Event Overview and Key Uncertainties

On 4 August 2026, Reuters reported — citing four people familiar with the matter — that the US Federal Communications Commission (FCC) is drafting a measure to restrict new models of data-centre optical transceivers made by Chinese companies from entering the US market, with officials hoping to introduce it within the year. Neither the White House nor the FCC has commented publicly. The proposal is not final, and neither the effective date nor any transition arrangements have been specified.

Three details will determine what the proposal actually means in practice. First, any restriction may apply only to new models, which would limit the immediate impact on products already approved. However, the definition of a new model, whether existing certifications carry over after a product upgrade, and whether a transition period applies have not been disclosed — so delivery and revenue risk cannot yet be ruled out. Second, the reported mechanism is to restrict all new models in principle and then grant exemptions to a large number of non-Chinese suppliers, structurally similar to earlier measures on drones and routers. Third, the stated rationale is infrastructure security: guarding against data theft, malware and service disruption before large-scale deployment occurs, rather than remedying existing harm.

The shape of the market reaction is more informative than its magnitude. The day after the report, the leading Chinese optical transceiver supplier fell 13–16% intraday; after the company noted that no rules had yet been issued, the shares rebounded and briefly traded up more than 9%. Peers fell roughly 5–10%, while upstream optical component suppliers — only indirectly affected — declined less. The sharp intraday volatility and subsequent rebound reflected the market’s difficulty in pricing an unfinalised proposal, rather than any same-day change in the underlying earnings outlook.

The initial market reaction primarily reflected the potential redistribution of supply-chain market share between Chinese and non-Chinese suppliers, rather than a deterioration in underlying AI demand. However, a shortage of replacement capacity could subsequently raise procurement costs, delay data-centre deployment and ultimately slow the pace of AI capital expenditure.

Directly Affected Suppliers: US Revenue Exposure and Substitution Difficulty

Revenue exposure is central to assessing the risk faced by this segment. Innolight Technology (300308.SZ), the global leader by datacom optical transceiver market share, generated RMB 21.897 billion of 2025 revenue from the US market, or 57.3% of the total. Mainland China contributed RMB 3.603 billion (9.4%) and other overseas markets RMB 12.740 billion (33.3%). In Q1 2026 the mix tilted further toward the US, at 61.7%, versus 34.6% for other overseas markets and 3.7% for mainland China. Customer concentration is similarly high: the top five customers accounted for 24.1%, 18.3%, 14.2%, 11.3% and 8.1% of 2025 revenue respectively.

With more than 60% of revenue tied directly to US data-centre customers and domestic revenue below 4%, an effective restriction covering the company’s main products in development would have a material impact on revenue, customer concentration and margins. This should not be dismissed as a purely sentiment-driven risk.

Disclosure: Innolight Technology was added to the US Department of Defense’s “Chinese military-related companies” list in June 2026. The company has denied the designation.

The ability of alternative suppliers to replace this capacity is constrained by several factors. Market research estimates put major Chinese suppliers at close to 50% of the global datacom optical transceiver market in 2025, against approximately 12–13% for the largest single US-based supplier. Replacing the displaced capacity would require alternative suppliers to expand production, complete customer qualification and ramp manufacturing yields. Policy support may accelerate investment, but it cannot eliminate the time required for capacity expansion, qualification and yield improvement.

On the other side, the Chinese market leader already operates high-end capacity in Thailand, Malaysia and at domestic sites in Suzhou and Tongling, with a Mexico facility potentially starting production in Q4. Whether that overseas footprint provides a buffer depends on whether the final rules define coverage by place of production, by ownership or by component origin. Our assessment therefore remains conditional on how the final rules define the scope of coverage.

Portfolio-Relevant Segments of China’s AI Hardware Supply Chain

AI data-centre build-out extends well beyond chips, to servers and system assembly, printed circuit boards, power and thermal management, and upstream semiconductor equipment and materials. China has strong at-scale delivery capability across several of these segments, while external dependencies persist in high-end compute chips, advanced memory and certain core components. China’s supply-chain strengths are therefore not confined to end-product assembly but extend into several critical infrastructure segments — and these segments also account for a substantial share of the index’s current weight.

Demand for these segments is underpinned by the rising share of domestic chips, servers and foundational software in China’s incremental computing deployment, driven by supply-security and cost considerations. That demand comes from domestic model training, inference and application deployment, and is not directly linked to US market-access conditions for Chinese optical transceivers.

