China’s optical chip sector has been gaining momentum recently, with supportive signals from policy, industry and capital markets. In June 2026, the Ministry of Industry and Information Technology formally issued the Implementation Opinions on Innovative Development of “AI + Information and Communications,” calling for stronger R&D in core technologies including high-end optoelectronic chips, CPO and all-optical switching. Overseas giants such as NVIDIA and Meta have committed billions of dollars to secure optical interconnect capacity. Falling prices and rising shipments of 800G optical modules, together with pilot production of 1.6T modules, are helping meet the bandwidth demands of AI computing power networks...

What is worth noting is that this round of industry tailwinds is not short-term concept speculation, but a direct reflection of substantive changes in the sector’s fundamentals. According to statistics, global optical communications sales reached $36.8 billion in 2025 and are expected to rise to $39 billion in 2026, with the market continuing to expand. The surge in AI computing power has changed the role of optical chips, upgrading them from traditional “communications pipes” into a “strategic resource” that determines the ceiling of computing power.

AI Reshapes Industry Logic as Supply and Demand Shift

The explosion in AI computing power is the core engine driving demand for optical chips. Cloud-based agents and large models are now often deployed on server clusters with 10,000 GPUs, 100,000 GPUs or even larger scale, placing extremely high requirements on data transmission speed, latency and stability. Traditional low- and mid-speed optical chips and related modules represented by 10G, 25G and 100G can no longer keep up with the operating needs of computing power clusters. High-speed 800G and 1.6T optical modules have become a must-have across the industry.

“The rise in the optical communications sector is not short-term thematic speculation. It is based on a substantive change in industry fundamentals. AI computing power demand is upgrading optical communications from a traditional ‘communications pipe’ into a ‘strategic resource’ that determines the ceiling of computing power. Its impact is not only a short-term demand boost, but points to a long-term and far-reaching restructuring of the ecosystem,” Peng Lingyong, chairman of optical communications VCSEL supplier Huaxin Semiconductor Technology Co., Ltd., told China Electronics News.

According to LightCounting, 800G and 1.6T optical modules will see rapid volume growth in 2026, with their combined market size expected to reach $14.6 billion, or about 64% of the overall optical module market. By shipments, global sales of 800G optical modules are expected to reach 33.5 million units in 2026, while demand for 1.6T modules will reach 8.6 million to 20 million units. Their scarcity could even exceed that of today’s hot HBM, or high-bandwidth memory, chips.

“In the short term, the need to relieve communications bottlenecks is driving scaled commercial use of 1.6T optical modules in 2026, giving the industrial chain extremely strong earnings visibility. Over the long term, the technology is evolving toward ‘optical communications 3.0.’ The future will move toward co-packaged optics, or CPO, to break through power consumption limits, and optical circuit switching, or OCS, to restructure network architecture. Copper cables have already entered the countdown to phase-out in AI high-speed interconnects,” Peng said.

Rapidly rising demand has made supply-side weaknesses increasingly visible. On one hand, the number of optical modules required by AI clusters is usually three to five times that of ordinary data centers. On the other, nearly half of China’s backbone network equipment has already been upgraded to 800G, while major computing power hubs and leading AI data centers have largely adopted 800G networks. With North American cloud providers, NVIDIA and other companies placing large long-term orders extending to 2028, demand at home and abroad is rising in tandem, and related domestic capacity has been running at full load.

“The optical communications industry is now seeing market expansion and capacity shortages at the same time. Goldman Sachs predicts the optical interconnect market will expand from $15 billion to $154 billion within two to three years, but supplies of upstream core materials such as high-end EML optical chips and indium phosphide substrates are tight, with supply-demand gaps of more than 30% and 70%, respectively,” Peng said, describing the state of the industry.

More specifically, this supply tightness is not simply a short-term capacity allocation issue, but a structural problem formed over time. Take indium phosphide wafers, which are essential for producing optical chips. Investment in a single 6-inch wafer production line exceeds 1 billion yuan. The cost is high, and the construction and expansion cycles are also very long. Low mass-production yields for high-end chips and the need to improve optoelectronic co-design capabilities have further widened the market gap.

