Goldman Sachs, in its latest series of reports, posits that the potential economic benefits from Chinese AI are significantly underappreciated by the market and are forming an investment narrative distinct from global technology stocks.

According to the report, since the "DeepSeek moment" in January 2025, Chinese AI stocks have surged by an average of 50%, adding over $3 trillion to the total market capitalization of related tech stocks. Analysis suggests current valuations still underestimate the potential value creation from AI by 50% to 100%, indicating the structural opportunity is far from over.

From an allocation perspective, global fund managers' holdings of Chinese AI stocks constitute only 1.2% of their global technology portfolios. This is substantially lower than China's approximate 10% share of the global AI market by capitalization and 16% by revenue contribution. Within the global competitive landscape, China possesses significant comparative advantages in power, infrastructure, and physical AI, making it an increasingly important source of diversification for global AI investment portfolios.

At the policy level, AI aligns with China's core objective of technological self-reliance. Profit growth for related stocks is expected to significantly outpace that of non-AI assets. Against a backdrop of attractive valuations, Goldman Sachs believes allocating to AI has become a necessary strategy to hedge against the risk of disruption to the real economy and traditional industries.

**DeepSeek Ignites the Fuse: AI Transforms the Game in the Past Year** Goldman Sachs states that since the release of DeepSeek R1 in January 2025, AI application in China has entered a phase of substantive acceleration. A significant drop in model inference costs has accelerated technological adoption, rapidly establishing AI as the dominant theme in Chinese equity markets. During this period, Chinese AI-related stocks rose an average of 50%, 103 AI companies completed IPOs in Hong Kong and mainland China, and the tech sector added over $3 trillion in total market cap, with AI stocks contributing approximately $3.4 trillion of this increase.

Concurrently, Chinese model capabilities have solidified their standing in global competition. Large language models from companies like DeepSeek, Alibaba, and ByteDance have ranked at the top in multiple global benchmark tests, positioning China as a globally competitive AI model exporter. The latest survey from Goldman Sachs' Asia Pacific Global Macro Conference revealed that 68% of attending investors identified AI as the top investment theme for 2026, far ahead of consumer sectors, global travel, and dividend strategies.

**Redefining the Chinese AI Value Chain: A $10 Trillion Ecosystem** In response to the rapidly evolving technological landscape, Goldman Sachs has redefined the investment universe for Chinese AI stocks using a triple-methodology cross-verification approach.

First, based on a global AI supply chain mapping, Goldman Sachs categorized over 700 leading AI companies from the US, North Asia, and Europe into 29 industries across five thematic layers—Power, Semiconductors, Infrastructure, Models, and Applications. Chinese listed companies with AI revenue linkages were then mapped to corresponding industries, constructing a global AI stock universe covering 3,715 companies with a total market capitalization of $36 trillion, equivalent to 25% of the global total market cap.

Building on this, Goldman Sachs introduced a revenue classification framework based on 2024 financial data to segment each company's revenue streams, objectively quantifying their true AI exposure. The results show that over 3,000 Chinese listed companies have traceable AI revenue linkages, with a total market cap of approximately $10 trillion. About half of this market capitalization is directly related to the AI value chain. By industry distribution, the software sector has an AI revenue exposure of 84%, electronic manufacturing stands at 60%, while sectors like medical services and mining are below 5%.

Finally, through bottom-up industry insights, Goldman Sachs' analyst team has identified 29 AI-enabling and AI-enabled industries. They project that by 2035, the total addressable market for Chinese AI companies could expand to $16 trillion.

**Four Investment Implications from Chinese AI** Goldman Sachs notes a significant structural mismatch for Chinese AI stocks. China accounts for 10% of the global AI market cap, contributes 16% of related revenue, and holds nearly a 20% share in capital expenditure and R&D investment. However, global mutual funds allocate only 1.7% to Chinese stocks overall, with holdings of Chinese AI tech stocks constituting just 1.2% of their global tech allocation. This substantial gap suggests considerable potential for capital inflows if global investors begin correcting this allocation bias.

From a global value chain perspective, the US and China each possess distinct advantages. The US dominates in semiconductors, AI models, and digital applications, whereas China's comparative advantages are concentrated in three areas: Power, Infrastructure, and Physical AI, which account for 38%, 26%, and 27% of the global AI revenue pool, respectively. Holding Chinese AI stocks can provide global investors with differentiated exposure and effective diversification benefits.

Heterogeneity in returns is also increasing. Since the DeepSeek release in January 2025, Chinese AI stocks have outperformed their US counterparts by 30%, while North Asian AI stocks have outperformed by 21%. More importantly, the 52-week rolling return correlation between Chinese AI stocks and US/global tech stocks is only 23%, significantly lower than the 69% correlation between the US and other regions, indicating that Chinese AI has formed an investment theme independent of the US narrative.

