AI Models Will Become Tiered: Chinese Labs Expected to Capture 35–40% of the Global Market

Overview

This report presents a new analytical framework for the commercialization and market segmentation of artificial intelligence (AI) models. Bernstein argues that the competitive landscape of AI models will no longer be determined solely by technological convergence but will depend on how human users perceive 'task completion' across different application scenarios. Once a particular application—such as consumer services—is 'solved' by AI, the focus of competition will shift from reasoning power to price, reliability, and developer trust. Based on this, the report predicts that the global AI market will bifurcate into two tiers: U.S. frontier labs serving high-premium, specialized domains, while Chinese AI labs, leveraging their cost advantages, are expected to capture 35–40% of the global market in broader general-purpose and enterprise applications. The report maintains a positive rating for Tencent and Alibaba.

Core views

The driving force behind AI model commercialization lies in user perception rather than pure technical metrics. The report notes that as AI performance improves, when task completion in specific verticals becomes reliable and scalable—for example, ordering bubble tea or booking flights—the marginal returns from investing further massive R&D funds in those areas will decline sharply. This implies that AI labs will reallocate resources to more complex, cutting-edge tasks. Consequently, consumer-oriented AI applications will be the first to achieve commercialization, followed by enterprise workloads with higher certainty—such as Excel processing—and finally, frontier scientific fields like drug discovery and nuclear fusion. The global AI market will exhibit a clear two-tiered division. The first tier comprises U.S. frontier labs—such as OpenAI and Anthropic—which, by continually unlocking new model capabilities, serve price-insensitive, highly specialized customer segments and enjoy pricing premiums. The second tier is the 'lagging fringe' AI market, primarily catering to routine enterprise and consumer needs. In this layer, due to geopolitical and data security concerns, some large European and American companies may shun Chinese models, but in other regions and among non-sovereign or small-to-medium enterprises, the 'good enough' yet significantly lower-priced reasoning capabilities offered by Chinese AI labs will be highly attractive. The market opportunities for Chinese AI labs have been underestimated. Despite geopolitical constraints, the report estimates that Chinese AI labs can still tap into roughly $320 billion of the global AI revenue TAM, accounting for 35–40% of the total ($1 trillion). This share comes mainly from markets outside the U.S., as well as from price-sensitive SMEs and individual users. Chinese labs benefit from using the world's state-of-the-art (SOTA) technologies as R&D beacons, reducing exploratory research costs, and leveraging lower developer salaries and hardware efficiency. An optimistic outlook on the economics of AI labs. If the aforementioned path to commercialization holds, growth in AI industry R&D spending may slow, as the range of use cases requiring exponential capital investment gradually narrows. This will help AI labs demonstrate operating leverage. For Chinese internet giants like Tencent and Alibaba, their deployments at the AI application layer—such as WeChat smart agents and Tongyi Qianwen—will benefit from this trend, prompting the report to maintain a positive rating and elevated price targets for both companies.

Analysis framework

The report employs a non-traditional analytical framework based on 'user perception' and 'task completion,' diverging from conventional convergence theories grounded purely in technical indicators. The firm first draws attention to ROI by observing recent shifts in AI pricing strategies—for instance, Claude Fable 5's high price limiting user adoption—and then infers that users will select models of varying costs depending on the marginal utility of each task. Next, the report segments the global AI market by geography (U.S., Europe, other regions) and customer type (government, large enterprises, startups/individuals), factoring in geopolitical and data security constraints to qualitatively assess the accessibility of each segment. Finally, by summing up the sizes of accessible markets, it quantifies the potential TAM share for Chinese AI labs and, combined with R&D cost structures, analyzes their impact on long-term profitability.

Methodology notes

• Industry/Industrial Analysis FrameworkSupply-demand framework

  Analysis of AI Model Commercialization and Segmented Market Supply and Demand

  Rather than treating AI as a single, homogeneous market, the report divides it into tiers based on 'task complexity' and 'willingness to pay.' This segmentation helps identify genuine demand differences across regions and customer groups, enabling a more accurate assessment of Chinese AI vendors' potential supply capacity in non-U.S. markets.

