Which semiconductor stocks are listed on the Hong Kong stock market? How does AI compute demand drive upgrades across China’s domestic chip industry supply chain?

The competitive focus across the global AI industry is undergoing a significant shift. In the first half of 2026, the market’s main narrative centered on large-model parameter scale, training costs, and inference efficiency; in the second half, the discussion has clearly shifted toward compute infrastructure. From NVIDIA’s GPU production capacity to SK hynix’s HBM storage, from TSMC’s advanced process technology to SMIC’s mature process expansion, the value allocation of the AI industrial chain is migrating from the application layer to upstream hardware.

Hong Kong-listed semiconductor stocks have become one of the key directions attracting capital attention during this narrative shift. On July 14, despite a pullback at high levels across the semiconductor sector overall, Semiconductor Manufacturing International Corporation (SMIC) (00981) and Hua Hong Semiconductor (01347) both saw intraday volatility and pullbacks, but there was no systemic capital withdrawal. On the day, southbound funds recorded a net inflow of about HK$9.038 billion, reversing the prior single-day net outflow.

Short-term sector volatility coexists with structural capital inflows, reflecting that the market’s mid-term consensus on China’s domestic chip industrial chain is taking shape. This article starts from the structural changes in AI compute demand, breaks down the end-to-end industrial-chain transmission logic from large models to semiconductor manufacturing, and outlines the core Hong Kong-listed semiconductor targets and investment risks.

AI compute demand shifts from a “software contest” to “hardware consumption”

The global AI industry’s development trajectory in 2026 shows distinct stage characteristics. In the first half, the key variables the market focused on were large-model parameter scale, multimodal capabilities, and user penetration; in the second half, the physical constraints on compute supply have begun to emerge as the primary bottleneck constraining industry development.

This judgment has strong data support. In a report released on July 4, 2026, the semiconductor industry research firm SemiAnalysis explicitly stated that after combining DRAM, NAND, and HBM, the share of memory spending within NVIDIA systems will exceed 30% by the end of 2026 and exceed 40% in 2027. The ongoing rise in memory chips’ share of capital expenditures for compute systems directly reflects the rigid demand for high-bandwidth storage for AI training and inference. Driven by AI demand, the share of memory spending within capital expenditures of ultra-large-scale cloud vendors and data center operators in 2026 will rise to about 30%, clearly higher than the roughly 8% level in 2023 to 2024.

At the same time, bottlenecks in the chip manufacturing link are becoming evident. On July 13, Intel announced an additional investment of €5 billion (about $5.7 billion) in its Leixlip campus in Ireland, to upgrade the manufacturing park, expand Xeon server processor capacity, and strengthen R&D capabilities. On July 9, Micron announced $500 million in strategic funding to GlobalWafers; both parties signed a 10-year long-term supply agreement to lock in wafer capacity in advance. Silicon wafers may become the next key bottleneck in AI compute hardware.

The dramatic volatility in the Korean market provides indirect confirmation of valuation pressure in the AI hardware segment. On July 14, SK hynix’s U.S. depositary receipts (ADR) fell 9.3%, nearly erasing the 13% gain recorded on the first day of listing last Friday, with the share price dropping to near the offering price of about $149. SK hynix ADR began trading on Nasdaq on July 10 at $149 per ADR. It issued about 177.9 million ADRs, raising roughly $26.5 billion, making it one of the largest IPOs by a foreign company in the United States. The second trading day saw a sharp reversal, reflecting market disagreement about whether a valuation bubble has been created by the AI chip boom.

The sell-off sentiment in Korea spread to Hong Kong’s semiconductor sector. On July 14, during intraday trading, GigaDevice (03986) fell by more than 15%, and Xinxing Insight (02166) dropped by nearly 13%. However, it is worth noting that Hong Kong’s broader market still rose on the day: the Hang Seng Index closed up 38.60 points. While some capital withdrew from the semiconductor sector, it did not leave the Hong Kong market—instead, it rotated into other sectors. This structural feature indicates that the sector’s pullback in Hong Kong’s semiconductor segment is more about reallocation of capital between sectors rather than the release of systemic risk.

Industrial-chain transmission logic: a complete closed loop from large models to wafer fabs

The boost from AI compute demand to the semiconductor industry is not a linear transmission, but instead expands step by step along a clear industrial chain path.

