Goldman Sachs' 10 Signals Reveal China AI Hardware's Sentiment Shift and Future Outlook

Woofun AI reports that Goldman Sachs released a comprehensive China strategy report on August 9, detailing a reassessment of market sentiment within the country's AI hardware sector. The analysis dissects ten distinct signals to determine whether recent corrections represent short-term profit-taking or deeper structural shifts in trading dynamics.

The first half of the year witnessed rapid gains for Chinese AI stocks, with hardware sectors averaging a 33% increase.

However, as global trading in AI hardware reversed, major indices including the Science and Innovation 50, ChiNext, and CSI 1000 have corrected by more than 20% from their recent highs. This divergence prompts a critical evaluation of whether the adjustment cycle is nearing completion or if further downside risk persists.

Signal one highlights a significant return divergence and style reversal between hardware and soft-tech assets. During the first six months, the return gap between the Science and Innovation 50 and the Hang Seng Tech Index exceeded 100 percentage points, mirroring the extreme dispersion seen at the peak of market divergence in early 2021. Small-cap, high-growth, and momentum styles drove these gains until July, when a rapid reversal occurred. As global AI hardware trading cooled, alpha from hardware, small-cap, and momentum factors declined sharply, while the performance gap between soft-tech and hardware stocks reverted to near long-term averages. Goldman Sachs notes that the period of fastest pullbacks and style reversals may be easing, indicating that the most crowded sectors have experienced significant cooling, though complete liquidation has not yet occurred.

Signal two addresses market breadth and the concentration of gains among top performers. Estimates indicate that 90% of the Science and Innovation 50's gains this year originated from the top 10 best-performing stocks. In contrast, this concentration was 23% at the beginning of 2021 and 46% in mid-2015 for ChiNext. Trading volume remains similarly concentrated, with the technology sector, ChiNext, and the Science and Innovation 50 accounting for 26%, 14%, and 6% of A-share cash trading volume, respectively—levels that are relatively high in recent years. Despite low correlation among individual A-share stock returns, which suggests a focus on niche themes like AI rather than macro-driven moves, the trading structure remains highly concentrated. AI continues to be the primary pricing factor, with neither returns nor volume truly diversifying.

Signal three evaluates valuation metrics, specifically forward P/E and PEG ratios relative to historical norms. At the June high, the market-cap-weighted forward P/E ratio of ChiNext stood at around 26 times, while the median forward P/E for Science and Innovation 50 stocks was approximately 50 times. Following price declines, valuation multiples for A-share hardware companies have dropped to near or below long-term averages. When adjusted for earnings growth, forward PEG ratios are also at or below historical averages, suggesting that some overheating has been addressed.

However, absolute valuations for the Science and Innovation 50 remain elevated. The valuation discount of Chinese AI stocks relative to overseas counterparts has narrowed significantly after global adjustments, reducing the previous safety margin of "Chinese assets being cheaper." Consequently, the market is expected to rely more on actual earnings realization rather than continued valuation expansion.

Signal four examines leverage trends, financing balances, and structural concentration risks. Goldman Sachs' data shows that A-share financing balances have decreased from around 3 trillion yuan to about 2.6 trillion yuan, with the proportion of free-floating market cap falling from 6.0% to 5.5%. This indicates that some leveraged funds have withdrawn, yet both metrics remain above historical norms. Compared to the more pronounced deleveraging observed in South Korea and Taiwan China recently, A-share deleveraging may still be in its early stages.

More critically, the distribution of leverage is highly skewed: the top 10% of stocks with the highest financing balances account for about 30% of all A-share financing balances, a new all-time high, with most concentration in the AI hardware sector. This structural pressure means that while overall market leverage risk may be lower than in 2015, continued declines in related stocks could amplify volatility through concentrated financing positions.

Signal five and six focus on retail sentiment cooling and institutional allocation behaviors. Retail investors still contribute about 70% of the average daily trading volume in A shares, making their sentiment a direct driver of short-term fluctuations. Goldman Sachs' retail sentiment gauge, which includes financing data, new account openings, and turnover rates, has dropped from about 1 standard deviation above the one-year average a month ago to around 0 standard deviations.

This shift indicates that risk-on sentiment has moved from hot to neutral or slightly pessimistic, releasing speculative pressure but not reaching extreme pessimism or a "panic bottom." Meanwhile, institutional behavior is complex: domestic public funds manage nearly 40 trillion yuan, with about 7 trillion yuan allocated to stocks, representing 6.6% of A-share total market cap and 15% of free-floating market cap.

