Chinese stocks face rotation temptation: A-share hard tech remains preferred, but H-share internet can begin to be accumulated on weakness

Overview

This report is Goldman Sachs' China equity strategy publication, “China Musings,” focusing on the topics currently receiving the most attention from investors: internal rotation within the Chinese market, the AI theme, divergence between A- and H-share performance, internet earnings pressure, IPO supply, foreign investor positioning, and sources of returns in 2H26. The report believes macro factors have temporarily receded in importance, with market pricing increasingly driven by AI and micro factors at the company and industry levels.

Core views

The report's central views are: first, continue to tactically favor A-shares over H-shares and hard tech over soft tech, as A-share hard tech is driven by themes such as AI infrastructure and domestic substitution, with stronger earnings momentum and investor attention; second, H-share internet/soft tech has already experienced significant derating, and if subsidy losses narrow, new opportunities such as cloud and agentic AI grow, and core e-commerce businesses maintain stable cash flow, the 2Q or 3Q earnings season could become a catalyst for valuation rerating; third, Chinese AI stocks are not broadly in a bubble, as the increase in market capitalization remains modest relative to potential AI-driven economic gains, although localized overheating exists in semiconductors and certain A-share hard-tech names, requiring control of concentration and valuation risks; fourth, China remains an important component of the global AI stock universe, but international investors remain underpositioned, particularly in power, infrastructure, and Physical AI.

Analysis framework

The report forms its strategy conclusions by combining investor meeting feedback, A-H market rotation models, hard-tech versus soft-tech relative-return models, earnings revision and valuation indicators, fund-flow and prime-brokerage positioning data, historical IPO-return research, comparisons between AI-related market capitalization and potential economic value, and thematic portfolio frameworks.

Methodology notes

• market_rotationA-H Market Rotation Model / Hard vs. Soft Tech Rotation Model

  Assessing rotation between A-shares and H-shares and between hard tech and soft tech

  The model uses fundamental, valuation, and liquidity factors to assess the possibility that H-share soft tech will improve relative to A-share hard tech over the coming months.

• earnings_sensitivityChinese Big Tech Profit Growth Sensitivity

  Internet leaders' share prices are more sensitive to actual profit growth

  The report believes that low valuations alone are insufficient to drive a sustained rebound; a durable rally requires confirmation from an underlying recovery in profits.

• theme_portfolioChina Select AI Portfolio / GS 15FYP Portfolio

  Thematic portfolios centered on AI and policy support

  The China Select AI Portfolio covers 50 companies across 22 AI subsectors and aims to provide balanced exposure across China's AI value chain; the GS 15FYP Portfolio represents high-conviction themes related to China's 15th Five-Year Plan.

Asset mapping & comparison

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

• Chinese A-shares

  Tactical overweight

  Strengths

  Better earnings momentum, prominent AI hard-tech and policy-support themes, and diversification value for international investors.

  Weaknesses

  Valuation premiums, concentration, and leverage risks have risen in some hard-tech and semiconductor segments.

  Comparison

  The report continues to prefer them over H-shares.

  Risks

  Localized overheating in AI trading, valuation pullbacks, and liquidity crowding.

• Chinese H-share internet/soft tech

  Begin gradual allocation

  Strengths

  Significant derating since the beginning of the year; some pessimistic assumptions about AI investment returns may already be reflected in prices, while future earnings recovery could trigger a rerating.

  Weaknesses

  Weak 1Q26 earnings, with internet subsidy losses and pressure from AI capital expenditures still significant.

  Comparison

  Underperformed A-share hard tech in the short term but may narrow the gap over the coming months.

  Risks

  Earnings recovery later than expected, uncertainty regarding AI investment returns, and continued subsidy competition.

• Chinese AI stocks

  High-conviction theme requiring diversification

  Strengths

  Potential efficiency gains and incremental profits/TAM may exceed what is discounted in current share prices; China holds an important position in global AI market capitalization, revenue, and token usage.

  Weaknesses

  International investors remain both underpositioned and underinformed, while valuations in some subsectors are already elevated.

  Comparison

  Compared with the Korean and Taiwanese AI chains, China's AI exposure is more diversified, with some links under-owned.

