In the past, discussions on AI computing power focused entirely on GPUs and servers, but in 2026, the landscape has completely shifted.
From top-level documents issued by the Ministry of Industry and Information Technology (MIIT) and the National Development and Reform Commission (NDRC), to a series of support plans in Guangdong Province and Shenzhen City, a succession of major policies has been intensively rolled out. An inconspicuous small component hidden deep inside the racks—the optical module, once regarded merely as a "communication accessory"—has been officially elevated to the strategic foundation for AI computing power and 6G networks.
Three Hardcode Numbers: Understanding the True Signals of the Policies
Many people interpret policies simply as "benefits, subsidies, and industry prosperity."
However, looking beyond the text, there are three irreversible industrial red lines hidden within.
「✅ Number One: 800G/1.6T Becomes the Standard for Newly Built 10,000-GPU AI Computing Centers」
The policies explicitly require that newly built 10,000-GPU intelligent computing clusters in China must be equipped with 800G and 1.6T high-speed optical modules as standard, while 400G low-speed modules will be gradually phased out of new procurement lists.
In simple terms:
In the past, speed upgrades relied on slow iteration driven by market demand; now, policies have set a clear timetable for upgrades.
Large-scale computing infrastructure in China no longer leaves room for low-speed products. 800G is the current rigid demand, 1.6T has shifted from a "future direction" to a task for scaled deployment, and 3.2T has entered the track of technological reserves.
For manufacturers still sticking to 400G capacity, the ceiling in the domestic computing power market is slowly lowering.
「✅ Number Two: Localization Rate of National-Level Intelligent Computing Clusters ≥ 70% 」
This is the most easily overlooked yet most profoundly impactful rule.
A 70% localization rate does not mean simply assembling domestic enclosures. The real bottleneck lies in high-speed optical chips.
Currently, many domestic module manufacturers still rely on overseas procurement for laser chips. If upstream chips cannot be independently controlled, it will be difficult to cross this localization threshold even if the final products are assembled domestically.
The industrial profit landscape will also undergo a major reshuffle.
In the past, dividends were concentrated in mid-stream optical module manufacturing; in the future, the focus of value will shift upstream: Indium Phosphide (InP) substrates, high-speed EML (Electro-absorption Modulated Laser) optical chips, and Thin-Film Lithium Niobate (TFLN) materials will become the main battlegrounds for industrial chain breakthroughs in the next 3-5 years. The National Indium Phosphide Special Fund is precisely targeting this shortcoming to strengthen the supply chain.
「✅ Number Three: Target of 45% Self-Sufficiency Rate for High-End Optical Chips by 2028 」
This is not just a slogan, but a task document with specific time nodes for critical breakthroughs.
High-speed optical chips have always been the weakest link in the domestic computing power supply chain. The clear targets set by the policies mean that over the next two years, R&D subsidies, demonstration projects, and computing power procurement orders will all lean towards enterprises with self-developed optical chips.
Three Major Technical Routes Locked In: No Multiple Choices for the Track
Reading through the full set of policies for 2026, it is not hard to see that the national technological direction is already very clear: CPO (Co-Packaged Optics), LPO (Linear-drive Pluggable Optics), and NPO (Near-Packaged Optics) for low-power optoelectronic interconnects, with silicon photonics modules as the long-term main focus.
Why are policies strongly supporting this route?
There is an unavoidable physical challenge in the development of computing power.
As optical module speeds advance from 800G to 1.6T and 3.2T, the power consumption of traditional pluggable optical modules will rise sharply. The interconnection of 10,000-GPU clusters brings massive energy consumption pressure due to the huge number of optical modules, leading to a steep increase in the pressure of Data Center PUE (Power Usage Effectiveness) management.
Short-term (1-2 years): 1.6T pluggable optical modules will remain the main force for market delivery;
Medium to long-term (2-4 years): CPO (Co-Packaged Optics) solutions will gradually enter large-scale intelligent computing centers to reduce the power consumption of computing cluster interconnections.
In short: pluggable is the present, and co-packaged is the future.
Enterprises that now deploy R&D in CPO, silicon photonics, and lithium niobate will significantly increase their chances of securing local R&D subsidies and demonstration projects.
Three-Tier Policy Division: Dividends Landing Layer by Layer
Many people wonder what the exact relationship is among national, provincial, and municipal policies. In fact, the division of labor among the three is very clear, forming a complete system of "top-level goals - supply chain shortboard supplementation - project support."
National Level: Setting Standards, Establishing Thresholds, and Overcoming Choke Points
Introducing hard rules for computing power procurement, localization red lines, and chip self-sufficiency targets, and establishing a special fund for the indium phosphide industry, leaning towards constraints and top-level breakthroughs.
