓洁、杨建海最新发声 近日，老牌私募源乐晟举办2026年度策略会，源乐晟资产创始人曾晓洁与源乐晟合伙人杨建海分享了对 2026年市场的展望及行业观察。 曾晓洁与杨建海均认为，目前无需担忧"AI泡沫"，2026年或成为AI行业标志性的一年。当前消费板块较 难出现大贝塔行情，但新消费领域结构性机会持续存在。化工行业当前估值处于低位，产品价格具备反 弹基础，化工板块具备较强的性价比。 整体而言，2026年有望走出"慢牛"行情，但"慢牛"不代表没有波动。在风险方面，地缘政治风险、美联 储降息滞后等因素仍需警惕。 消费行业仍有结构性机会 化工板块具有较强的性价比 "在新消费领域，每年都会有结构性的机会。"杨建海表示，如餐饮中的一些细分赛道，在新兴业态下的 翻台率大幅突破过往。新消费领域不乏这类结构性机会，但这类机会的天花板相较以往更低，相关标的 市值达到一定程度后，其发展的持续性会引发担忧。 他认为，白酒行业那种长期持有"躺平"的逻辑已不再适用于新消费投资，而是需要每年在不同SKU中筛 选出当期锐度最强的方向，同时紧密跟踪标的变化、动态评估。 在杨建海看来，当前消费板块较难出现大 ...

Core Viewpoint - UBS's 2026 U.S. Economic Outlook report indicates that the seemingly robust growth of the U.S. economy is heavily reliant on the AI sector, and a potential AI bubble burst could raise the probability of a recession to 50% [1] Economic Growth and Risks - The foundation of U.S. economic growth is extremely narrow, with AI-related equipment investment increasing by approximately 17% over the past four quarters, while non-AI equipment investment has declined by about 1% [3] - Non-residential construction investment has contracted for six consecutive quarters, and residential investment has decreased in four out of the last five quarters [3] - Without the contribution from AI, the U.S. economy is essentially stagnating [3] - Consumer spending shows a K-shaped divergence, with high-income groups benefiting from AI-driven stock market wealth effects, while middle and low-income groups face erosion of purchasing power due to inflation and tariffs [3] Labor Market Conditions - The labor market's weakness is more pronounced than surface data suggests, with non-farm employment declining by an average of 41,000 per month in the last four months of 2025 when excluding healthcare and social assistance sectors [3] - The broader U-6 unemployment rate has risen to 8.43%, significantly above pre-pandemic levels [3] - There are indications that current employment data may overestimate job growth by about 60,000 per month, suggesting that labor market weakness poses a risk to the economy [3] Monetary Policy Insights - UBS believes that the market has mispriced the timing of the Federal Reserve's interest rate cuts, with current expectations suggesting a delay until June [4] - Factors such as actual labor market weakness and potential soft signals from the February non-farm report indicate a higher necessity for rate cuts before June than the market anticipates [4] - UBS forecasts two rate cuts of 25 basis points each in 2026, bringing the federal funds rate target range down to 3.00%-3.25% [4] Market Divergence - There is a clear divide among market participants regarding UBS's outlook, with optimistic views suggesting a 2.7% GDP growth in 2026 and a 36% year-on-year increase in AI-related capital expenditures [5] - Some institutions agree with UBS's assessment of economic fragility, highlighting the risks associated with over-reliance on AI and the hidden weakness in the labor market [5] Tariff Policy Implications - The current weighted average tariff rate in the U.S. has risen to 13.2%, equivalent to a tax burden of 1.1% of GDP, which is gradually being passed on to consumers and pushing core PCE inflation higher [6] - Inflation is expected to peak in the summer of 2026 and remain around 3% in the long term, which limits the Federal Reserve's ability to cut rates [6] - Fiscal policy may provide short-term support, but the impact of tax rebates in Q2 2026 is expected to diminish quickly [6] - Key variables influencing the U.S. economic trajectory and Federal Reserve policy include the conversion of AI investment into productivity gains, the visibility of labor market weakness, the extent of tariff-induced inflation, and the Fed's balancing act between inflation and employment goals [6]

当世界目光聚焦于瑞士的冰雪小镇，一场关乎未来命运的对话正在激烈上演。2026年1月19日至1月23 日，以"对话的精神"为主题的世界经济论坛于达沃斯召开，汇聚了来自130多个国家的近3000名政商领 袖。 在这场全球顶级思想盛宴中，人工智能毫无意外地占据了舞台重要一角。 论坛之上，西方科技巨头的认知呈现鲜明撕裂：一面是以微软、谷歌掌舵人为代表的"技术乌托邦"信 念，另一面则是如Anthropic、Palantir等掌门人对失业潮与社会动荡的严峻预警。 上述高层级的观点对立，凸显了AI革命进程中的巨大不确定性。然而，更关键的破局变量已然出现 ——中国AI力量的加速崛起，正在成为改写全球游戏规则的关键因子。 AI是泡沫还是未来引擎？ 世界经济论坛上，一个根本性问题被反复提及：当前如火如荼的AI浪潮，究竟是一场坚实的革命，还 是一个危险的泡沫？海外科技领袖们就此展开了激烈交锋。 悲观与警示的声音不容忽视。OpenAI董事会主席Bret Taylor直指，AI"很可能"是泡沫，大量"聪明 钱"与"盲目资金"正涌入技术栈的每一层，催生泡沫膨胀。亚马逊CEO安迪·贾西同样警告存在泡沫风 险，他指出模型商与基础设施提供商之 ...