• Competition and Strategy FrameworkMoat / competitive advantage

  Competitive Barriers Based on Cost and 'Good Enough' Performance

  The report points out that, in general-purpose AI tasks, competitive advantage will shift from 'absolute intelligence level' to 'per-task cost' and 'reliability.' By adopting a follower strategy to reduce R&D exploration costs and offering 'good enough' services at lower prices, Chinese AI labs establish competitive barriers in price-sensitive markets.

• Valuation MethodPE/PEG valuation

  Relative Valuation Based on Price-to-Earnings Ratio (PE)

  When valuing Tencent and Alibaba, the report uses forward-looking P/E multiples—such as Tencent's FY+1 PE of 20—to determine target prices. This is a standard approach for assessing mature internet technology companies, reflecting the market's pricing of their earnings-growth expectations.

• Valuation MethodSOTP Segmental Valuation

  Sum-of-the-Parts (SOTP) Valuation

  For diversified companies like Alibaba, the report applies SOTP, separately valuing the future revenues and profits of its core e-commerce and cloud businesses before summing them up. This method more clearly reflects the independent value of each business segment, avoiding distortions caused by a single valuation metric.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Tencent Holdings (700.HK)

  Beneficiary; its deployments in consumer applications such as WeChat smart agents will be among the first to realize the commercialization and scaled rollout of AI tasks

  Strengths

  Vast user base, strong social ecosystem integration, rich AI application deployment scenarios

  Weaknesses

  None

  Comparison

  Compared to pure model providers, Tencent has more direct consumer-facing monetization channels and stronger user stickiness

  Risks

  Macroeconomic risks, fluctuations in user engagement, gaming and advertising competition, antitrust regulatory risks

• Alibaba Group (BABA/9988.HK)

  Beneficiary; the promotion of Tongyi Qianwen in enterprise and consumer applications, along with infrastructure support from its cloud business

  Strengths

  Leading cloud computing infrastructure, abundant enterprise-level application scenarios, influence of its open-source model ecosystem

  Weaknesses

  None

  Comparison

  Alibaba Cloud provides the computational foundation for AI inference, while e-commerce scenarios offer extensive data feedback loops

  Risks

  Macroeconomic risks, fluctuations in platform user engagement, intensifying competition, regulatory pressures, losses in innovative businesses

Key data

• Proportion of Global TAM Accessible to Chinese AI Labs35–40%Corresponds to roughly $320 billion in potential revenue, even accounting for geopolitical restrictions and challenges accessing the U.S. market
• Total Addressable Market (TAM) for Global AIApproximately $1 trillionLong-term proxy estimate of AI revenue TAM
• Tencent Target PriceHKD 780Based on a FY+1 PE of 20, implying significant upside from current levels
• Alibaba Target PriceUSD 180 / HKD 176Derived from SOTP valuation of its core e-commerce and cloud businesses
• Expected R&D Spending Growth Rate for China's Top AI Labs50% CAGR (five years)R&D expenditure will continue to rise rapidly in the short term but should decelerate over the long term as the scope of solved tasks narrows

Impact & implications

For investors, this framework suggests focusing not only on the absolute technical parameters of AI models but also on their implementation costs and reliability in specific business contexts. Although Chinese internet giants—such as Tencent and Alibaba—may lag slightly behind top U.S. labs in cutting-edge foundational models, they possess immense commercialization potential and cost advantages in the vast 'lagging fringe' market and consumer application scenarios. This supports their long-term profitability and valuation recovery thesis. Meanwhile, the global AI profit pool will be redistributed, with U.S. labs occupying high-end niche markets and Chinese labs capturing share in large-scale general-purpose markets through volume and price competitiveness.

Risks

• Macroeconomic risks (e.g., credit tightening, weak retail consumption)
• Fluctuations in user engagement
• Competition from rivals in platforms, gaming, and advertising demand
• Regulatory risks (especially China's antitrust laws)
• Geopolitical and data security concerns limiting Chinese models' penetration into large overseas enterprises
• Losses in innovative businesses and other divisions

What to watch

• How AI users choose models based on the marginal cost of different tasks ('rational return to token-maximization behavior')
• The pace at which Chinese AI labs expand their market share in non-U.S. markets
• Whether AI R&D spending growth slows as the scope of solved tasks narrows
• Progress in the commercial deployment of Tencent's WeChat smart agents and Alibaba's Tongyi Qianwen