A rumor on August 4 sent shockwaves through global capital markets. Reports indicate that the US government is drafting a regulation to prohibit imports of new-generation data center components from China, aiming to secure critical infrastructure powering artificial intelligence. This news triggered a surge in US optical component stocks, with AAOI jumping 19% and Coherent, Lumentum each rising 15% during trading, alongside an 8% gain for Corning. On the other side of the Pacific, Chinese optical module stocks plunged on August 5. Zhongji Innolight (A-shares: SZ300308, price CNY 947.74, market cap CNY 1.11 trillion; HKEx: 03308, price HKD 1,116.00, market cap HKD 1.31 trillion) opened down about 8% in both markets, while Eoptolink Technology (SZSE: 300502, price CNY 424.30, market cap CNY 591.6 billion) and TFC Communication (SZSE: 300394, price CNY 216.85, market cap CNY 236.5 billion) also declined. By close, these stocks moved -7.27%, -5.18%, -5.29%, and +2.29%, respectively. On August 5 morning, an investor relations representative from a leading Chinese optical module manufacturer told reporters: "Our investigation confirms the FCC has not issued any restrictive documents in this area. Orders remain robust both domestically and internationally, with normal production and delivery operations."
If confirmed, this would mark the second US restriction on Chinese optical modules in a short period. Following the addition of Zhongji Innolight to the US Department of Defense’s 1260H list in June 2026, this latest move underscores tightening policies. A central question remains: Can the US truly decouple from Chinese optical modules? Through interviews with listed companies, analysis of broker reports, and corporate filings, this article maps the industry’s factual landscape.
Industry Reality: Chinese Dominance in Manufacturing, Inevitable Overseas Dependence
Strong demand growth drives the sector. Eoptolink Technology’s 2025 annual report noted that rapid AI development in North America, with four major cloud providers increasing capital expenditures by over 40% year-on-year to surpass USD 360 billion in 2025 for AI data centers. Zhongji Innolight reported that CSP providers globally are boosting capex to meet surging inference and training computing demands. In Q4 2025 alone, Microsoft, Amazon, Meta, and Google saw combined capex rise 64% year-on-year to USD 118.6 billion. According to FactSet consensus estimates as of January 31, 2026, these four cloud providers’ total capex may grow 53% year-on-year to USD 570.8 billion. TrendForce forecasts that eight global cloud service providers will invest USD 602 billion in 2026, a 40% increase, sustaining a boom cycle for optical module production lines.
Demand directly impacts top performers. Zhongji Innolight reported 2025 revenue of CNY 38.24 billion, up 60.25%, and net profit of CNY 10.797 billion, up 108.78%. Eoptolink Technology posted 2025 revenue of CNY 24.84 billion, forecasting H1 2026 net profit between CNY 7 billion and 8 billion, representing 77.56% to 102.93% growth. TFC Communication’s 2026 H1 performance forecast shows net profit between CNY 1.124 billion and 1.304 billion, up 25% to 45% year-on-year. A banker noted that global high-speed optical modules are overwhelmingly supplied by Chinese manufacturers, making it impossible for overseas peers alone to meet North American demand. Institutional analysts deemed the rumored policy "difficult to implement," echoing this view. According to them, fully replacing existing optical modules would be prohibitively costly. Light Counting’s 2025 global top ten optical module companies list includes Zhongji Innolight, Eoptolink Technology, Accelink Technologies, Nazhen Technology (a Hisense subsidiary), HG Genuine, Source Photonics, and Cambridge Industries—seven Chinese firms in total. To mitigate geopolitical risks, leading players have globalized over the past five to six years. For instance, TFC Communication’s Thailand Plant A building passed customer validation in 2024 for passive products, with Plant B delivered in 2025. Eoptolink Technology’s Thailand Plant Phase II has been expanding since early 2025, accelerating in 2026.
