Recently, the emergence of DeepSeek has taken both the technology and asset management industries by storm. Within days, DeepSeek climbed to the top of the free app rankings in both the U.S. and Chinese Apple App Stores, while tech giants like Amazon and Microsoft announced integrations with its model.
The rapid rise of DeepSeek validates a key observation from the ARK CIO Office: as we enter a highly volatile and uncertain 2025, global Chinese wealth managers will face two key open-ended challenges—geopolitical and economic uncertainties, symbolized by the new U.S. administration, and the revolutionary impact of AI, digital assets, and other disruptive technologies.
So, what investment advice would DeepSeek offer? What is its outlook on AI sector investments? And does it align with the strategies and trends highlighted by the ARK CIO Office?
With these questions in mind, Noah Holdings (HKEX: 6686, NYSE: NOAH), through its global brand ARK, conducted a special interview with DeepSeek, asking the following three questions:
How should a $10 million portfolio be allocated?
What are the key investment opportunities in AI-related industries in 2025?
Does DeepSeek’s rise fundamentally alter AI market investment logic?
Let’s explore how this AI phenomenon responded.
Disclaimer
The following responses are AI-generated and have not been reviewed by professional advisors or regulatory authorities. They do not consider specific investor profiles and should be used for informational purposes only. These responses should not be taken as financial advice. Investing involves risks, and past performance does not guarantee future results.
Question 1: If You Had $10 Million, How Would You Invest It?
If you were an experienced wealth advisor, how would you allocate a $10 million portfolio for 2025? My key requirements are as follows:
Aim for the highest possible returns while keeping overall drawdown within 10%.
Diversify across industries to ensure a well-balanced allocation.
Include international assets.
Carefully assess the risk-return profile of each asset class.
DeepSeek’s Response:
Based on your requirements—high returns, controlled drawdowns, sector diversification, and including international assets—here is a customized 2025 asset allocation strategy, referencing the latest market analysis and institutional strategies:
I.Strategic Asset Allocation Framework
Asset Class Allocation Investment Rationale & Objectives
Equities 60% Core return driver, diversified across A-shares and global markets, balancing sector and style risks
Fixed Income (including Convertible Bonds) 25% Stable cash flow, smoothing volatility, convertible bonds offer equity-debt hybrid opportunities
Alternative Assets 10% Gold as a hedge against geopolitical risks, REITs benefiting from rate cuts
Cash & Money Market Funds 5% Liquidity reserve to capture market opportunities
II.Sector-Specific Investment Strategy
1.Equities (60%)
1)A-Shares (40%)
Sector allocation across five key themes:
Consumer Upgrade (20%): Premium consumption & staples, benefiting from domestic recovery and
defensive characteristics.
Tech Growth (25%): Semiconductors, renewable energy, and AI-driven fintech, boosted by policy support
and domestic substitution.
Healthcare & Life Science (15%): Medical devices & CXOs, driven by aging demographics and biotech
innovation.
Dividend Stocks (20%): High-yield banks & coal leaders, offering attractive yields in a low-rate environment.
Cyclical/Manufacturing (20%): Advanced manufacturing & carbon neutral transition, benefiting from
economic recovery and supply-side reforms.
Investment Vehicles:
Index ETFs (40%): Large-cap blue chips & high-dividend ETFs.
Active Funds (30%): Balanced growth and consumer sector leaders.
Direct Stock Picks (30%): Selection of 10 industry leaders, each with ≤5% weighting.
2)Overseas Equities (20%)
Hong Kong Stocks (10%): Leading internet and consumer service firms, benefiting from attractive
valuations and Southbound capital inflows.
U.S. Stocks (6%): S&P 500 ETF (VOO), Nasdaq 100 ETF (QQQ), focusing on AI hardware and applications (e.g., Nvidia).
Europe/Japan (4%): European healthcare ETFs and Nikkei 225 ETFs, supported by low valuations and accommodative monetary policies.
2.Fixed Income+ (25%)
Government Bonds (30%): 10-year Treasury ETFs for defensive positioning, benefiting from potential rate cuts.
Corporate Bonds (40%): High-grade corporate bond funds with stable coupon income.