First layer: AI model companies and cloud computing platforms. Global leading AI model enterprises continue to expand capital expenditures to procure GPU clusters and build data centers. In 2026, cloud vendors such as Meta and Amazon have locked in upstream fiber-optic capacity through long-term agreements. This “long-term contract lock-in” means that visibility into orders for upstream hardware suppliers increases significantly, and earnings certainty is strengthened.

Second layer: GPUs and AI chips. GPU suppliers such as NVIDIA and AMD occupy a central hub position in the compute industry chain. Strong demand for AI chips directly drives capacity at foundry fabs. Worth noting is that, besides advanced process technology, mature process technology also benefits from the diffusion effect of the AI industry—demand for AI supporting chips such as power management chips, driver chips, and analog chips is exploding across the board.

Third layer: semiconductor manufacturing. This is the most core link in China’s domestic chip industrial chain and also the most direct mapping to Hong Kong-listed targets. In the first quarter of 2026, SMIC achieved sales revenue of $2.505 billion, up 11.5% year over year, with a gross margin of 20.1%. Capacity utilization reached as high as 93.1%. Even more notable is the second-quarter guidance: revenue is expected to grow 14% to 16% quarter over quarter, and gross margin is expected to be between 20% and 22%. Consumer electronics revenue grew 27% year over year, while industrial and automotive revenue surged 63%.

SMIC management said that, based on customer demand and the order book, the company is more optimistic about its full-year 2026 operating outlook. The company is negotiating price increases with customers, and expects both second-quarter shipment volume and average selling prices to rise meaningfully. Hua Hong Semiconductor’s statement is even more explicit: the company expects its average product selling price for full-year 2026 to increase by 10% to 15%, and the full-year price increase for products with strong demand may exceed 20% to 25%.

Fourth layer: semiconductor equipment and materials. Capacity expansion by wafer fabs requires supporting equipment and materials. Domestic semiconductor equipment companies continue to benefit from this round of domestic substitution. For the week from July 6 to July 10, the Sci-Tech Innovation-themed ETFs saw a large net subscription; the net subscription amount for the semiconductor equipment segment reached RMB 25.508 billion. Among 18 semiconductor equipment thematic ETFs, 15 achieved net purchases, totaling RMB 29.168 billion.

Fifth layer: a complete closed loop of domestic supply chains. From AI models to compute chips, from foundry outsourcing to equipment and materials, China’s domestic semiconductor industrial chain is forming an increasingly complete investment narrative. With this closed loop in place, Hong Kong’s semiconductor sector is no longer merely a passive mapping of the Philadelphia Semiconductor Index; it now has an independent logic for value re-rating.

Near-term pullback versus mid-term logic: the tug-of-war

The overall pullback in Hong Kong’s semiconductor sector on July 14 needs to be understood in a broader macro context.

On external factors: tensions in the Middle East escalated, international oil prices rose significantly, U.S. Treasury yields moved higher, and this dragged on the performance of U.S. tech stocks and chip stocks. The Philadelphia Semiconductor Index fell sharply on the same day as well. These external shocks influenced Hong Kong’s semiconductor sector in the short term through sentiment transmission.

On internal factors: after a strong rally since April, the semiconductor sector accumulated substantial gains. This pullback was jointly catalyzed by profit-taking pressure and institutions’ portfolio rebalancing ahead of the interim report window. In addition, the capital outflow effect from newly unlocked shares from AI large-model IPOs also cannot be ignored.

But the mid-term logic has not been reversed by a single-day pullback. Southbound funds turned into a net inflow of about HK$9.038 billion on the day, ending the prior single-day outflow pattern. SMIC topped the list of the 10 most actively traded stocks on the Stock Connect (Shanghai). Capital is still being structurally allocated around domestic compute and the AI industrial chain.

From a valuation perspective, Hong Kong valuations remain at historic lows, with a PE of about 10.3x. This valuation level provides a margin of safety for medium- to long-term allocation.

Risk factors and observation dimensions

Although the mid-term narrative for Hong Kong’s semiconductor sector is clear, the following risk dimensions still need to be continuously tracked.

Global semiconductor cycle timing. AI infrastructure investment has significantly raised semiconductor demand, but SK hynix ADR’s sharp pullback reminds the market that the valuation bubble triggered by the AI chip boom is under pressure and starting to unwind. Doubts among global investors about whether the AI rally can continue are growing.