国泰君安期货·君研海外 境外权益（港美股）周度策略报告 国泰君安期货研究所· 海外研究 戴璐 Z0021475 联系人 杨藤 F03151619 日期：2026年1月25日 Guotai Junan Futures all rights reserved, please do not reprint 美股：本周地缘风波引发美股波动，能源板块领涨 本周地缘风波引发美股波动，随后TACO（标普500和VIX均呈现V字形反转） 地缘风波助推下，本周美股能源板块领涨 u 上周六特朗普威胁要对支持格陵兰的8个欧洲国家加征关税，随即引发美股在周二触发年内的最大单日调整。而周三特朗普又宣布取消加 征关税引发"TACO交易"，市场随即回补周内跌幅。地缘政治风险溢价扩大，能源板块本周领涨。 资料来源：彭博，国泰君安期货研究 Special report on Guotai Junan Futures 美股：下周迎来"超级周" ，市场主线回归宏观与财报预期差 标普500成分股：年初至今涨幅前12的标的（存储和半导体设备为代表的硬件龙头显著领涨全市场） | 序号 代码 | 名称 | 涨跌幅 | 年初至今▼ | 时间 | 成交额 | ...

Core Viewpoint - The article discusses the unprecedented rise in memory prices, particularly DDR5 memory, and its implications for various industries, including consumer electronics and automotive sectors. The surge in prices is attributed to the increasing demand driven by AI developments and the strategic decisions of major memory manufacturers. Group 1: Memory Price Surge - The price of a single 256GB DDR5 memory module has skyrocketed to 42,000 yuan, making it comparable to gold bars in value [5][7] - The prices of other computer components, such as solid-state drives and graphics cards, have also increased significantly, with a 1TB SSD now costing around 1,200 yuan, up from 300 yuan two years ago [10][11] - The rising costs are severely impacting consumers looking to build PCs, as well as manufacturers of smartphones and laptops, who have raised prices by approximately 500 yuan for new models [12][13] Group 2: Impact on Automotive Industry - The automotive sector is also feeling the pressure from rising memory costs, with estimates suggesting that the average cost per vehicle may increase by 1,000 to 3,000 yuan due to higher memory prices [20] - Major automotive companies, including NIO and Xiaomi, have acknowledged that the cost pressures from memory are significant, affecting their pricing strategies [16][18] Group 3: Supply Chain Dynamics - OpenAI has signed substantial contracts for graphics cards, leading to a projected demand for 4-5 million units from NVIDIA and 240,000 from AMD, which has strained the supply chain [54][55] - Major memory manufacturers, including Samsung and SK Hynix, have shifted their production focus from consumer-grade DRAM to server-grade HBM memory, exacerbating the supply shortage for consumer memory products [58][61] - Approximately 60-70% of consumer-grade DRAM capacity is now directed towards AI server memory, leading to significant price increases for consumer memory [63] Group 4: Market Behavior and Future Outlook - Despite the rising prices, major manufacturers are hesitant to expand production capacity significantly, with Samsung and Hynix planning only modest increases of 4.5% and 8.5%, respectively [69][70] - The cautious approach to capacity expansion is attributed to the potential for market oversupply in the future, reflecting a broader trend of greed within the memory industry [80] - The article suggests that the current memory price surge may persist until at least 2030, driven by ongoing AI demand, although some analysts warn of a potential bubble that could burst [25][26][110] Group 5: AI and Energy Concerns - The development of AI is expected to lead to a massive increase in energy consumption, with predictions indicating that global data centers could consume 945 TWh by 2030, significantly impacting energy resources [114][115] - The article raises concerns about the sustainability of AI growth, highlighting the potential for energy shortages and increased electricity costs as data centers expand [127][128] - The interplay between AI advancements and energy demands is framed as a critical issue that could shape the future of both industries and society at large [130][131]

