The U.S. Federal Communications Commission (FCC) officially published its highly anticipated final rule, 'Protecting Against National Security Threats to the Communications Supply Chain Through Equipment Authorization Programs,' in the Federal Register. The rule, approved on July 22, is set for formal publication on September 11 and will take effect 30 days thereafter.

Following the rule's implementation, the most feared scenario by the market did not materialize: Core Chinese optical communications companies, including CIG Photonic, YOFC, Dongshan Precision, and TFC Communication, were not directly placed on the FCC's restricted list.

The controversy did not arise without reason. On July 22, the FCC added optical transceivers to the Cover List, banning the import of equipment containing logic hardware components produced by Chinese 'Covered List Entities.' On August 4, reports emerged about a draft proposal to prohibit the import of new Chinese optical transceiver models. Now, with the rule finalized, the most extreme scenario of a complete blockade has not yet occurred, but the competition over supply chain security and technological substitution continues.

More notably, the optical module industry is undergoing changes far more profound than a mere restricted list: Co-Packaged Optics (CPO) technology is on the verge of mass production, with industry giants like NVIDIA and Broadcom leading a new wave of transformation in data center optical interconnects. NVIDIA's CPO innovation abandons traditional pluggable optical modules, opting instead for silicon photonics devices integrated with ASIC chips. Compared to traditional solutions, this technology can improve device energy efficiency by fivefold, significantly enhance network reliability, and extend continuous operation time for business applications by the same factor.

01 'Shooting Oneself in the Foot'

Although the final rule did not restrict Chinese companies as rumored, this game theory (geopolitical maneuvering) has exposed an unavoidable issue: U.S. AI infrastructure construction is highly dependent on the Chinese optical module supply chain, and a forced blockade would result in mutual harm.

From a global industrial perspective, Chinese manufacturers have achieved absolute dominance in the optical module sector: Seven of the world's top 10 optical module companies are Chinese; in the high-speed optical module markets for 800G and 1.6T, Chinese manufacturers account for over 70% of shipments; CIG Photonic alone holds approximately 27% of the global data center optical transceiver market. While U.S. manufacturers Coherent and Lumentum possess technological competitiveness, their production capacity is far from sufficient to replace Chinese suppliers. Industry consensus holds that overseas optical component suppliers are generally reluctant to engage in businesses with gross margins below 50%, and China's complete optical industry chain and efficient delivery capabilities remain irreplaceable in the short term.

The costs of a forced import ban would be directly borne by U.S. downstream industries. On one hand, procurement costs for U.S. cloud providers like Amazon Web Services (AWS) would rise significantly; on the other, during the current window of accelerated AI infrastructure construction, supply shortages would directly slow down data center construction and deployment. Analysts at Bank of America point out that a broad import ban would not only disrupt existing supply chain order but could also delay the implementation of next-generation data center architectures, running counter to U.S. goals of accelerating AI computing power construction.

Meanwhile, core components such as DSP chips and analog chips for optical modules predominantly come from U.S. companies like Marvell and Broadcom, creating deep interdependence in the industrial chain. A blanket ban would ultimately impact the revenue of U.S. upstream chip manufacturers. Since 2020, the market has repeatedly expressed concerns that Chinese optical module manufacturers' market share could be replaced by North American companies. However, after years of validation, even with sustained U.S. industrial policies to support domestic supply chains, Chinese manufacturers' global market share has continued to grow, demonstrating that industrial law (market forces) far outweigh administrative interventions.

02 Proactive Strategic Layouts by Domestic Optical Module Companies

Notably, Chinese optical module companies have not been caught unprepared. As early as five or six years ago, the industry began preparing for geopolitical factors.

First, they are advancing global production capacity Layout (deployment) to hedge against market access risks. Leading manufacturers have long initiated capacity relocation in response to market access risks. Since 2024, suppliers like CIG Photonic have expanded production capacity in Thailand, Malaysia, and other locations. By 2025, their non-Chinese factories will independently meet U.S. market demand, mitigating origin-based regulatory risks at the manufacturing level. Simultaneously, North American leading cloud providers have engaged in deep joint R&D with Chinese companies on next-generation NPO and Coherent Lite products. The R&D cycles and technical barriers involved cannot be easily replicated, further solidifying bilateral ties.

