以下文章来源于中国金融四十人论坛 ，作者蔡昉 中国金融四十人论坛 . 聚焦金融热点，速递论坛动态，独家发布论坛课题成果，连载书系新书、好书。 文/ 中国金融四十人论坛（CF40）学术委员会主席、中国社会科学院国家高端智 库首席专家 蔡昉 当前养老保险再次成为重点关注问题。首先需要清楚， 影响社保可持续 性的三大因素为人口、劳动力市场与劳动生产率 ，分解来看，老龄化与 高龄化趋势不可逆转，劳动力市场存在结构性就业矛盾、非正规化趋势、 岗位破坏与人工智能冲击。但是，人工智能同时提升了劳动生产率的增长 潜力，未来要完善社会保障，并不缺物质财富，而是缺少制度安排。 现在养老保险再次成为重点关注问题，固然有老龄化加剧、经济增长减速、企业遭遇困难等原因，这些既是全球问题也有中国表现，尤其是中国的 老龄化、少子化导致抚养比明显上升，因此产生了" 精算忧虑 "，就是说一算账大家就预期要出现养老赤字。另外，最高人民法院的《司法解释 二》进一步强化了社保缴纳问题，其更加倾向于保护缴纳社保，更加严格地限制不缴社保的任何协议，一些人担心这会成为压倒小微企业的最后一 根稻草。关于社保问题，不只是企业，政府、老百姓、学者都反复陷入这种担 ...

Core Viewpoint - AI provides a historic opportunity for China's economy, aiding in overcoming the middle-income trap and addressing the challenges of an aging population. The transition from a policy-driven phase to a performance-driven phase is underway, indicating a significant economic transformation ahead [1][31]. Group 1: Economic Growth Paradigm - China's economy is entering a new growth paradigm, with a shift in capital market dynamics towards "innovation-efficiency" as evidenced by the rise of companies like Cambricon [2]. - The government's strategic focus on AI development has been solidified with the release of the "AI+" action plan, marking AI's elevation to a national strategy [2][5]. Group 2: Market Validation and AI Impact - Current market trends suggest that the assumptions made in AI models are being validated, with a recognition that both "dreams" and "reality" are being traded in the market [5]. - AI's penetration is expected to significantly mitigate the decline in potential economic growth rates, with projections indicating that a 20% AI penetration could sustain growth at around 5.8% by 2035, compared to a baseline of 4.6% [6][7]. Group 3: Capital Structure Transformation - The transition from "land finance" to "computing power finance" is a profound and irreversible trend, reshaping local government financial structures [10][11]. - The sustainability of this transition relies on increasing computing power utilization and the ability to generate revenue from AI-related assets [12][14]. Group 4: Addressing the Solow Paradox - AI has the potential to address the Solow Paradox, where technological advancements do not immediately translate into productivity gains. Key indicators include the ratio of AI capital expenditure and the revenue-to-cost ratio [15][16]. - A systematic measure called Elasticity of Compute-to-Output (ECO) is proposed to assess AI's impact on productivity, with a threshold of ECO greater than 0.25 indicating effective productivity enhancement [16]. Group 5: Market Valuation and Pricing Models - Traditional valuation metrics are inadequate for AI companies, which are often priced based on future earnings potential rather than current profitability [19][20]. - A more robust valuation approach involves using compute rent discount models and focusing on cash flow from AI-related revenues [20]. Group 6: Application and Commercial Viability - The most promising areas for AI applications that could achieve commercial viability include AI in financial services, industrial software, and biopharmaceuticals, with financial services expected to lead in generating cash flow [22][23]. - The criteria for identifying sectors likely to achieve commercial success include rigid demand, quantifiable ROI, and established data barriers [23]. Group 7: Strategic Outlook and Market Signals - The goal of achieving over 70% penetration of new intelligent applications by 2027 is aimed at creating a substantial domestic market for AI, fostering competition and profitability [25]. - Key signals to monitor for potential market overheating include financing ratios, regulatory attitudes, and insider selling behaviors [26][27].

Core Viewpoint - The Chinese economy is entering a new growth paradigm driven by innovation and efficiency, with artificial intelligence (AI) becoming a national strategic priority as evidenced by recent government policies and market movements [2][3]. Group 1: AI's Impact on Economic Growth - AI development is projected to significantly enhance technological progress and economies of scale, potentially stabilizing the decline in China's potential economic growth rate [2][5]. - By 2035, under a scenario where AI penetration reaches 20%, the potential growth rate could be maintained at approximately 5.8%, compared to a baseline scenario of about 4.6% [5][8]. - The current market dynamics reflect a transition from speculative "dreams" to tangible "reality," with AI's impact on productivity becoming a critical observation point [3][4]. Group 2: Market Validation and Valuation - The market is currently pricing both "dreams" and "realities," with significant capital inflow into AI companies like Cambrian, which has seen its stock price soar [3][8]. - The valuation of AI companies is shifting from traditional metrics like price-to-earnings (PE) ratios to new models based on compute rent and cash flow projections [16][17]. - Cambrian's high valuation reflects market expectations of its future cash flows rather than current profitability, indicating a speculative phase that may transition to a more sustainable growth model as productivity gains materialize [16][17]. Group 3: Structural Changes in Capital Allocation - The shift from "land finance" to "compute finance" signifies a deep transformation in China's capital structure, with local governments moving from land sales to monetizing computational power [9][10]. - This transition is seen as sustainable but requires time and policy support to fully realize its potential [9][10]. - The success of "compute finance" hinges on increasing compute usage rates and the ability to generate revenue from AI assets [10][12]. Group 4: Sector-Specific Opportunities - The most promising applications of AI that are likely to achieve commercial viability include AI in financial services, industrial software, and biopharmaceuticals, which are expected to generate solid cash flows and establish market barriers [19][22]. - The financial sector is identified as the most likely to achieve rapid commercial success, followed by manufacturing, indicating a strategic focus for investment [23][24]. Group 5: Future Outlook and Strategic Considerations - The government's goal of achieving over 70% penetration of new intelligent applications by 2027 aims to create a substantial domestic market for AI, fostering competition and profitability [24]. - Investors should monitor key signals such as market leverage, regulatory attitudes, and insider selling to gauge potential market tops in the AI sector [25][26]. - The long-term significance of AI for the Chinese economy lies in its potential to overcome structural challenges and enhance productivity, marking a critical phase in economic development [29].

