Goldman Sachs: Chinese stocks enter “rotation temptation”; A-share hard tech remains superior, H-share internet and others see profit recovery

The main theme of Chinese stocks this year is very clear: A-share hard tech is strong, H-share internet is weak. STAR50, representing onshore AI hard tech, has outperformed HSTECH, which focuses on offshore internet platforms, by 68 percentage points since the start of the year; the ChiNext Index has outperformed the Shanghai Composite and CSI 300 by 19 and 17 percentage points, respectively. This gap is already at an extreme range in the history of the Chinese stock market.

According to Chase Wind Trading Desk, Goldman Sachs Asia portfolio strategists Kinger Lau and others wrote in their China portfolio strategy research released on July 13: “We reiterate a broad tactical preference for A-shares over H-shares and hard tech over soft tech, but will begin to gradually allocate to some H-share soft tech/internet names after the valuation correction, and expect earnings recovery in the coming months.”

This is not simply a choice between A-shares and H-shares. The core difference in the framework is: hard tech is closer to the supply side of AI infrastructure, while soft tech and internet platforms are more like the AI capex bearers. The former benefits from demand for computing power, semiconductors, equipment, and infrastructure; the latter still needs to prove that AI investment and subsidy spending will not continue to eat into profits.

Profitability is the key barrier for H-share internet. MSCI China’s Q1 profits fell 8% year-on-year, with a significant drag from the internet sector; this sector accounts for about 35% of the index’s earnings weight and has accumulated over RMB 180 billion in subsidy losses since Q2 2025, while also being the main capex bearer for domestic AI. Relevant AI capex may exceed USD 100 billion and USD 120 billion in 2025 and 2026, respectively.

A-share hard tech remains the main theme, but rotation pressure has emerged

Profit divergence among Asian stocks this year occurs not only in mainland China. South Korea and Taiwan benefit from memory, foundry, and AI infrastructure, attracting international investor attention. Internal divergence in mainland China is more subtle: the gap between A-share hard tech and H-share internet is far larger than many overseas investors anticipated.

HSTECH, which focuses on offshore internet platforms, rebounded 11% in the past two weeks, but is still down 14% year-to-date. This rebound mainly comes from several factors: improved news for some internet leaders, compressed starting valuations, and continued progress in AI application and commercialization.

The model suggests that in the coming months, H-share soft tech’s underperformance relative to A-share hard tech may narrow. Variables such as valuation, earnings expectations, and liquidity support this judgment.

However, sustained rebound is still missing a key piece: profit. Some leading internet companies' valuations may already reflect pessimistic scenarios of continued devaluation of core business and negative AI investment net present value or ROI. But for the market to reassess valuations, a recovery in underlying profits is still needed, which may take several months or even quarters.

Without profits, H-shares find it hard to sustain a rally

MSCI China’s performance recently has been dragged by profitability. Six weeks ago, MSCI China rating was downgraded from overweight to neutral, with poor earnings delivery as a key driver.

In 2025, index earnings will grow only 7%, marking the sixth consecutive year of below-consensus results. By Q1 2026, profit pressures remain, with subsidies and AI investments in the internet sector further amplifying them.

Subsequent profit recovery depends mainly on three things.

First, whether subsidy losses can be narrowed. Over the past year, internet platforms have invested heavily in fast commerce and other areas, pressuring their profit statements.

Second, whether new AI-related opportunities can contribute income, including cloud, Agentic AI, AI tokens, etc.

Third, whether traditional e-commerce businesses can maintain stable cash flow. Only if these businesses continue to generate cash, will AI investment not be seen as simply a profit black hole by the market.

If operating profit turning points appear in Q2 or Q3 earnings, internet companies' valuation methodology may change. The market will no longer only discount blended profits, but is more likely to use segment valuations to measure core business, AI input, and new business options.

China AI is not an overall bubble, but some trades are crowded

Regarding whether “A-share AI is overheated,” the answer is not simply yes or no.

Since the DeepSeek moment, the market value of China AI-related stocks has increased, but based on top-down estimates, the economic benefits from AI through efficiency gains, new profits and potential market space may be 50-100% higher than current share prices reflect. This is why “China AI stocks are not considered an overall bubble.”

Risks are local.

The semiconductor sector and some A-share hard tech targets are already at high valuations relative to historical levels and global comparables. Meanwhile, concentration and leverage risk are rising. This means it’s still possible to follow the AI main theme, but you can’t treat all AI assets as the same trade anymore.