Summary of Key Points from Conference Call Industry Overview - The conference discusses the global AI investment theme, highlighting significant growth from late 2023 to late 2025, with the A-share AI theme index rising approximately 1.2 times, the Hang Seng Tech Index increasing over 46%, and the market capitalization of the seven major US tech companies growing about 174% to reach $21.6 trillion [1][2]. Core Insights and Arguments - **Rational vs. Irrational Bubbles**: - Rational bubbles are driven by investor expectations of future price increases, supported by loose liquidity and technological advancements. In contrast, irrational bubbles are influenced by psychological biases and herd behavior, detaching from fundamental valuations [1][4]. - **Comparison with 2000 Internet Bubble**: - The current AI market shares similarities with the 2000 internet bubble, such as technological benefits and fundamental improvements. However, the macroeconomic environment is more complex, necessitating caution regarding liquidity changes and market sentiment [1][5]. - **Valuation Concerns**: - Current AI-related assets exhibit signs of overvaluation, with the Shiller P/E ratio indicating the Nasdaq is in a dangerous zone. The Tobin Q ratio is high, and the investment in information processing and software as a percentage of GDP is nearing levels seen during the 2000 bubble [16][20]. - **Liquidity and Market Dynamics**: - The decline in long-term risk-free rates in the US has created a favorable environment for tech stock valuation expansion, attracting capital inflows and boosting the dollar index [7][8]. - **Risk Indicators**: - Key risk indicators include the Buffett Indicator exceeding historical highs, equity risk premiums falling into negative territory, and total stock market capitalization compared to financial data being close to 2000 levels [13][16]. Additional Important Content - **Investment Trends**: - By 2025, approximately 50% of global AI industry data investment will come from large enterprises' own funds, contrasting with the 54% from venture capital in 2000 [20]. - **Employment Impact**: - AI technology enhances efficiency but also compresses certain job roles, necessitating attention to changes in employment structure to assess whether the AI bubble could evolve into a detrimental bubble [21][22]. - **Market Sentiment and Future Outlook**: - The AI investment landscape is expected to remain robust, with significant opportunities in 2026 and 2027, particularly in core asset pricing models. The US market, led by companies like Nvidia, shows strong growth potential [23][24]. - **China's Market Dynamics**: - The Chinese mainland market, especially the ChiNext board, mirrors the US market's trends, indicating a dual rise in valuation and profitability, with no immediate downturn expected [24]. - **Technology Cycle Theory**: - The current phase of technology investment may be in the mature stage, with potential for a peak following widespread adoption. Evaluating tech companies should consider their technological barriers and commercialization paths to identify resilient assets [25][26].

Group 1 - Elon Musk highlighted that the development of AI technology in the U.S. is constrained by insufficient power supply, while China has already addressed this issue [1][3] - Musk warned that the U.S. may soon face a situation where there is an oversupply of chips that cannot be utilized due to power shortages [3] - The production of AI chips is experiencing exponential growth, but the lack of power supply is affecting the efficiency of AI data centers in training and deploying large models [3] Group 2 - Musk pointed out the significant differences in energy supply structures between the U.S. and China, noting that China has a clear lead in power capacity, especially in solar energy [3] - The aging U.S. power grid and decades of underinvestment in infrastructure are increasingly problematic, slowing down the advancement of AI applications and raising concerns about an "AI bubble" in the U.S. [3] - President Trump mentioned at the Davos forum that his administration encourages tech companies to build nuclear power plants, with related projects expected to receive government approval within three weeks [4]