Sector rotation is accelerating. From a global perspective, AI leadership is diffusing from semiconductors towards power and infrastructure, reflecting a market focus shift from computing power construction to supply bottlenecks. In China, the infrastructure sector has demonstrated strong performance throughout both the ChatGPT and DeepSeek market cycles, highlighting the competitive advantage of the local market in technology hardware manufacturing.

**Valuation Reassessment: Why Chinese AI is Not a Bubble** Goldman Sachs explicitly states that Chinese AI stocks are far from entering bubble territory; current valuations may be underestimating the potential value addition and profit creation from AI by 50% to 100%. This core judgment is based on calculations across macroeconomic, total addressable market, and corporate profit dimensions, all pointing to the same conclusion: the economic value generated by AI far exceeds the gains reflected in current market capitalizations.

At the macro level, generative AI is projected to deliver a cumulative 8% boost to Chinese labor productivity over a decade, equating to approximately $1.6 trillion in current economic value added. In present value terms, the total economic benefit from AI could reach $6 to $7 trillion, with about $3 trillion potentially accruing to Chinese companies as capital income.

At the industry level, by 2035, Chinese companies' revenue share in 21 specific AI-related global industries could expand to $16 trillion. Assuming a 15% net profit margin and discounting at a 10% cost of equity, the present value of this potential profit pool is approximately $2.4 trillion.

At the corporate profit level, widespread AI adoption could boost Chinese corporate profits by 3 percentage points annually over the next decade through cost savings and new market opportunities. For the entire listed sector, this implies an incremental profit contribution of about 6 percentage points, with a present value of approximately $800 billion.

Goldman Sachs argues that compared to these potential profit and value increments, the net increase in Chinese AI market capitalization since the "DeepSeek moment" appears relatively modest, further supporting the view that valuations are not stretched and the AI theme retains upside potential.

**The Cost of Missing Out: Four Risks of Not Investing in AI** Goldman Sachs' latest report indicates that in the current era where AI is reshaping global industrial structures, failing to invest in Chinese AI itself constitutes a risk that requires careful assessment.

The first risk is misallocation. China accounts for 10% of global AI market cap, 16% of revenue, and nearly 20% of R&D investment, yet global funds' holdings of Chinese AI represent only 1.2% of their tech allocations. This systematic underweight suggests that if global capital begins correcting this偏差, the cost of missing the trend could far exceed the risks associated with valuation.

The second risk is structural misinterpretation. Markets often benchmark Chinese AI against the US narrative, but their respective areas of strength differ significantly. China commands 26% to 38% of the global revenue pool in Power, Infrastructure, and Physical AI. Applying a Silicon Valley stock-picking logic to Chinese companies risks missing the true local strengths.

The third risk is growth divergence. Since the DeepSeek release, Chinese AI stocks have collectively outperformed US peers by 30%, but returns within the sector are highly differentiated. Infrastructure has remained strong through both market cycles, while the application layer has lagged due to unclear monetization paths. Simply betting on the "AI concept" is no longer effective.

The fourth risk is valuation lag. Goldman Sachs estimates that the potential economic value from AI-driven efficiency gains and new profit creation is 50% to 100% higher than the expectations reflected in current stock prices. Over the next decade, profit growth for Chinese AI companies is projected to outpace non-AI peers by 140 percentage points, yet the market's pricing of this growth differential remains conservative.

**How to Position in Chinese AI?** Goldman Sachs systematically evaluated the risk-return profiles of various thematic layers within Chinese AI by combining TAM forecasts, Residual Income Models, and fundamental industry data. The report indicates that the market-implied growth expectations for the Power and Infrastructure sectors are relatively conservative, with model-implied EPS CAGR around 3% and 11% respectively. In contrast, Goldman's TAM-based growth forecasts are significantly higher at 23% and 31%, suggesting substantial undervaluation.

Conversely, the Application layer—particularly in consumer services, healthcare, and autonomous driving—already embeds high growth expectations, resulting in relatively narrower valuation safety margins. The AI Models sector currently trades at a P/E of only 17x, but recognition is gradually improving as new IPOs come to market. The Semiconductor sector maintains a solid global position but offers a relatively moderate growth trajectory.

For hedging against AI disruption risks, Goldman Sachs recommends focusing on industries with the following characteristics: high tangible asset ratios, clear AI revenue exposure, intensive R&D and capital expenditure, and high barriers to entry—such as sectors with significant state-owned enterprise presence. Based on a composite ranking, sub-sectors like wafer foundry, semiconductor equipment, and optical modules lead across these dimensions.