Successive rounds of external restrictions have historically reinforced domestic customers’ focus on supply-chain security and accelerated the adoption of domestic alternatives in equipment, materials and compute chips. At the same time, in segments still dependent on overseas technology, restrictions can raise R&D costs, extend qualification cycles and constrain near-term supply. The strategic direction toward greater supply-chain localisation has therefore become clearer, though the pace of realisation still depends on technological maturity and commercialisation capability (discussed at greater length in “Global Compute Resonance, Domestic Substitution Accelerates”). Progress at the model and application layers, and China’s relative advantages in power supply and compute–power coordination, form part of the broader industry backdrop rather than a direct argument at the index holdings level.

Allocation Framework and Exposure Check

Our framework, set out in “Market Update: Q2 GDP +4.3%, Growth Momentum Rotates Toward AI and Semiconductors”, comprises three layers: the GF CSI All-Share IT ETF (159939.SZ) for AI hardware, the Hang Seng Tech ETF (3067.HK) for digital platform monetisation, and the T. Rowe Price China Evolving World Fund (LU2187417386) for physical infrastructure. Whether this development changes the earnings outlook underpinning any of these layers can be checked on two dimensions.

Industry classification and index constituents. Under the current CSI classification, datacom optical transceiver suppliers sit within communications equipment and are not included in the CSI All-Share Information Technology Index. No current Hang Seng Tech Index constituent has data-centre optical transceivers as its principal business. On that basis, the two passive vehicles have low direct exposure to this development. Classifications, company business mixes and index constituents can all change, so the outcome of periodic index reviews warrants monitoring. For actively managed funds, only the most recently disclosed holdings can be used as a reference, and those holdings may since have changed.

Holdings composition and revenue sources. The index’s core weight sits in semiconductor equipment and materials, domestic compute chips, display panels and printed circuit boards. Demand in these segments comes mainly from the domestic market and from orders driven by supply-chain localisation, so import restrictions on optical transceivers do not create a direct revenue transmission channel. Diversification provides a further buffer: as at 31 December 2025 the index had close to 200 constituents, with the top ten at roughly 30% and a single-name cap of around 4%.

Constituents already placed on various restriction lists are mostly domestic-substitution names, and most have continued to grow revenue or profit in recent years on domestic demand — indicating that external restrictions have not closed off their commercialisation path, and that these designations are already-known information reflected in market pricing. Some constituents nonetheless carry overseas customer and global AI capex exposure. If the rules were extended to printed circuit boards, server components, power or thermal management, the impact would move from sentiment to earnings.

Evolving Regulatory Approach and Key Developments to Monitor

The analysis above addresses the portfolio’s current exposure. The more important issue, however, is how the US regulatory approach is evolving. Unlike traditional export controls, which restrict the outflow of US technology to China, the measures discussed here operate through US equipment authorisation, import restrictions and supply-chain security reviews. Reuters previously reported that the US Department of Commerce shelved a set of China-focused import restrictions, including on data-centre equipment, following an easing in trade relations; the FCC subsequently advanced measures on drones and routers through the covered list and equipment authorisation mechanisms. US restrictions on Chinese technology are increasingly extending beyond designated entities to the market access of specific product categories, though how measures across different agencies relate to trade negotiations awaits confirmation in formal documentation.

The expansion of the regulatory perimeter is no longer hypothetical. On 22 July 2026 the FCC adopted a Report and Order bringing logic components made by covered-list entities within scope: equipment containing such components cannot obtain authorisation even if produced by a third party and assembled outside China. The rule applies from the date of adoption with no transition period; software and firmware are not currently included. That component rule applies only to list entries defined by producer, and does not yet apply to newly added categories defined by place of production — and the optical transceiver proposal is closer to the latter. Both bases for determining coverage are widening in parallel. The key risk is therefore not limited to optical transceivers themselves, but lies in whether the same regulatory logic is extended to product categories with meaningful index weight.

There is, however, a potential countervailing catalyst. The leaders of the United States and China are reported to be meeting in the US on 24 September, with artificial intelligence on the agenda; foreign ministers and vice-ministers have already held preparatory talks. An understanding on AI-related issues could slow the pace of further restrictions. Equally, disagreements over frontier model regulation, intellectual property and export controls could complicate the summit, and we would not treat it as a one-way positive for the sector.

We identify three developments that would require us to reassess our current view: final measures extending to complete systems or upstream components containing the relevant parts; restrictions extending to printed circuit boards, server components, power or thermal management, which carry meaningful index weight; or supply-chain adjustment materially slowing global AI capital expenditure. Should any of these occur, we would reassess the earnings exposure and portfolio role of the relevant vehicles.

Near-Term Impact and Longer-Term Investment Implications

Near-term impact depends on which entity is being assessed. For Chinese optical transceiver suppliers selling directly to US data-centre customers, the risk sits on the revenue line. For the passive index vehicles discussed here, based on current constituents, the impact would be transmitted indirectly through risk appetite, sector valuations and portfolio rebalancing. The two are different in nature and should not be conflated.