Core Insights - The richness of application ecosystems is becoming a key way for large model vendors to showcase their capabilities, with domestic foundational models rapidly penetrating various scenarios [1] - The head effect of foundational models is becoming more pronounced as the "hundred model battle" shrinks, with DeepSeek, Tongyi, and Tencent's Hongyuan ranking among the top ten globally according to the Chatbot Arena [1] - Foundational models are evolving into a new digital infrastructure that can spawn numerous applications, shifting the competitive landscape from individual companies to ecosystem battles [1][2] Industry Analysis - The ability of a large model platform to attract more developers and build a vibrant application ecosystem may lead to a "winner-takes-all" scenario, raising new challenges for antitrust authorities regarding market definitions [2] - Regulatory frameworks should be cautiously defined to balance intervention and market incentives, allowing private enterprises maximum innovation space in digital infrastructure, as long as national information security and fairness for individuals and small businesses are not compromised [2] - Although the ecosystem of foundational models is expanding rapidly, the impact of AI on macro productivity may take time to manifest, reflecting the classic Solow paradox where technological advancements do not immediately translate into productivity gains [3] - AI agents are emerging as a popular application direction for large models, with predictions that 20%-30% of office tasks could be automated, potentially reallocating labor to new technology-driven fields [3]

Core Insights - The article discusses the "Solow Paradox" in relation to artificial intelligence (AI), highlighting the lack of significant productivity gains despite the widespread presence of AI technology [1] - Predictions about AI replacing jobs have been prevalent, yet the actual outcomes have not aligned with these forecasts, as seen in the case of IBM's Watson and the increasing number of radiologists in the U.S. [2][3] - The profitability of AI, particularly large language models (LLMs), is questioned, as they struggle to provide reliable answers in high-stakes applications like healthcare and law [3][4] - The current hype around AI is deemed unprecedented, with many companies not disclosing AI-related revenues, raising concerns for investors [5][6] - Overall, the AI industry's revenue is estimated to be between $30 billion and $35 billion, with growth projections that may not support the current capital expenditures in data centers [7] Group 1: AI Predictions and Reality - Bill Gates predicts that AI will replace many jobs within a decade, but historical predictions about AI have often been overly optimistic [1][2] - IBM's Watson was expected to revolutionize cancer treatment but was ultimately dismissed due to safety and accuracy issues [2] - Prominent figures in AI have made bold claims about job displacement, yet the actual job market has not reflected these predictions [2][3] Group 2: Profitability and Revenue Concerns - LLMs have limited profitability despite their capabilities, as they often generate unreliable outputs in critical fields [3][4] - Companies like Microsoft and IBM acknowledge that AI will not replace programmers in the foreseeable future, indicating a gap between AI capabilities and market needs [3][4] - The estimated revenue for leading AI startups in 2024 is projected to be under $5 billion, raising questions about the overall financial health of the AI sector [5][6] Group 3: Market Dynamics and Future Outlook - Major tech companies have not reported significant AI-related revenues, suggesting a lack of substantial business impact from AI initiatives [6] - Analysts estimate that the AI industry's total revenue could reach $210 billion by 2030, which may not justify the current capital expenditures in data centers [7] - The article draws parallels between the current AI hype and the internet bubble of the early 2000s, suggesting that a similar correction may occur in the future [7]

第三， 每一轮技术革命可进一步划分为导入期（Installation Period）和展开期（Deployment Period）。导入期又进一步包括爆发阶段与狂热阶段。在爆发阶 段，新技术初步出现并引发投资热潮，金融资本主导，技术创新与投机泡沫并存；至狂热阶段，技术应用快速扩散，但市场泡沫逐步退潮。展开期进一步包括协同 阶段与成熟阶段。协同阶段的特点是技术逐渐成熟，开始与实体经济深度融合，社会制度（如政策、法规）开始适应新技术，形成稳定的"技术-经济范式"；至成 熟阶段，技术已经全面普及，并开始相对稳定地助力经济增长 。 第四， 关于技术革命对经济增长的影响，经典的理论主要是卢卡斯和罗默的"内生增长理论"、熊彼特的"创造性破坏理论"。内生增长理论把技术进步视为内生变 量，通过人力资本积累、研发投入、知识溢出等因素推动，技术是一种可积累的公共品，会不断地作用于经济。创造性破坏理论则认为技术创新的本质是结构性变 革，是新技术对旧技术的替代，以及它带来的市场垄断权的更迭和产业结构的重组。简单来看二者区别，内生增长理论之下，技术革命是线性的、连续的；创造性 破坏理论之下，技术革命是非连续的、跳跃性的。从"创造性破坏 ...