Convertible Bonds (30%): Low premium options, balancing equity and debt exposure for rebound
opportunities.
3.Alternative Assets (10%)
Gold (5%): Hedge against geopolitical risks and inflation surprises.
REITs (5%): Domestic public REITs, benefiting from lower interest rates and policy support.
4.Cash & Liquidity Management (5%)
Money Market Funds: Maintain liquidity reserves to facilitate portfolio rebalancing and capitalize on market opportunities.
III.Risk Management & Tactical Adjustments
1.Drawdown Control
If equity drawdown exceeds 15% within a calendar year, reallocate 10% of equity exposure to fixed
income+ to mitigate downside risk.
Implement index futures for tactical hedging against extreme market volatility (requires professional
execution).
2.Rebalancing Framework
Quarterly portfolio reviews, with rebalancing triggered if asset class allocations deviate beyond ±5% of target weights.
3.Scenario-Based Adjustments
Stronger-than-expected economic recovery: Increase exposure to cyclical sectors (e.g., industrials,
construction machinery) and high-dividend Hong Kong equities.
Heightened geopolitical tensions: Increase gold allocation to 10% as a risk hedge while reducing equity
exposure.
IV.Expected Returns & Risk Assessment
Target Annualized Return: 8–12%, with equities serving as the primary return driver and fixed income+
providing income stability and downside protection.
Maximum Drawdown: Managed within 10% through strategic diversification and dynamic risk hedging.
Key Risks: Potential disruptions from U.S. tariff policies under a Trump administration, slower-than-
expected domestic economic recovery, and valuation corrections in U.S. equities.
V.Additional Considerations
Global Diversification: Gain exposure to emerging markets (India, Southeast Asia) through QDII funds or
Hong Kong-recognized mutual funds to enhance portfolio resilience.
Tax Efficiency: Optimize after-tax returns by leveraging preferential dividend tax policies (e.g.,
exemptions for holdings exceeding one year).
Institutional-Grade Portfolio Management: Engage leading private equity firms or establish a family office
to implement sophisticated investment strategies and dynamic asset allocation.
The above portfolio strategy employs a 'core-satellite' approach to optimize risk-adjusted returns while
ensuring sector diversification and global exposure. Regular portfolio reviews and dynamic adjustments in
response to policy and market shifts are recommended to achieve sustainable long-term capital
appreciation.
ARK CIO Commentary
The ARK CIO Office underscores the importance of adopting a multi-dimensional strategic framework that extends beyond the conventional focus on safety, liquidity, and returns. DeepSeek’s emphasis on portfolio diversification and dynamic risk management closely aligns with this philosophy.
Moreover, DeepSeek’s recommendation to engage family offices or top-tier private equity firms for professional portfolio management and strategic asset allocation highlights the critical role of institutional wealth management in optimizing risk-adjusted returns and ensuring long-term capital preservation.
Question 2: Where Are the Key AI Investment Opportunities in 2025?
DeepSeek’s response highlights a diverse range of AI investment opportunities in 2025, spanning infrastructure, hardware innovation, application layers, and vertical industry adoption. Below is a breakdown of the key areas and their strategic implications.
I.AI Infrastructure & Computational Demand
1.Compute Chips & Hardware
The explosion of generative AI is fueling demand for cloud and edge computing power, benefiting AI servers, GPUs (e.g., Nvidia), DPUs, and advanced packaging technologies.Leading tech giants such as Microsoft and ByteDance continue to ramp up capital expenditures in AI. Microsoft alone is expected to allocate $80 billion in AI-related investments in its 2025 fiscal year, further driving demand across the supply chain, including optical modules (e.g., InnoLight, Zhongji Innolight) and thermal management solutions.
Domestic AI compute alternatives are accelerating, with Hygon, SMIC, and other Chinese firms making notable advancements in chip design and manufacturing.
2.Data Centers & Network Architecture
Microsoft and other cloud providers' transition to CXL architecture could boost growth in high-speed connectivity players like Credo, while the rising penetration of AI servers will drive demand for PCBs, switches, and networking components.