  Risks

  Localized overheating in semiconductors and A-share hard tech, concentration risk, leverage risk, and weaker-than-expected AI commercialization.

• Hong Kong IPO market

  Alpha opportunity

  Strengths

  Strong short-term post-listing returns for new stocks year to date; issuance size, standalone H-share listings, and moderate cornerstone ownership explain much of the performance dispersion.

  Weaknesses

  An active and lengthy IPO pipeline has raised concerns about liquidity diversion.

  Comparison

  Compared with the broad-market beta environment, IPOs have become a popular channel for investors seeking alpha.

  Risks

  Higher-than-expected follow-on issuance, declining market liquidity, and mean reversion in new-stock returns.

Key data

• STAR50 relative to HSTECH68 percentage points ahead year to dateThe report treats the STAR50 as an onshore AI hard-tech proxy, while the HSTECH is primarily composed of offshore internet and platform companies.
• ChiNext relative to SHCOMP/CSI30019 percentage points/17 percentage points ahead year to date, respectivelyThis shows that return dispersion within Chinese stocks is at an historically extreme level.
• Recent HSTECH performanceRebounded 11% over the past two weeks but remains down 14% year to dateThe rebound was driven by improved news flow for some internet leaders, low valuations, and progress in AI applications and monetization.
• MSCI China 1Q26 profitsDown 8% year over yearThe internet sector accounts for approximately 35% of earnings weight and was the primary source of drag.
• Internet subsidy lossesMore than RMB180 billion cumulatively since 2Q25Narrowing subsidy losses is an important condition for future earnings improvement.
• China's global share of AI exposureApproximately 11% of AI-related market capitalization and 18% of revenueThe report considers China an important component of the global AI stock universe.
• Token usage of Chinese AI modelsClose to half of the global totalBased on the top ten model companies by usage on OpenRouter.
• Hong Kong IPO fundraising year to date100 companies, with a combined US$35 billion raisedMedian one-month and three-month post-listing returns were 32% and 30%, respectively.
• Estimated new issuance for the remainder of the yearApproximately US$25 billion in IPOs and approximately US$45 billion including follow-on offeringsThe report believes concerns about liquidity drainage are overstated relative to the scale of dividends and buybacks by listed companies.
• Listed-company cash returnsApproximately US$500 billion in FY2025 and US$560 billion in FY2026Includes dividends and buybacks, which are important offsets to new issuance supply.

Impact & implications

The investment implication is that Chinese stock returns in 2H26 are more likely to come from structural themes and alpha than from broad beta. Portfolios can continue to retain exposure to A-share hard tech and policy-supported themes while beginning to monitor and gradually allocate to large-cap H-share internet companies where valuation derating is largely complete and earnings-recovery visibility is improving. Outside AI, materials, capital goods, insurance, shipbuilding, new consumption, healthcare/biotech, the property chain, brokers, and the hog cycle are also viewed as potential sources of alpha.

Risks

• If earnings recovery among H-share internet companies is delayed, low valuations may be insufficient to support a sustained rebound.
• Popular A-share hard-tech, semiconductor, and AI-related trades face localized overheating, valuation, and concentration risks.
• Large AI capital expenditures by internet companies, along with uncertainty over negative NPV or ROI, may continue to suppress valuations.
• If IPO and refinancing supply exceeds expectations, concerns over Hong Kong market liquidity could intensify.
• Although macro factors have temporarily receded in importance, housing, consumption, the renminbi, exports, oil-price shocks, and policy responses could again influence market pricing.
• Although China-US relations and geopolitical risks are priced at relatively low levels by the market, sudden changes remain possible.

What to watch

• The 2Q26 and 3Q26 earnings seasons for Chinese internet leaders, particularly whether subsidy losses narrow.
• Revenue and profit contributions from new AI opportunities such as cloud, agentic AI, and AI tokens.
• Whether the performance gap between the HSTECH and A-share hard tech continues to narrow.
• Changes in valuation, trading concentration, and leverage risks in A-share semiconductors and hard tech.
• The scale of Hong Kong IPO and refinancing supply, and the extent to which dividends and buybacks offset liquidity effects.
• Changes in net/gross China equity exposure among EM funds and global hedge funds.
• Policy signals following the July Politburo meeting, as well as trends in housing, consumption, exports, and the renminbi.