1. Report Industry Investment Rating No information provided in the content. 2. Core Viewpoints - The return variance of funds has widened, and convertible bond funds have outperformed active products. The scale of public funds' convertible bond holdings decreased by 2.6% to 30.83 billion yuan, less than the overall market scale decline of -7%. Most convertible bond funds achieved positive returns, and there was an obvious trend of funds flowing from passive to active products [1][13]. - In Q4 2025, funds mainly increased positions in bank, military, photovoltaic industry targets and semiconductor new - issue bonds, and reduced positions in battery convertible bonds. Convertible bond funds led active - type funds, and the performance variance of active equity - oriented funds was extremely large [2]. - Among the high - performing products, Fund A, which ranked first in returns, adopted a quantitative strategy with outstanding asset - switching ability. Fund B, which ranked second, had excellent bond - selection ability, heavily invested in securities and technology, and significantly outperformed the index during the rising stage [3]. 3. Summary According to the Directory 3.1 Fund Holding Convertible Bond Scale and Fund Type Distribution - The convertible bond holding scale of first - and second - tier bond funds and flexible allocation funds slightly increased, while that of partial - debt hybrid and convertible bond funds decreased. The total asset value of convertible bond funds decreased from 67.85 billion yuan at the end of Q3 to 65.459 billion yuan [12]. - In Q4 2025, affected by multiple factors, the market mostly oscillated until late December. The scale of the convertible bond market continued to shrink, and the supply - demand contradiction was prominent. The convertible bond holding scale of public funds decreased by 830 million yuan to 30.83 billion yuan, with a decline of 2.6%, less than the overall market scale decline [13]. - The number of funds with a decreased convertible bond position in Q4 2025 was more than those with an increased position, and the ratio of adding - position funds to reducing - position funds was 0.74. The number of funds with a convertible bond position of more than 5% decreased significantly compared with Q2 and Q3 2025 [15][17]. - By fund type, first - tier bond funds, second - tier bond funds, convertible bond funds, partial - debt hybrid funds, and flexible allocation funds were the main forces in convertible bond allocation. More first - tier bond funds began to allocate convertible bonds in Q4 2025, and the scale of second - tier bond funds expanded most significantly [20][25]. - In terms of the price of convertible bonds held by public funds in Q4 2025, the proportion of balanced convertible bonds slightly increased, the position of bank convertible bonds remained stable, and the proportion of high - priced convertible bonds slightly decreased [27]. 3.2 2025 Q4 All Types of Fund Return Statistics - In Q4 2025, convertible bond funds led active - type funds, and the performance variance of active equity - oriented funds was extremely large. The average quarterly returns of ordinary stock funds and partial - equity hybrid funds were - 1.59% and - 1.94% respectively, with performance variances of 6.72% and 7.5% respectively. The average quarterly return of convertible bond funds was 0.86% [2][43]. - Most convertible bond funds achieved positive returns, and the trend of funds flowing from passive to active products was obvious. The median return of 41 convertible bond funds (including convertible bond funds, Xingquan Convertible Bond, and two ETFs) in Q4 was + 1.09%, and the return of convertible bond ETFs was 1.45%. The median return of these 41 products in the past year was 1.21% [46]. - Among the high - performing products, Fund A, which ranked first in returns, adopted a quantitative strategy, with the proportion of convertible bonds in the fund's total assets increasing significantly in Q4. It had outstanding asset - switching ability. Fund B, which ranked second, had excellent bond - selection ability, heavily invested in securities and technology, and significantly outperformed the index during the rising stage [3][51][62].

纳德拉表示，AI必须走出科技公司的范畴，实现更广泛的渗透。否则，过去数年提振股市、催生历史 性估值的AI热潮，从定义上来说必然会成为一场泡沫。"从定义而言，要避免形成泡沫，就必须让AI带 来的收益更均衡地惠及各方。" 为平息对AI泡沫的进一步担忧，纳德拉表示，基于过去20年云计算和移动设备的广泛普及，他对AI技 术的快速扩散充满信心。非科技领域的企业对AI更广泛的应用将推动经济增长，这种增长必须由企业 运用AI提升自身营收来驱动，而非仅靠科技公司投资AI基础设施的资本支出，资本支出驱动的经济增 长"正是我们当前所见的现象"。 在企业管理方面，大型企业是否会因AI高昂的开发成本而具备先天优势？纳德拉表示，初创公司能从 零开始构建AI体系，而传统巨头虽在数据、规模和客户关系方面保有优势，却受困于变革管理难 题，"没有任何一方能安于现状。" 此外，纳德拉表示，欧洲若要在AI时代取得成功，就需要有更广阔的全球视野。"欧洲的竞争力在于其 产出的全球竞争力，而不仅限于欧洲内部。"他认为，欧洲经济在过去300年里繁荣是因为能够生产出世 界所需的产品，要想再次做到这一点，就需要投资能够驱动AI的能源与token。 卡内基梅隆 ...

Group 1 - The core argument presented by Jensen Huang, CEO of Nvidia, is that the current investment in AI is not a bubble but rather the largest infrastructure build-up in human history, requiring trillions of dollars in investment [1][5] - The AI industry is experiencing a significant transformation, moving from a "light asset software industry" to a "heavy asset heavy industry," with substantial investments in hardware and infrastructure [7][8] - The semiconductor market is witnessing a surge, with major companies like Intel and Micron seeing significant stock price increases, indicating strong demand for AI-related hardware [7][8] Group 2 - In 2025, the AI sector is projected to see investments of approximately $1.5 trillion, surpassing expenditures in nearly all other fields [2] - Despite the massive investments, a report from MIT indicates that 95% of organizations in the U.S. have not seen returns on their investments in generative AI, raising concerns about the effectiveness of these expenditures [3] - The AI industry is structured like a "five-layer cake," with each layer requiring substantial investment to ensure the functionality of the layers above it, emphasizing the need for foundational infrastructure [4][5] Group 3 - The global semiconductor market is experiencing unprecedented growth, with companies like TSMC and Micron investing heavily in new factories to meet the rising demand for AI capabilities [7][8] - The price of high-end memory chips has skyrocketed, with some products seeing price increases of over 600%, driven by limited production capacity and growing demand for AI servers [8] - Shenzhen-based company Demingli has projected a net profit increase of 85.4% to 128.2% for 2025, reflecting the strong growth potential in the storage solutions sector [9]