Second, they are pursuing multi-track layouts in cutting-edge technologies to reduce reliance on a single track (market segment). While current restrictions primarily target pluggable modules, Chinese manufacturers are not confined to traditional pluggable optical modules but are advancing on multiple fronts to secure positions in next-generation technologies. Near-Packaged Optics (NPO), an open ecosystem primarily driven by cloud providers, has seen domestic manufacturers establish significant leads. In April 2026, Google placed orders for 12 million NPO optical modules for its next-generation TPU v7/v8/v9 supercomputing clusters, with CIG Photonic and YOFC securing 60% and 40% of the shares, respectively, capturing the entire order. In CPO, leading companies like CIG Photonic and YOFC have clearly articulated strategies to transition from 'selling modules' to 'selling optical engines.' CIG Photonic has developed silicon photonics chip technology and is advancing 3.2T optical engine testing, aiming to become an indispensable optical component supplier in the CPO ecosystem. YOFC has launched 1.6T optical module products covering all technical routes, including VCSEL/EML, silicon photonics, and thin-film lithium niobate, while simultaneously deploying high-speed LPO and CPO-related products.

Third, they are upgrading business models by establishing overseas joint venture platforms to break through geopolitical barriers and access global high-end supply chains. The joint venture between TFC Communication and SuperX exemplifies this approach. On April 21, 2026, TFC Communication's wholly-owned subsidiary, Singapore TFC, signed an agreement with NASDAQ-listed SuperX and APEX VERVE LIMITED to jointly invest SGD 2 million in establishing SuperX Optical Communications Pte. Ltd. in Singapore, with SuperX holding 45%, Singapore TFC holding 35%, and APEX VERVE LIMITED holding 20%. The joint venture completed business registration and full capital contribution on July 1, 2026. This structure serves three purposes: (1) circumventing geopolitical restrictions by registering in Singapore and covering global markets outside mainland China, Hong Kong, and Macau, directly bypassing the FCC's Cover List restrictions on mainland Chinese entities; (2) deeply binding with the core CPO ecosystem, as TFC Communication was publicly listed as a core partner in NVIDIA's Quantum-X and Spectrum-X silicon photonics CPO switches announced at the 2025 GTC Conference, alongside TSMC, Coherent, and Corning; and (3) achieving industrial upgrading, as TFC Communication transitions from an upstream passive component supplier to directly participating in the global sales of silicon photonics modules and optical engines through the joint venture platform, entering core links of the industrial chain.

03 Reverse Leverage at the Supply Chain Level

Beyond proactive layouts by optical module companies themselves, China also possesses inherent industrial leverage in the upstream raw materials for optical chips. Indium phosphide (InP), widely used in DFB lasers, EML lasers, and photodetectors, is an essential material for 800G/1.6T and even next-generation 3.2T optical modules. The production capacity for high-purity indium metal required for InP optical chip manufacturing is largely concentrated in China, with even U.S. company AXT directly locating its InP factory in China. AXT's subsidiary, Beijing Tongmei, alone accounts for approximately 36% of the global InP substrate market.

Moreover, InP is currently facing severe supply-demand mismatches. As optical module speeds advance to 1.6T and 3.2T, AI data center demand for InP substrates has surged. Data shows that the global InP substrate supply-demand gap exceeded 70% in 2025, with this tightness expected to persist until 2027. Consequently, InP prices have risen significantly this year, with average prices for 4-inch substrates increasing by about 50% since the beginning of the year and upstream refined indium prices surging by over 80%. NVIDIA has requested suppliers to expand InP laser production capacity to 20 times current levels by 2030.

In January 2026, China's Ministry of Commerce announced a complete ban on the export of dual-use items (including InP, indium, gallium, and germanium) to Japanese military users and applications, with civilian exports subject to strict licensing and end-user reviews. This creates a reverse leverage dynamic in industrial game theory (competition): If the U.S. bans Chinese optical module imports, China could retaliate by imposing export controls on upstream critical materials. Without Chinese indium supplies, U.S. domestic optical chip manufacturers would halt production, further reducing the feasibility of extreme import bans.