During adjustments, equity-based public funds increased their cash ratios, moderating risk, yet their allocation to technology stocks such as semiconductors, hardware, and software remains at historical highs. Systematic strategies, including quantitative funds, have shown more evident risk reduction, with weakened activity in small and medium-cap stocks. Thus, while quantitative and short-term funds have shrunk, traditional institutions have not significantly reduced core technology holdings, meaning AI hardware stocks cannot be considered fully de-congested.

Signal seven and eight analyze corporate actions and policy stance neutrality. Corporate buybacks rose to multi-year highs in the third quarter of 2026, with the number and value of announced transactions increasing by 35% and 59% year-on-year, respectively. This suggests management believes stock prices are below intrinsic value or are willing to support shareholder returns.

Additionally, warnings about abnormal stock price fluctuations, which peaked in June before the correction, have decreased significantly. Transactions by major shareholders and executives have shifted from net selling in the first half of the year to a more balanced state. These internal actions no longer lean toward selling or overheating warnings, signaling a positive shift. On the policy front, Goldman Sachs uses LLMs to analyze regulator statements, measuring support and tightening risks. Concerns about market overheating and policy tightening peaked in the first quarter of 2026 but have since declined to a more neutral level. Policies are currently neither a significant driver of sentiment nor a major source of pressure, remaining more neutral compared to price, leverage, and earnings factors.

Signal nine and ten cover "national team" trading shifts and global AI capital expenditure forecasts. The broader 'national team' holds about 5 trillion yuan in A-share assets, equivalent to 5% of total market cap. After selling about 1.5 trillion yuan in the previous six months, the "national team" switched to net buying worth over 140 billion yuan in the past three weeks, including small amounts in Science and Innovation 50 ETFs.

This shift indicates a changed assessment of market risks, providing a downward cushion, though it does not confirm a medium-term bottom. Historical backtests show that weekly net buying must exceed 1.5 standard deviations to likely correspond to a medium-term bottom. Regarding fundamentals, Goldman Sachs predicts AI spending by nine major U.S. and Chinese hyperscalers and cloud service providers will exceed 900 billion dollars this year, rising to 1.3 trillion dollars next year, accounting for 1.7% and 2.3% of the combined GDP of the two countries, respectively. Capital expenditure forecasts for 2026 and 2027 for eight listed hyperscalers have been raised by 32% and 75%, driving upward revisions in earnings forecasts for China's hardware sectors by 12% and 22% for fiscal years 2026 and 2027. This confirms that the rise in China's AI hardware sector is supported by capital expenditure and earnings growth, not just sentiment.

However, the pace of upward revisions in capital expenditure and earnings forecasts has slowed, with momentum showing signs of peaking in the short term. In contrast, soft-tech sectors still have room for periodic improvements in earnings revision momentum.

Synthesizing these ten signals, the current position of China's AI hardware sector is characterized by a mix of positive and cautious indicators. Positive developments include the narrowing return gap between hardware and soft-tech, rapid reversals in momentum and small-cap styles, valuations returning to long-term averages, retail sentiment cooling to neutral, increased corporate buybacks, balanced major shareholder selling, and the 'national team' switching to net buying. These changes indicate that much of the speculative positioning, valuation overheating, and leverage pressures from the first half of the year have been relieved.

However, cautious signals persist: absolute valuations of the Science and Innovation 50 remain high, trading volume and returns in A-share technology stocks are highly concentrated, financing balances are above historical norms, leverage concentration has hit a new high, and public fund allocation to technology stocks remains at historical highs. Fundamentally, global AI capital expenditure remains strong, and earnings forecasts for hardware sectors are being raised, but the momentum for upward revisions is slowing.

Consequently, Goldman Sachs maintains an 'overweight' rating for A shares and remains long-term optimistic about AI hardware sectors structurally. In the short term, the firm emphasizes rotation and diversified allocation, recommending gradual increases in exposure to selected Hong Kong soft-tech stocks, stocks benefiting from policies, and those focused on self-reliance. Investors are also advised to pay attention to stocks with rising earnings, IPOs, and cash returns from dividends and buybacks.

These ten signals do not simply indicate that 'AI has bottomed out' or that the trend is over; rather, they provide a sentiment thermometer showing that the hottest times may have passed and the steepest declines may be slowing, but concentrated positions and earnings expectations still need to be addressed. Future heating of AI trading in China will depend more on corporate earnings and AI monetization rather than relying again on valuation, leverage, and chasing rising prices.