A more appropriate approach is to control exposure in any single direction, diversify across different segments of the AI value chain, and prioritize the ability to deliver profits. Pure thematic trading still has room, but risk-return is no longer as clean as at the start of the year.

Foreign capital hasn't left mainland China, but prefers Alpha over Beta

A common question is: has AI rallies in Korea and Taiwan siphoned funds away from mainland Chinese stocks?

Data shows a more complex picture. The directional return correlation of mainland Chinese stocks with Korea and Taiwan are near historical lows, indicating investors do not simply treat the three as the same “North Asia AI trade.”

Hedge funds have net risk exposure at historic highs in emerging Asian markets excluding mainland China, while net risk exposure to mainland stocks sits at the bottom of the range. However, total exposure is near cyclical highs, meaning funds have not fully withdrawn, just shifted more toward market-neutral and single-stock Alpha.

Mutual funds show another side. Especially emerging market funds, allocations to mainland China have risen to multi-year highs, with a mild overweight for the first time since 2011. This change may relate to China’s declining weight in the emerging market benchmark.

Hong Kong IPOs also show foreign participation. Foreign cornerstone investor participation this year is near 2021 highs. Funds are still there, but not rushing to buy index Beta, and prefer structural opportunities.

Hong Kong IPOs are hot, but “siphoning” concerns may be exaggerated

This year’s Hong Kong IPOs have become an important channel to capture Alpha.

So far this year, 100 companies have listed on HKEX, raising USD 35 billion in total. Median stock price returns after 1 and 3 months post-listing are 32% and 30%, respectively.

The best performing IPOs commonly share several traits: larger issuance size, listing as independent H-shares, and moderate cornerstone investor shareholding. A cornerstone ratio of 30-50% has strong explanatory power for subsequent performance.

The market fears too many IPOs will drain liquidity. According to available estimates, the remaining Hong Kong IPOs this year will total about USD 25 billion; including follow-on offerings, total size is about USD 45 billion. By contrast, cash returned to shareholders through dividends and buybacks is much larger—about USD 500 billion and USD 560 billion in FY2025 and FY2026, respectively.

A-share IPOs are also accelerating. Chinese DRAM maker CXMT reportedly plans to list on the STAR Market to raise USD 4.3 billion; if successful, it will be the largest A-share IPO since May 2022.

AI overshadows macro, but macro hasn’t disappeared

Recently, investor discussion has clearly shifted towards AI. Issues around China’s AI ecosystem are more specific: progress in building an indigenous ecosystem, bottlenecks in EDA, lithography, advanced packaging, HBM, cost-efficiency differences in AI capex between China and the U.S., and the global competitive pressure from China in large models, robotics, and AI tokens.

China is already an important part of the global AI equity pool. Chinese companies account for about 11% of global AI-related market cap, and about 18% of revenue. But international investor allocation to China AI stocks is still low, especially in power, infrastructure, and physical AI layers.

By OpenRouter’s standard, China’s AI model token usage is close to half the global total. This means China AI is not just a supply chain story—it has now progressed into application and consumption.

Macro topics are temporarily in the background but remain on investor watchlists. Discussion includes: whether the real estate market will stabilize in 2027, China’s economic resilience amid global oil price shocks, continued weakness in domestic consumption, China’s export performance, etc.

Second half: A-shares still preferable, H-share internet can be monitored gradually

In the second half of 2026, tactical preference remains with A-shares at the market level. Main reasons: first, earnings momentum cycle is more favorable; second, for international investors, A-shares still offer underappreciated diversification value.

However, large H-share internet stocks have entered a watch zone. After a significant devaluation this year, if profits recover in the next few quarters, these stocks could shift from “low-value traps” to “profit recovery trades.”

At the sector level, materials, capital goods, and insurance are listed as overweight directions. Outside AI, a few non-AI micro clues can also be noted: shipbuilding, new consumption, medical health/biotech, real estate extension industry chain, brokerages, and the pig cycle.

~~~~~~~~~~~~~~~~~~~~~~~~

The above content is from Chase Wind Trading Desk.

For more detailed analysis, including real-time interpretations, frontline research and more, please join [Chase Wind Trading Desk ▪ Annual Membership]

Risk warning and disclaimerThe market carries risks; investment requires caution. This article does not constitute personal investment advice, nor does it take into account the specific investment objectives, financial status, or needs of individual users. Readers should consider whether any opinions, views or conclusions in this article suit their specific situation. Investment based on this is at your own risk.