II.AI Hardware Innovation at the Edge
1.AI Glasses & Wearable Devices
AI-powered smart glasses are projected to surpass 5 million units in sales by 2025, featuring functionalities such as real-time translation and automated meeting transcription.
Qualcomm’s AR1 chip and suppliers like Goertek (hardware assembly) stand to benefit, alongside innovative brands showcased at CES, such as Thunderbird and Liviko.
2.AI PCs & Smartphones
The adoption of AI-enhanced PCs is accelerating, while smartphones integrating on-device AI models (e.g., Apple's and Xiaomi’s localized AI applications) are driving demand for advanced chipsets.
Supply chain beneficiaries include Luxshare Precision and Lens Technology.
III.Generative AI & the Rise of the Application Layer
1.Content Creation & Interactive AI
PGC/UGC tools: AI-driven content generation is unlocking new opportunities in video production (Runway, HeyGen) and AI-assisted writing (Wanderboat for travel planning), with potential billion-dollar enterprises emerging in niche segments.
Enterprise AI Agents: The evolution from Copilot-style assistance to fully autonomous AI agents is accelerating. Notable applications include AI customer support (GigaML), AI sales agents (GTM Agent), and AI-powered enterprise search tools (Glean).
2.Media & Entertainment
AI-powered content production in film, gaming, and digital marketing is unlocking commercial monetization opportunities.
Social commerce innovations (e.g., WeChat-based viral marketing) are driving revenue growth in AI-assisted book publishing and e-commerce marketing.
IV.Industry-Specific AI Adoption
1.Healthcare & Regulatory Compliance
AI adoption is addressing labor shortages in healthcare, particularly in automated diagnostics and administrative workflow optimization.
High-growth verticals include AI-driven legal compliance tools (Harvey) and AI-powered research search engines (Consensus).
2.Education & Financial Services
AI language tutors (e.g., Speak) and personalized learning platforms are becoming more mainstream.
AI-enhanced risk assessment and client management solutions are seeing growing adoption in financial services.
V.Humanoid Robots & Autonomous Driving
1.Humanoid Robotics
The convergence of AI, advanced manufacturing, and new materials is making humanoid robots the next consumer electronics frontier.
Key beneficiaries include upstream component suppliers (sensors, actuators) and midstream assemblers (e.g., Tesla, UBTech Robotics).
2.Autonomous Driving
Advanced driver assistance systems (ADAS) are becoming standard in new energy vehicles, driving demand for vehicle cameras, LiDAR sensors, and autonomous driving SoCs.
BYD Electronics and other supply chain players stand to benefit.
VI.The Evolution of AI Search & Information Distribution
The Fragmentation of Search Markets
The dominance of traditional search engines (e.g., Google) is being disrupted by AI-driven vertical search engines (e.g., Perplexity, Consensus), which redefine information retrieval through semantic understanding and knowledge synthesis.
VII.Investment Logic & Risk Considerations
1.Focus on ROI & Commercial Viability
The AI investment landscape is shifting from theme-driven speculation to revenue realization, requiring investors to prioritize companies with:
Strong technological leadership (high R&D spending).
Clear business models (e.g., performance-based SaaS monetization).
2.Domestic AI Substitution & Policy Support
China's AI compute ecosystem (chips, servers) and humanoid robotics sector remain key policy-driven opportunities with strong growth potential.
3.Risk Considerations
Slower-than-expected AI compute demand growth.
Technology obsolescence risks due to rapid iteration.
Intensifying industry competition.
Conclusion
AI investments in 2025 require a dual focus on infrastructure robustness and application-driven innovation. Key areas of interest include compute infrastructure, edge AI hardware, generative AI tools, industry-specific AI solutions, and humanoid robotics. Investors should adopt a dynamic allocation strategy, balancing technological differentiation and commercialization readiness.
ARK CIO Commentary
The ARK CIO Office notes that global institutional investors are accelerating their allocations to AI infrastructure and digital assets, while Chinese wealth managers continue to favor risk-averse assets such as government bonds and insurance products.
