The potential decision by the Federal Communications Commission (FCC) regarding high-speed optical modules has once again drawn attention to the supply chain underpinning every AI data center. According to LightCounting's global ranking of optical module suppliers, seven Chinese companies have entered the top ten, with Innolight and Eoptolink holding the first and second positions, respectively.
In the 400G, 800G, and 1.6T segments, which are essential for hyperscalers to train and infer AI, Chinese manufacturers have moved beyond being mere competitors—they constitute the majority of supplies.
The position held by these companies has become structurally difficult to change. Senior analyst Liu Minyang of Omdia asserts that even as American capacities such as Coherent, Lumentum, Nokia, AOI, and Corning expand, the base of optical communication production and supply in China and the Asia-Pacific region remains virtually indispensable in the foreseeable future. The division of labor has reached maturity: Chinese suppliers handle large-scale production of modules, packaging, testing, and rapid delivery, while high-speed DSP chips and analog chips primarily come from Marvell and Broadcom, and premium EML lasers and parts of the silicon-photonics stack come from Coherent and Lumentum.
The bottleneck lies further down the technological chain. Module manufacturers, including Accelink, Source Photonics, HGTECH, and CIG, complete the upper echelon of Chinese companies. Innolight, Eoptolink, and TFC have significant manufacturing capacity in Thailand; DSBJ utilizes 69.7 percent of its capacity in mainland China, with the remainder in Taiwan, routing products intended for the US through Thai and Taiwanese facilities, aiming for a target of 35 million units of 800G and 1.6T by 2027.
The next source of tension is indium phosphide. This substrate is crucial for 400G, 800G, and 1.6T modules. Industry estimates suggest demand for 2026 will be between 2.6 and 3 million wafers with an effective capacity of about 750,000, and prices for 2-inch InP optical substrates have sharply risen this year. EML lasers, silicon-photonic components, DSP chips, and InP substrates have become four main choke points that everyone is trying to pre-book.
The race extends beyond just modules. Ciena reported a 40 percent revenue increase in its second fiscal quarter, with an order backlog reaching $7.7 billion extending to 2027; Nokia's optical network segment showed a 20 percent year-over-year growth in the second quarter. In China, Huawei, ZTE, and FiberHome dominate the domestic market for backbone networks and data center interconnects, where AI-OTN is being upgraded from single-wavelength 400G to 800G and 1.6T. China Mobile and China Telecom have already launched new tenders for 400G and 800G optical transmission equipment for OTN.
As AI shifts from model competition to infrastructure competition, optical communication ceases to be a single product line. It is a complex system encompassing pluggable modules, switches, optical chips, coherent optics, CPO, DCI, and backbone transmission. Whoever controls the top-tier raw materials, core devices, and long-term supply plan will determine the winner in the next decade of AI infrastructure.
Artificial intelligence has become one of China's new export directions, and the country is leading this race by developing more efficient, low-cost, and open-source models.
There is a trend where American companies are beginning to switch to Chinese AI solutions. Furthermore, the rest of the world is moving in the same direction.
Major Chinese internet companies are actively implementing robotaxis, delivery drones, warehouse robots, and embodied AI agents across transportation, logistics, manufacturing, and service sectors.
These companies are bringing robots, autonomous vehicles, drones, and embodied AI to a mass level. For instance, Meituan's delivery robots have completed millions of autonomous deliveries, while JD.com has modernized its logistics using warehouse robots. ByteDance has invested in robotics startups, and Baidu's Apollo robotaxis have covered tens of millions of kilometers of autonomous driving.
This structural transition is driven by three interconnected factors. Firstly, business models initially based on digital technologies face growth limitations in pure internet markets, forcing companies to seek opportunities to create value in the physical economy, where the level of digitalization remains lower. Secondly, advancements in AI perception, decision-making, and actuation have made real-world automation technically and economically feasible on a large scale for the first time. Thirdly, the competitive advantage of internet companies—vast data sets, user engagement platforms, and capital resources—naturally transfers to physical world applications, where data-driven optimization and economies of scale are equally applicable.
Companies are not just using technology; they are building vertically integrated stacks of physical AI—from sensors and computing hardware to algorithms and operational platforms. Meituan demonstrates the most advanced example by creating a complete physical AI infrastructure that covers autonomous delivery vehicles, drones, in-store service robots, and fleet management cloud systems. Meituan's robots can navigate complex urban environments, including pedestrian zones, intersections, and building entrances, without human intervention.
JD.com has integrated robotics throughout its supply chain, utilizing autonomous forklifts in warehouses and AI-optimized routing for trucks. ByteDance maintains a distributed strategy, investing through its venture fund and internal R&D in startups developing humanoid robots, collaborative robots, and specialized industrial automation systems.
Beyond these leaders, Baidu offers Apollo autonomous driving, embodied AI, and Kunlun chips. Cainiao logistics robots from Alibaba and DAMO research are also mentioned, while Tencent Robotics X focuses on embodied AI. This drive is comprehensive across all major internet companies. Companies combining digital intelligence with physical operations are gaining cross-domain value. Traditional industries are feeling pressure from AI-focused companies but are simultaneously discovering partnership opportunities. The expansion of the Chinese internet into the physical world has global implications, as the combination of AI, production scale, and deployment data creates new competitive advantages.