US Replacement Capability: Strong but Insufficient
Upstream, China faces bottlenecks in chips and materials. Key components in the optical module supply chain remain dominated by overseas firms. A leading manufacturer disclosed that its products target North America and Europe, with some critical raw materials sourced abroad. Significant changes in exchange rates or trade policies could reduce demand for optical modules and increase procurement difficulty. Addressing tight supply for 200G EML materials, TFC Communication noted in July 9 investor relations activities that shortages are gradually easing, with ongoing supplier engagement for more deliveries. Despite these challenges, the US has some replacement capability. Frost & Sullivan predicts the global electronic measurement instrument market could exceed CNY 200 billion by 2029, with a CAGR of 8% from 2025 to 2029, while China’s market may surpass CNY 80 billion. However, mid-to-high-end testing instruments are dominated by overseas players like Keysight, Tektronix, and Rohde & Schwarz, with low domestic substitution rates. In upstream optical chips, Coherent CEO Jim Anderson stated in a May 2026 conference call that orders are experiencing "step-function" growth, with backlog hitting record highs. Meanwhile, 6-inch indium phosphide capacity upgrades are progressing ahead of schedule, achieving double capacity a quarter early, and 1.6T optical modules are accelerating rapidly. On March 2, NVIDIA announced two major CPO collaboration projects, investing USD 2 billion each in Lumentum and Coherent, with substantial purchase commitments and future capacity rights. US manufacturers still face structural weaknesses in supply. According to Dongwu Securities’ June 25 report, 2026 global optical module shipments could reach 70 million units, with over 52 million units at 800G and above, including 41.57 million units at 800G and 11.19 million units at 1.6T. The localization rate for high-end optical module testing instruments in the US is below 20%. Upstream, US firms like Coherent and Lumentum have deep technical capabilities in optical chips, receiving large investments from NVIDIA. However, building manufacturing capacity from chips to complete modules remains a significant short-term gap. As analysts point out, a policy can close the import door with a single document, but capacity ramping, yield improvement, and customer validation all take time. In this market reality, leading Chinese players like Zhongji Innolight source most North American shipments from overseas capacity, providing ample room for maneuvering even under extreme scenarios.
Decoupling Stress Test: Impossible Short-Term, Costly Mid-Term
Based on industry realities, the timeline for US-China decoupling in optical modules breaks into three layers. Zhongji Innolight’s 2025 annual report notes that suppliers typically require customer qualification and product code certification, involving site visits and audits. The certification cycle makes decoupling virtually impossible in the short term (2026). Industry data shows that switching optical module suppliers takes at least 18 months, with replacement costs 30% higher. On August 5 morning, market analysts cited by The Paper noted that after tariff hikes in 2025, optical modules were among the first items exempted due to opposition from US cloud providers, suggesting any future policies would include many exemptions, keeping the impact manageable. The banker and institutional analysts agree that Chinese leaders significantly outperform overseas peers in technology, product progress, cost control, and capacity layout, making a short-term gap unfillable. Full decoupling would only delay North American data center builds, harming larger interests. If short-term decoupling is difficult, mid-term partial substitution faces feasibility but at high costs. According to CITIC Construction Investment, 1.6T optical modules began shipping in 2025, with volume growth expected in 2026, and 800G modules maintained high growth from 2024 to 2026. Zhongji Innolight’s 2025 report states that demand for high-speed modules like 800G and 1.6T will dominate the market over the next three years. Light Counting predicts that 800G and 1.6T module shipments will surge in 2026, with combined market size reaching USD 14.6 billion, about 64% of the total. With NVIDIA advancing CPO technology, ASE data shows CPO solutions can reduce power consumption by 60% and costs by over 30% compared to traditional pluggable modules, with large-scale deployment possibly starting in 2028. Guotai Haitong Communications team believes that the optical communication industry relies heavily on the global ecosystem. Chinese firms benefit, but upstream US chip and component companies also capture significant profits, and downstream cloud providers gain critical support. Blanket restrictions would harm globalization, making such a scenario unlikely. Beyond the impossibility of short-term and cost of mid-term, the long-term logic may shift. Goldman Sachs reports that optical modules are moving toward silicon photonics, with 1.6T silicon photonics penetration at 80% and 3.2T fully using silicon photonics, with silicon photonics revenue share reaching 62% by 2028. Zhongji Innolight indicates that 3.2T modules may begin ramping from 2028. Dongwu Securities notes that given the fragmented nature of testing instrument sub-markets and high R&D costs, there are usually no mature commercial chips available, forcing self-development. The surge in AI infrastructure demand drives volume and price growth for matched testing instruments. Due to the specialty of scientific instruments, expansion is difficult, and foreign leaders struggle to meet incremental demand, creating opportunities for domestic leaders with technology and channel reserves. Returning to the present, this rumor serves more as another stress test than a ban preview. For China’s optical module industry, every rumor acts as a stress test and a catalyst for supply chain consolidation. The real challenge lies not in whether a single ban can be implemented, but in how to accelerate upstream core technology self-reliance amid policy uncertainty, turning "can't decouple" dependency into "indispensable" competitiveness.
Comments