This industry reality also gives the local supply chain valuable room to grow and iterate. In the short term, the industry’s core task is to digest a huge volume of orders and ease supply pressure. Over the long run, the entire sector will continue upgrading toward lower power consumption, higher bandwidth and greater integration. The domestic supply chain is also expected to use this round of industry change to steadily close gaps, refine its technologies and keep improving its competitiveness.

A Clearly Tiered Domestic Supply Chain Enters an Upmarket Window

Continued changes in market supply and demand are forcing Chinese optical chip companies to sharpen their capabilities. After years of technical accumulation and market testing, domestic optical chip products have formed a clear tiered structure: low- and mid-end products have secured a foothold in the market, while high-end products are still in an accelerated catch-up phase. The industry is now entering a valuable window to break into higher-end segments.

From actual applications and market performance, the product tiers are easy to see.

For low-speed optical chips used in conventional communications scenarios, domestic technology, capacity and supply chains have become highly mature and can fully meet market demand.

For 10G optical chips, which have broader applications, domestic products hold a considerable market share, and their bulk delivery capabilities have been tested over a long period. They operate stably and reliably.

High-speed optical chips at 25G and above, focused on AI computing power clusters and high-speed interconnect scenarios, are now the industry’s main area of effort. These chips are also core hardware for supporting GPU clusters at the 10,000-card scale. At this stage, domestic products account for only 4% to 5% of the market, leaving significant room for improvement.

In response to the online claim that “the overall localization rate for optical chips has reached 70%,” Peng offered an objective clarification: “That usually comes from including low- and mid-end products and passive components in the calculation. It does not apply to the high-end active chip scenarios required by AI.”

As he put it, the high-end track is the core direction for the domestic industry’s next concentrated push.

The industry is currently facing a series of systemic challenges. Beyond the need to strengthen high-end chip design and manufacturing capabilities, the sector also commonly faces tight high-end capacity, insufficient mass-production yields and constrained supplies of core materials. At the same time, heavy reliance on overseas suppliers for key supporting components such as upstream DSP electrical chips has also become a major obstacle to the entire supply chain’s move upmarket.

Overall, China’s optical chip industry today has a solid foundation, remaining weak links and clear opportunities ahead. Low- and mid-end products have secured the market base. High-end products still lag, but with support from market demand, policy backing and technological innovation, the conditions for an upward breakthrough are now in place. The years 2026 and 2027 will also be a golden period for the domestic industry to sprint into the high-end track.

Frontier Technologies Await Breakthroughs as Multiple Routes Define the Future

For China’s optical chip industry to truly establish itself in the high-end market, the decisive competition will be in technological innovation. The industry has now identified three mainstream development routes. Domestic companies and research teams are advancing in parallel, making steady progress in new materials, device architectures and integrated packaging. These varied technical directions are not only addressing today’s industry problems, but also clearly outlining where the sector is headed.

Thin-Film Lithium Niobate: A Next-Generation Chip Material

As transmission speeds in AI servers keep rising, traditional chip materials are gradually approaching their performance ceilings, with slower transmission and higher power consumption becoming more prominent problems. Against this backdrop, thin-film lithium niobate has become widely recognized in the industry as a next-generation premium material, and it is also a key focus for domestic players.

Discussing the technical advantages of thin-film lithium niobate, Yang Zhiwei, senior product director at TuringQ, told China Electronics News: “For silicon photonics-based optical communications, the bandwidth ceiling is 60GHz. To reach 100G, heterogeneous integration is required, and that is still at the laboratory stage. Thin-film lithium niobate can easily exceed 100GHz, which is one of its natural advantages. It also has lower optical insertion loss, higher modulation efficiency and lower driving power consumption, making it better suited to the ultra-high-speed, low-power optical interconnect needs of high-speed computing power clusters.”

Beyond its technical advantages, thin-film lithium niobate’s cost benefits will gradually be released as 8-inch thin-film lithium niobate wafers enter mass production. To accelerate commercialization, national innovation platforms, universities, research institutes and backbone companies are jointly building public foundry and pilot production platforms, establishing unified process standards and helping the industry improve yields so more companies can participate in R&D and production. The industry has also split into two R&D approaches: some teams focus on standalone chip products, while others are trying to combine the material with mature silicon to draw on the strengths of both.

