<p>The logic for domestic chips hasn't changed; why did SMIC and Hua Hong fall first?</p>

On October 8, SMIC and Hua Hong Semiconductor saw significant stock price pullbacks, with chipmakers like GigaDevice and Montage Technology also weakening in tandem. The correction in the domestic semiconductor sector has brought a question back into investors' focus: while domestic substitution remains the long-term industrial theme and AI computing demand is spreading to memory, interconnect, and manufacturing segments, why are related companies still under pressure?

Looking at the latest financial reports, the industry fundamentals have not broadly weakened. SMIC's Q2 2026 sales revenue reached $3.006bn (+36.1% YoY); Hua Hong Semiconductor's sales revenue hit $717.5mn (+26.8% YoY); and Montage Technology's H1 operating revenue reached ¥3.335bn (+26.7% YoY). These figures indicate that domestic chip firms continue to benefit from growing manufacturing demand, server upgrades, and local supply chain construction.

However, revenue growth does not guarantee equal profit returns for all companies. Foundry operations require continuous investment in equipment and capacity, memory chips are subject to supply-demand cycles and price fluctuations, and high-speed interconnect chips rely more heavily on technology iterations and customer adoption. Even within the same industry chain, different enterprises exhibit distinct differences in cost structures, profit elasticity, and cash flow quality.

The long-term direction of domestic substitution remains unchanged, but the quality of growth within the industry requires re-examination. Whether SMIC can convert capacity expansion into profits, if Hua Hong can continuously improve its product mix, whether GigaDevice can sustain memory business growth, and if Montage Technology can translate next-gen interconnect products into scaled revenue—these operational issues will determine how much of the industrial opportunity ultimately gets realized.

Foundry growth is materializing, but profitability post-expansion is key

SMIC's latest earnings report reflects continued support for domestic foundry demand. In Q2 2026, wafer shipments grew 20.1% YoY, capacity utilization reached 93.7%, sales revenue increased 36.1% YoY, and gross margin rose to 25.3%. The company guided for Q3 revenue to grow 2-4% QoQ, with gross margins expected between 26% and 28%.

Revenue growth and improved capacity utilization suggest SMIC continues to capture downstream demand. However, wafer fabrication is a typical heavy-asset business; order growth requires coordination from equipment, facilities, and process capabilities, and the benefits of expansion must withstand tests from depreciation and costs.

In Q2, SMIC's capital expenditure reached $1.836bn, with depreciation and amortization at $925mn. Continuous investment helps meet customer needs and enhance manufacturing capabilities, but it also means the company must maintain high capacity absorption levels. If new capacity releases faster than demand growth, or if intensified competition in mature processes puts price pressure on revenues, revenue growth may not translate proportionally into profits.

Hua Hong Semiconductor's growth path focuses more on specialty processes. In Q2 2026, sales revenue grew 26.8% YoY, gross margin rose to 16.5%, and capacity utilization hit 102.8%. Embedded non-volatile memory revenue grew 41.8% YoY, while standalone non-volatile memory revenue surged 149.3% YoY. Improved memory demand and product mix shifts have provided new growth momentum for Hua Hong.

Hua Hong's advantage lies in its specialty process platform and accumulation in niche markets, but high capacity utilization does not imply sustained margin expansion. Future observation should focus on order visibility, product pricing, and ROI on new capacity. While SMIC holds advantages in scale and manufacturing capability, Hua Hong continues to expand in specialty processes; both companies' performance must be judged based on specific products and profitability.

For foundries, domestic substitution brings customer opportunities, but true long-term competitiveness still depends on process, cost, delivery, and capital efficiency. Expansion is just the starting point; converting capacity into profit is the critical factor.

Memory and high-speed interconnects: AI demand creating divergent growth curves

The growth logic for memory chips differs significantly from foundries. GigaDevice's H1 2026 operating revenue reached ¥11.566bn (+178.67% YoY), with net profit attributable to shareholders reaching ¥6.857bn (+1091.50% YoY). The company disclosed that niche DRAM and SLC NAND Flash benefited from market supply changes, seeing simultaneous volume and price increases, making memory business a key source of performance growth.

This data reflects that domestic memory firms are benefiting from supply-demand improvement and local supply chain opportunities. Applications in industrial control, network communications, and automotive electronics provide a broader demand base for memory products. However, the cyclical nature of the memory industry remains prominent; supply changes can push up prices but may also rebalance the market after expansion.

Therefore, assessing whether GigaDevice's growth can continue requires monitoring product shipments, market share, and price trends. Temporary volume-price surges bring strong profit elasticity but cannot be directly extrapolated as long-term growth rates. As industry supply and demand shift, product competitiveness and customer structure will determine earnings stability.

Montage Technology's opportunity stems from server memory architecture upgrades. AI servers demand increasing memory bandwidth and data transmission efficiency. DDR5 iterations and MRCD/MDB, PCIe Retimer, CKD, and CXL-related products have created new market space for high-speed interconnect chips.

In H1 2026, Montage Technology's operating revenue reached ¥3.335bn (+26.7% YoY), with net profit attributable to shareholders at ¥1.997bn (+72.3% YoY). Interconnect chip revenue reached ¥3.111bn (+26.4% YoY), with four new interconnect products contributing ¥538mn (+80.7% YoY). The company disclosed that its fifth-generation DDR5 RCD chips have begun mass shipments, with sixth-generation products sampled by customers in June.

Compared to foundries, Montage Technology's core competitiveness derives more from chip design, technology iteration, and customer certification. New product ramp-up offers opportunities to improve revenue structure and increase participation in the server memory value chain. However, sampling does not equal scaled revenue, and product iteration requires ongoing R&D investment. Subsequent shipment volumes, gross margins, and customer expansion for new products remain key indicators for judging growth quality.

The three types of companies thus present different growth curves: SMIC and Hua Hong rely on manufacturing demand and capacity efficiency, GigaDevice is influenced by memory supply-demand cycles, and Montage Technology benefits from server architecture upgrades. Domestic substitution provides a common industrial backdrop but cannot replace specific analysis of corporate business models.

Domestic substitution enters deep water; growth must ultimately reflect in financials

The collective pullback on October 8 occurred against the backdrop of overall pressure on Hong Kong tech stocks. A single day's decline is insufficient to prove a reversal in semiconductor industry trends, but it reminds investors that there is a gap between industry direction and individual stock performance. Industrial demand growth is merely the starting point; a company's cost structure, competitive moats, and ability to realize profits determine long-term returns.

Going forward, AI server expansion is expected to continue driving demand in memory, interconnect, and manufacturing segments, while industrial and automotive electronics offer new application spaces for domestic chips. However, industrial opportunities will not be evenly distributed among all enterprises. SMIC needs to continuously validate expansion returns, Hua Hong needs to improve its specialty process product mix, GigaDevice needs to navigate memory cycle changes, and Montage Technology must translate technology iterations into sustained commercial revenue.

For investors, subsequent focus should be on revenue growth rates, gross margins, capital expenditures, new product shipments, and operating cash flows for each company. For industry chain enterprises, clearly explaining order sources, product progress, and investment returns will also help the capital market understand their long-term value.

Domeestic chips still have room for growth, but industry competition is entering a finer-grained stage. Enterprises that can convert technological accumulation into customer orders, transform revenue growth into profits, and then turn profits into cash flow are better positioned to weather industry cycles. Domestic substitution provides development opportunities, but operational quality determines how much of those opportunities are ultimately realized.