Fiber Standards Shift to Preform Core in AI Era

Introduction

AI computing clusters demand massive data throughput, pushing optical interconnect capacity to its physical limits. Traditional single-mode fibers are nearing their theoretical ceiling, and the industry is now racing to develop advanced fiber architectures, particularly at the preform stage, which determines the fiber's entire performance profile.

The preform is the glass rod from which optical fiber is drawn. Its refractive index profile, core geometry, and dopant distribution encode every transmission characteristic of the final fiber. In the AI era, control over preform design has become the primary competitive battlefield, and standard-setting bodies are increasingly focusing on preform-level specifications rather than only measured fiber parameters.

The Preform Battleground

Why has the standard war moved upstream? Because once a preform design is patented and embedded into an international standard, all downstream fiber manufacturers must either license that patent or design around it. This shift makes the preform the most defensible intellectual property chokepoint in the optical fiber value chain.

AI data centers require fibers with extreme bend resistance, low latency, and high modal bandwidth. Each requirement translates into specific preform characteristics, such as trench-assisted refractive index profiles or multi-core arrangements. Consequently, the companies that control preform patents effectively control the future of AI interconnect standards.

Core Patents: Multi-Core Structure

Multi-core fiber (MCF) is a frontrunner technology for space-division multiplexing, where multiple independent cores in a single cladding multiply data capacity. Key patents cover core placement, crosstalk suppression through trench structures, and coupling techniques at connectors. The international standardization effort, particularly within ITU-T and IEC, is now defining parameters for heterogeneous multi-core fibers.

For Chinese enterprises, the strategy is to file patent families around novel core arrangements, especially those using asymmetrical layouts that reduce crosstalk without requiring complex manufacturing processes. It is also essential to claim design methods that allow standard-compliant fibers to be produced with conventional chemical vapor deposition equipment.

Ultra-Low Loss Processes

Ultra-low-loss fibers rely on minimizing Rayleigh scattering and absorption, achieved by ultra-pure silica preforms and specialized dopant profiles. Patents here often cover specific dehydration methods, fluorine doping techniques, and core-cladding interface treatments that eliminate microbending losses. These processes are difficult to reverse-engineer and create strong trade-secret protection when combined with patents.

Standard bodies now include loss coefficient thresholds for specific wavelengths used in AI interconnect, such as O-band and E-band windows. Companies that hold process patents for achieving these thresholds can license them to competitors, creating a recurring revenue stream while maintaining dominance in the standard-setting committee.

Standard-Setting Strategies

To influence standards, Chinese firms should participate actively in ITU-T Study Group 15 and IEC SC86A working groups, submitting contributions that align with their patent portfolios. It is critical to propose measurable parameters, such as effective area, cutoff wavelength, and crosstalk limits, that match the capabilities of their own preform designs.

Additionally, companies should pursue joint research projects with academic institutions to generate foundational patents, then publish non-essential findings to build credibility. This open-innovation approach can help shape the standard's direction while keeping core claims protected in strategic jurisdictions.

Strategic Recommendations

Chinese enterprises should map the existing patent landscape for preform fabrication and identify white-space opportunities, especially in novel dopant combinations and additive manufacturing of preforms. Creating a defensive patent pool with peer companies can reduce litigation risk and increase bargaining power in cross-licensing negotiations.

Enterprises should also invest in high-throughput preform characterization tools and machine learning algorithms to optimize process parameters. By linking process patents to standardized certification methods, they can force competitors to adopt their testing protocols. Finally, it is essential to file patents early in both China and major markets like the US and Europe, ensuring that standard contributions do not become prior art against their own claims.

Conclusion

The shift of fiber standard competition to the preform stage reflects the deeper complexity of AI-era optical requirements. Multi-core structures and ultra-low-loss processes are now the decisive patent domains, and international standards are being written around these capabilities.

Chinese companies that combine aggressive patent filing with active participation in standard bodies and collaborative research will be well-positioned to define the next generation of fiber infrastructure. The race is not simply about making better fiber, but about controlling the intellectual property that makes standards possible.