The industry sees 2026 as the first year of mass production for thin-film lithium niobate. At the Optical Fiber Communication Conference, or OFC, held in March this year, a consensus emerged: silicon photonics and indium phosphide can meet demand for 800G and lower optical modules, but 1.6T and 3.2T high-speed optical modules and CPO co-packaging technology cannot do without thin-film lithium niobate. In the past, it was used only for experiments and was difficult to mass-produce. Now, processes have improved sharply, and products are entering the market in batches. By 2027 to 2028, 3.2T optical module products are expected to become more widely adopted, and thin-film lithium niobate will become an industry standard, fully unlocking its commercial value.

LPO: Simplifying Device Structures to Forge a Distinctive Path

The core idea behind LPO technology is subtraction. It retains the form factor of traditional pluggable optical modules, but removes the most power-hungry and expensive DSP chip inside the module, using a simpler combination to complete data transmission. It features low power consumption, low cost, low latency and easy maintenance.

LPO technology has several advantages. First, low power consumption. According to data from semiconductor product supplier MACOM, the power consumption of an 800G multimode optical module with DSP functionality can exceed 13W. An 800G multimode optical module using MACOM PURE DRIVE technology consumes less than 4W, a reduction of 70%. Second, low cost. Some industry institutions have analyzed that in an 800G optical module, the bill-of-materials cost is about $600 to $700, while the DSP chip costs about $50 to $70, so removing the chip can effectively lower costs. Third, low latency. Without the DSP, one processing step is removed, reducing data transmission latency, which is especially important for AI computing and supercomputing scenarios. Fourth, easy maintenance. LPO packaging does not change significantly, supports hot swapping, simplifies fiber cabling and equipment maintenance, and is more convenient to use.

In terms of use cases, LPO is better suited to short-distance data transmission inside server cabinets and data center clusters, generally within about 500 meters. That happens to be the core use case for today’s AI computing power clusters. With its streamlined architecture and lower power consumption, LPO technology has already been deployed at scale in multiple domestic AI data centers and major internet companies, and has become a notable differentiator for China’s optical chip industry.

CPO: Chip Integration Becomes the Mainstream Evolution Path

Optical interconnect technology is key to breaking through bandwidth bottlenecks in computing power clusters. Among these technologies, co-packaged optics, or CPO, is seen as the next-generation mainstream technology. Compared with pluggable optical modules and near-packaged optics, or NPO, CPO directly integrates the optical engine and computing chip on the same package substrate, sharply increasing physical integration and compressing electrical signal transmission to the millimeter level, improving signal integrity while reducing power consumption.

Major global technology companies are now investing heavily in this technology, and China is moving closely behind, with R&D mainly focused on glass substrate packaging, microchips and heterogeneous integration. According to LightCounting, the CPO market is expected to reach $10 billion in 2030 and grow rapidly from 2027 to 2030. The core driver of that high growth is the AI computing power arms race, which is forcing breakthroughs in power consumption and bandwidth limits. That is being reinforced by advances in packaging yields and processes, close ties between cloud providers and leading customers, and localization of the industrial chain, forming a closed loop from demand and supply to the broader ecosystem.

The three technology routes have a clear division of labor, forming a complete layout that balances short- and long-term needs. In the short term, LPO and thin-film lithium niobate are being used to address current commercial demand. Over the medium and long term, the industry will focus on advancing CPO technology to target future markets. Continued refinement of chip materials and precision production processes remains the basic guarantee for all of these technologies to be implemented.

Looking across the optical chip industry, the AI wave has brought not just a short-term surge in orders, but a comprehensive reshaping of supply-demand structures, product systems and technology routes. The years 2026 and 2027 will be an inflection point for the adoption of 1.6T optical modules and CPO technology, as well as a golden window for domestic high-end optical chips to make a leap forward. Building on the broader shift toward optical communications 3.0, China’s optical chip supply chain will continue to close gaps and strengthen its advantages, moving steadily ahead as it integrates more deeply into the global industrial system and creates new value through this long-term industrial transformation.