Geopolitics and supply-chain security. Geopolitical events such as the escalation of conflict between the United States and Iran indirectly impact the semiconductor sector by affecting energy prices and risk appetite. While domestic substitution provides structural opportunities for domestic semiconductor companies, geopolitical risks in the supply chain remain an important source of uncertainty.

Capacity expansion and supply-demand balance. Current foundry capacity is in a tight balance, with SMIC’s capacity utilization at 93.1%. However, whether large-scale expansion could cause the future supply-demand relationship to reverse is a variable that needs ongoing observation.

Stock-level differentiation. Hong Kong’s semiconductor sector is not one uniform block. Foundry companies (SMIC, Hua Hong Semiconductor) benefit from the price-increase logic and full capacity utilization, so their earnings certainty is relatively higher. Meanwhile, some design companies may face pressure from inventory adjustments and intensifying competition. Fine-grained analysis at the stock-selection level is indispensable.

Conclusion

The pullback of Hong Kong’s semiconductor sector on July 14, 2026 is a collision between short-term sentiment disruptions and a mid-term industrial trend. From AI compute demand to chip manufacturing, from wafer price increases to full capacity utilization, the transmission logic of the domestic chip industrial chain is moving from narrative toward earnings execution. Data such as SMIC’s first-quarter revenue exceeding $2.5 billion and capacity utilization above 93% provides verifiable fundamental support for this logic.

The structural shift in the global AI industry from a “software contest” to “hardware consumption” is the core framework for understanding the current Hong Kong semiconductor market. Under this framework, short-term volatility should not obscure the mid-term trend—demand for AI compute and chips is not a cyclical pulse, but a structural uplift. As an important financing and pricing platform for the domestic chip industrial chain, Hong Kong’s semiconductor sector’s value re-rating may still be at an early stage.

Of course, the timing of the global semiconductor cycle, geopolitical uncertainty, and the sustainability of capacity expansion are all risk variables that need to be continuously monitored. While investors focus on investment opportunities across the industrial chain, they should also maintain full awareness of the risks mentioned above.

FAQ

Q1: What are the main representative stocks in Hong Kong’s semiconductor sector?

The core targets in Hong Kong’s semiconductor sector include SMIC (00981) and Hua Hong Semiconductor (01347), as well as GigaDevice (03986), Shanghai Fudan (01385), Xinxing Insight (02166), and others. Among them, SMIC and Hua Hong Semiconductor benefit from wafer price increases and full capacity utilization, so their earnings certainty is relatively higher.

Q2: How does AI compute demand transmit to semiconductor manufacturing?

AI model training and inference require large amounts of compute-chip components such as GPUs and HBM storage. Manufacturing these chips depends on foundry capacity. The diffusion effect of AI demand also boosts demand for supporting chips such as power management, drivers, and analog components, allowing mature process technology to benefit as well. SMIC’s capacity utilization of 93.1% in the first quarter of 2026 directly reflects this transmission logic.

Q3: How has SMIC performed recently?

In the first quarter of 2026, SMIC achieved sales revenue of $2.505 billion, up 11.5% year over year, with a gross margin of 20.1%. The company expects second-quarter revenue to grow 14% to 16% quarter over quarter, and gross margin to rise to 20% to 22%. Consumer electronics revenue grew 27% year over year, while industrial and automotive revenue increased 63%.

Q4: What are the main risks for Hong Kong’s semiconductor sector right now?

Major risks include: uncertainty in the global semiconductor cycle (a sharp drop in SK hynix ADR signals valuation pressure), geopolitical events impacting risk appetite, large-scale capacity expansion potentially reversing future supply-demand relationships, and differentiation in underlying fundamentals among companies within the sector. Investors need to make differentiated judgments based on the earnings certainty of specific targets.

Q5: How much impact do southbound funds have on Hong Kong’s semiconductor sector?

Southbound funds are an important source of liquidity for Hong Kong’s semiconductor sector. On July 14, southbound funds recorded a net inflow of about HK$9.038 billion, reversing the prior single-day outflow trend. SMIC topped the list of the 10 most actively traded stocks on the Stock Connect, indicating that mainland investors’ allocation willingness toward China’s domestic chip industrial chain remains sustained.