Guangdong Province Level: Deploying Long-Term Tracks and Making Up for Supply Chain Shortboards (2026-2035)
Incorporating optical communications, optical chips, and optical computing into the province's new track planning. Continuing the optical chip support plan, issuing R&D rewards for indium phosphide and optical chip tape-out projects, and creating demonstration scenarios for domestic high-speed optical modules.
Shenzhen City Level: Substantial Financial Support for Mass Production and Deployment
Launching the AI server industry chain action plan, encouraging the mass production of 800G-1.6T, providing up to CNY 10 million in subsidies for CPO R&D projects, with a maximum R&D funding limit of CNY 30 million per project. The goal is to build a nationally leading complete local industry chain cluster of "servers - optical modules - optical chips."
Differentiation of Industry Landscape
Under the wave of policies, structural differentiation in the industry has already begun.
✦ Enterprises Welcoming the Dividend Window
Mid-stream module manufacturers that have completed 800G mass production, achieved smooth capacity ramp-up for 1.6T products, and proactively deployed 3.2T R&D;
IDM (Integrated Device Manufacturer) enterprises with in-house R&D capabilities for high-speed EML optical chips, capable of meeting the localization procurement threshold;
R&D teams that have proactively deployed new technical routes such as CPO, silicon photonics, and thin-film lithium niobate.
✦ Enterprises Facing Direct Pressure
Those with capacity concentrated in 400G and below low-speed modules, lagging in high-speed product iteration;
Those completely relying on external procurement for high-end chips, lacking alternative domestic supply chain options;
Those with insufficient R&D investment in the next-generation low-power interconnect track.
At the same time, the market is also divided into two tracks. In the domestic computing power market, the home-court advantage of domestic manufacturers is becoming increasingly obvious; the overseas market remains a fully competitive global arena. The growth path of "domestic + overseas dual circulation" for leading enterprises is becoming clearer.
Finally, Let's Talk About the Risks Behind the Policy Dividends
Dividends do not equal easy wins; we must also see the hidden boundaries.
1. The 70% localization threshold only applies to national-level intelligent computing clusters. Commercial self-built data centers and overseas orders are not subject to this mandatory rule.
2. Government project breakthroughs and demonstration rewards ≠ guaranteed orders. Whether a technology can be commercialized ultimately depends on whether its performance, power consumption, and cost are competitive.
3. Product iteration speed is extremely fast, with continuous iterations from 800G to 1.6T to 3.2T. R&D investment costs remain high, and falling behind in the iteration rhythm will result in being left behind by the market.
Conclusion
The rise of optical modules has never been the result of single-market speculation.
When policies elevate it from an inconspicuous transmission accessory to the strategic height of the AI computing power foundation, we see a clear long-term mainline: speed upgrades + domestic substitution + low-power optoelectronic interconnect transformation. These three engines drive the optical module industry into a brand-new cycle.
In the computing power competition, when it comes down to it, it is not just about competing in GPUs, but also about the speed of the "conveyor belt."
And this computing power conveyor belt is now ushering in its own era!
Understanding the policies is the only way to truly understand the industry cycle! Like and bookmark to lock in the core logic of this optical module industry issue. Follow us to keep up with the latest iteration trends in the AI optoelectronic industry.
Interactive Topic: When do you think 1.6T will achieve large-scale commercialization? Welcome to leave your views in the comments section.
References
| Policy | Issuing Authority |
| "Action Plan for the Upgrading of the AI Computing Power Industry Chain" | Ministry of Industry and Information Technology (MIIT), National Development and Reform Commission (NDRC) |
| "Implementation Opinions on the Innovative Development of 'AI + Information and Communications' (2026-2028)" | MIIT |
| "White Paper on the Development of the High-Speed Optical Module Industry (2026)" | MIIT |
| Action Plan for Stable Growth in the Electronic Information Manufacturing Industry 2025-2026 | MIIT |
| Catalog of Encouraged Industries for Foreign Investment (2025 Edition) | NDRC, Ministry of Commerce |
| "Action Plan of Guangdong Province on Accelerating the Cultivation of New Tracks and Leading the Construction of a Modernized Industrial System (2026-2035)" | People's Government of Guangdong Province |
| "Action Plan for the Innovative Development of the Optical Chip Industry in Guangdong Province (2024-2030)" (Continued Implementation) | People's Government of Guangdong Province |
| "Action Plan of Shenzhen Municipality on Accelerating the High-Quality Development of the AI Server Industry Chain (2026-2028)" | Shenzhen Municipal Bureau of Industry and Information Technology |
| Shenzhen 2026 Key Industry R&D Funding Policy (Special Fund for Optical Computing and Optical Carrier Information) | Shenzhen Municipal Bureau of Science and Technology Innovation |