全球互联网_人工智能模型的商业化框架和市场细分_18页_942kb

Global Internet: AI Model Commoditisation and Market Segmentation Summary

Core Content

This document provides an analysis of AI model commoditisation and its implications for global market segmentation and competition. The focus is on how different end-use scenarios will influence the adoption and pricing of AI models, as well as the role of Chinese AI labs in the global market.

Main Points

• AI Model Commoditisation Framework: The authors suggest that AI model commoditisation will be driven by human perception of task completion and the reliability of models at scale, rather than by the convergence of underlying intelligence. As models become “good enough” for specific tasks, competition shifts from performance to cost and availability.

• Market Segmentation: The market is expected to split into two tiers:

  • Frontier AI: High-cost, high-performance models used for complex and mission-critical tasks such as frontier science and advanced engineering.
  • Trailing Edge AI: Lower-cost models that will dominate consumer and enterprise applications once they are deemed “good enough” for routine tasks.

• Competition Dynamics: US and European enterprises are likely to avoid Chinese models due to data security and geopolitical concerns, leaving them to be used by non-developed market enterprises and SMEs. Chinese labs, on the other hand, will benefit from their lower token costs and potentially capture 35-40% of the global AI TAM.

• R&D Economics: As AI models become more commoditised, the focus of AI labs will shift from broad R&D to more specific and complex tasks. This could lead to more efficient use of R&D budgets and better operating leverage for AI labs.

• Token Cost and ROI: The high cost of Claude Fable 5 tokens has led developers to reconsider AI usage, highlighting the importance of aligning token costs with the marginal benefit of task completion. This is expected to drive the adoption of cheaper, “good enough” models for simpler tasks.

Key Information

AI Commoditisation Timeline

• Consumer-focused AI (e.g., ordering bubble tea, booking hotel rooms) will commoditise quickly.
• Enterprise use cases (e.g., Excel work, coding) will take longer to become “good enough”.
• Frontier science (e.g., drug discovery, nuclear fusion) will likely remain a niche market with high willingness to pay.

Market Segmentation

• US and Europe: Will primarily use US-based SOTA models due to data security and geopolitical concerns.
• Rest of the World (ex. China): Will see significant adoption of Chinese models due to their lower costs and broader accessibility.
• China: Will serve a large portion of the global AI TAM, especially in consumer and enterprise markets, due to cost advantages.

Valuation and Market Multiples

• Tencent (700.HK): Rated Outperform, with a price target of HKD 780, and a projected 2026E P/E of 13.5x.
• Alibaba (BABA, 9988.HK): Rated Outperform, with a price target of USD 180, and a projected 2026E P/E of 28.9x.
• Other Companies: A range of valuation multiples and price targets are provided, reflecting different market segments and AI use cases.

Implications for AI Labs

• R&D Focus: As simpler tasks become commoditised, AI labs will need to focus on more complex and high-value use cases.
• Operating Leverage: Lower token costs and higher efficiency in model training could lead to improved operating leverage and profitability for Chinese AI labs.

Investment Implications

• Token Cost Alignment: AI users will select models based on the marginal cost of tokens relative to the marginal benefit of task completion.
• Market Share Shifts: Trailing edge AI models, especially from Chinese labs, will likely capture significant market share in consumer and enterprise applications due to cost efficiency.
• Geopolitical Constraints: US and European enterprises may avoid Chinese models, limiting their market access but allowing Chinese labs to expand in other regions.

Conclusion

The document outlines a framework for understanding AI model commoditisation and market segmentation, emphasizing the role of user perception, cost, and reliability in shaping the future of AI adoption. It suggests that while US labs will continue to lead in high-value, complex tasks, Chinese labs will gain traction in more accessible and cost-sensitive markets. The analysis also highlights the potential for AI labs to shift focus and achieve better economics as the market evolves.