04 The Real Crisis for Domestic Optical Module Companies

Objectively, the impact of external bans is relatively limited. The real crisis for domestic optical module manufacturers stems from technological iteration. In August 2026, Gilad Shainer, NVIDIA's Senior Vice President of Networking, officially announced that Co-Packaged Optics (CPO) had entered mass production. Spectrum-X CPO switches have been delivered to core customers and deployed in NVIDIA's proprietary AI factories.

Currently, expected regulatory targets focus on pluggable optical modules, which happen to be the strong suit of domestic companies. In the short term, pluggable optical modules will not be completely replaced by CPO, but the medium-to-long-term industrial evolution direction is clear. It must be recognized that in the CPO era, core industrial chain discourse power (influence) lies with companies possessing full-stack 'optoelectronic integration' manufacturing capabilities, where domestic industries still have significant shortcomings.

Overseas, TSMC leads in Co-Packaged Optics (CPO) through its COUPE platform. Leveraging advanced process capabilities, it focuses on high-end silicon photonics (PIC) integration processes, emphasizing 3D stacking and optoelectronic co-packaging technology routes. Deeply bound with top clients like NVIDIA and Broadcom, TSMC is the core high-end silicon photonics foundry supplier for next-generation CPO architectures, with its photonics integrated circuit capacity rapidly expanding alongside AI computing demand.

GlobalFoundries has become one of the largest pure silicon photonics foundry players by acquiring Singapore's AMF. Its Fotonix platform specializes in data center and AI optical interconnects, with dual production capacities in the U.S. and Singapore. With over a decade of silicon photonics technology Layout (deployment), it covers full technical routes for pluggable optical transceivers and CPO devices. In 2026, it received USD 300 million in U.S. government funding to support next-generation silicon photonics R&D, targeting single-channel rate increases to 400Gb/s, making it a core carrier of the U.S. domestic silicon photonics supply chain.

UMC focuses on mature process silicon photonics solutions. Allying with IMEC, a global semiconductor R&D center, UMC handles front-end wafer foundry processes using a 28-22nm process platform, with trial production expected in 2026 and volume production in 2027, marking a key step in shifting mature processes toward high-value applications. On July 14, 2026, UMC and Singaporean photonic chip company SILITH Technology jointly announced that UMC's Singapore 12-inch wafer fab, Fab 12i, had completed its first batch of silicon photonics wafer mass production deliveries, with their 1.6T silicon photonics platform transitioning from technology development to commercial volume production.

Tower Semiconductor, an Israel-based wafer foundry with factories in the U.S. and Japan, differs from foundry giants like TSMC and Samsung, which focus on advanced logic chips. Tower specializes in high-value analog and mixed-signal manufacturing, with core differentiators in specialty processes like silicon photonics (SiPho) and silicon-germanium (SiGe). In February 2026, Tower officially announced a collaboration with NVIDIA to provide silicon photonics platform support for its 1.6T data center optical modules. This partnership signifies global recognition of Tower's technological prowess in AI optical interconnects by the world's leading AI chip company.

In contrast, domestic companies still face significant gaps in silicon photonics foundry capabilities.

05 Conclusion

Overall, domestic optical modules are at a critical juncture where geopolitical pressures intersect with next-generation technological transformations. On September 11, 2026, the FCC's final rule on related equipment officially took effect but did not restrict Chinese companies as previously rumored, nor did it impose a complete ban on Chinese optical modules. External pressures have not yet escalated into extreme bans, and domestic leading companies have proactively hedged against geopolitical risks through overseas capacity relocation, diversified cutting-edge technology Layout (deployment), and overseas joint venture platforms, solidifying their advantageous positions in the global high-speed pluggable optical module market.

External risks are not the industry's greatest threat. As CPO technology nears mass production, the optical interconnect industry is about to shift tracks. Future industrial competition will center on full-stack optoelectronic integration manufacturing capabilities, with overseas foundries like TSMC, GlobalFoundries, and UMC having already completed technological and client preparations in silicon photonics foundry services. In contrast, domestic silicon photonics foundry capabilities remain a prominent weakness. For domestic optical module manufacturers, relying solely on the scale advantages of traditional pluggable optical modules is unsustainable. While defending their existing global market positions, accelerating the development of foundational process weaknesses like silicon photonics foundry and transitioning from optical module suppliers to optical engine and optoelectronic integration solution providers will be key to seizing the initiative in the next-generation AI optical interconnect era.