The ARK CIO Office highlights the need to balance short-term defensive positioning with long-term technological upside—a trend reflected in DeepSeek’s recommendation to invest in AI compute hardware (e.g., Nvidia, Hygon) and generative AI applications (e.g., Runway, AI-driven healthcare). Additionally, digital assets should be considered as part of an alternative investment strategy.
Question 3: DeepSeek’s Impact: Has It Reshaped AI Investment Logic?
DeepSeek’s rapid emergence has significantly influenced investment logic in the AI market, with profound implications across several key areas.
I. Challenging the "Compute Supremacy" Paradigm and Rethinking Hardware Demand
DeepSeek has disrupted the long-held belief that high-performance AI models require expensive high-end chips by leveraging innovative training methodologies, such as reinforcement learning and model architecture optimization. With a development cost of less than $6 million, DeepSeek successfully trained a model comparable to ChatGPT-4o.
This breakthrough has directly impacted leading chipmakers like Nvidia, whose stock plummeted 16.97% in a single day, erasing nearly $600 billion in market capitalization—an all-time record for U.S. equities.
The market has begun reassessing the long-term demand for high-end AI chips and questioning the necessity of continued large-scale hardware investments.
II. Questioning the Justification for Big Tech’s Capital Expenditures
In 2024, Microsoft, Meta, Google, and Amazon collectively invested $218 billion in AI data centers, with plans to increase spending to $300 billion in 2025.
However, DeepSeek’s low-cost model development and open-source strategy have significantly lowered barriers to AI model access, prompting investors to scrutinize the sustainability of big tech’s massive capital expenditures.
The market is increasingly challenging the notion that AI dominance is exclusive to tech giants and is exploring more capital-efficient investment pathways.
III. Shifting Investment Focus from Hardware to Software and Applications
Goldman Sachs and other institutions note that DeepSeek’s rise signals a shift in AI investment priorities, moving from hardware infrastructure to software and applications.
The open-source model and cost-efficient training framework have enabled small and mid-sized enterprises to develop AI-driven applications more efficiently, accelerating adoption in robotics, consumer electronics, and autonomous vehicles.
Stock market trends further validate this transition—application-focused companies (Apple, Google, Meta) have outperformed hardware manufacturers, reflecting a capital rotation into AI software, edge computing, and data services.
IV. Repricing of Chinese Tech Assets and Narrowing the Global Expectation Gap
DeepSeek’s success has reshaped global perceptions of China’s AI innovation capabilities, narrowing the expectation gap between Chinese and U.S. technology ecosystems.
Chinese AI stocks—particularly AI application-layer companies—are experiencing valuation recovery and growth momentum.
The MSCI China Index has benefited from its higher exposure to AI software firms, while asset managers like abrdn now recognize that China’s tech giants have the capability to develop top-tier AI models.
With low valuations and improving earnings, Chinese equities are becoming increasingly attractive to global investors.
V. Jevons Paradox and Accelerated AI Adoption
DeepSeek’s cost-efficiency breakthrough may trigger Jevons Paradox, where increased efficiency leads to a surge in overall demand.
By significantly lowering the cost of AI model development, DeepSeek has expanded AI accessibility, accelerating AI integration across healthcare, education, and industrial automation.
Over the long term, this could drive increased compute demand, particularly for AI inference workloads.
Huatai Securities predicts that while short-term AI training demand may decline, the expansion of AI applications will sustain overall demand for compute resources.
Conclusion and Market Debate
Optimists believe that DeepSeek has accelerated AI democratization, shifting industry dynamics from hardware-driven investment to application-driven monetization, ultimately benefiting the entire AI value chain.
Cautious investors argue that hardware advancements (e.g., next-gen chip iterations, data center expansions) remain foundational to AI’s long-term trajectory, and that short-term market fluctuations do not negate the necessity of hardware investment.
Geopolitical risks persist: The U.S. may escalate AI industry protectionist measures in response to rising competition from alternative AI development models, potentially reshaping global AI supply chain dynamics.
Final Takeaway
DeepSeek’s emergence is not only reshaping AI investment logic but also driving a broader reassessment of global technology asset valuations.
Investors must balance technological innovation, commercial adoption, and regulatory risks while identifying opportunities at the intersection of hardware efficiency and software-driven AI enablement.