The medium- to long-term investment case remains intact and, in some respects, has become clearer. Global AI development is increasingly taking place across two parallel ecosystems: one built around frontier chips and large-scale capital expenditure by leading cloud providers, and another focused on cluster scale, system-level computing efficiency and hardware–software co-design under constraints on access to advanced process nodes. These ecosystems are not substitutes for one another and should be treated as distinct investment exposures.

As market-access and certification regimes diverge, the coexistence of two increasingly separate supply chains may become structural rather than temporary. In such an environment, greater control over critical technologies and supply chains becomes a strategic necessity rather than an optional objective. China’s domestic AI infrastructure and application ecosystem is therefore likely to continue expanding, with domestic suppliers capturing a rising share of the required hardware. At the same time, supply-chain fragmentation will raise build costs on both sides and create increasingly distinct addressable markets — effects that are likely to unfold over several years rather than appear immediately in reported earnings.

Recommended Vehicles and Portfolio Framework

At this stage, the development does not justify revising the existing portfolio framework, and we maintain all three layers unchanged.

Vehicle	Ticker	Market / Domicile	Benchmark Index	Portfolio Role
GF CSI All-Share IT ETF	159939.SZ	Shenzhen Stock Exchange	CSI All-Share Information Technology Index	AI Hardware
Hang Seng Tech ETF	3067.HK	Hong Kong Stock Exchange	Hang Seng Tech Index	Digital Platforms
T. Rowe Price China Evolving World Fund	LU2187417386	Luxembourg (UCITS)	MSCI China All Shares Index	Physical Infrastructure
Source: Fund managers and index providers (public disclosures), compiled by iFAST Research.

These positions reflect a medium- to long-term view: current AI capital investment is expected to translate into revenue and profit over the next two to three years. Investors can set the relative weighting of each layer according to their own risk tolerance and investment horizon. Short-term price moves driven by a single policy headline, absent a change in the earnings outlook or the valuation anchor, are generally not sufficient on their own to justify revising a long-term framework. Where such moves materially change valuations, position concentration or the portfolio’s risk budget, a rebalancing assessment remains warranted.

Related Articles

• China’s AI hardware opportunity: Global compute boom meets accelerating domestic substitution
• China's Structural Bull Market Case Still Holds: From Valuation Repair to the AI Engine
• China's 2Q GDP slowed, but AI and Semiconductors are driving the next growth cycle 

Risk Factors

• Uncertainty over the final scope of the rules. The proposal has not been finalised. The definition of a new model, the treatment of existing certifications, the handling of overseas capacity and whether upstream components are covered all remain unspecified. A stricter final scope than currently reported would widen the affected universe.
• Extension of restrictions to index-relevant segments. Coverage is being defined both by producer and by place of production, and both bases are widening. An extension to printed circuit boards, server components, power or thermal management would move the impact on affected constituents from valuation and sentiment to revenue.
• Two-sided geopolitical risk. The outcome of the September US–China leaders’ summit is uncertain. Both an understanding and an escalation in friction are possible, with opposite effects on sector risk appetite.
• Slower global AI capital expenditure. Supply-chain fragmentation raises build costs on both sides. If it delays data-centre delivery schedules, earnings forecasts for the affected sectors would face downward revision.

This report has been prepared by the iFAST China Research team for general information purposes and is intended for public distribution. It does not constitute, and should not be construed as, an offer, solicitation, or investment recommendation to buy or sell any securities, funds, or investment products, and does not take into account the investment objectives, financial situation, or specific needs of any particular recipient.

The information, views, projections and estimates contained in this report are based on publicly available information and independent judgement that the team considers reliable; however, the Company makes no express or implied warranties as to their accuracy, completeness, or timeliness. Forward-looking statements in this report are subject to uncertainty, and actual results may differ materially from those expressed.

Past performance is not indicative of future returns. The value of investments and any income derived from them may fall as well as rise, and investors may not recover the amount originally invested. Where investments are denominated in a foreign currency, exchange rate movements may also adversely affect the value of the investment.

Investors should independently and carefully assess the relevant risks before making any investment decision, and should seek independent financial, legal, or tax professional advice as appropriate to their individual circumstances.

The Company and related licensed entities within the iFAST Group (including iFAST Platform Limited (IFPL) and others) are regulated by the relevant supervisory authorities in their respective jurisdictions. For further information about the Company and this report, please visit the official iFAST website.

Analyst Disclosure

As at the date of publication of this report, the author Ian Li, CFA, and his immediate family members hold no position or interest (NIL) in any of the securities or investment products referred to in this report. The author’s compensation is not directly linked to the specific views expressed in this report.

Artificial Intelligence Disclosure

This report may have used AI-assisted tools in the course of drafting, data compilation, and chart preparation. All outputs have been independently reviewed and verified by the research team.