ZTE’s Fang Hui: AI x Optical, Unveiling a New Era of Connected Intelligence

PRESS RELEASE: A network O&M engineer receives a new alert issued automatically by an AI-based optical component health forecast system, indicating that a backbone optical amplifier may degrade in 20 days. This is not science fiction, but a real-world system already running on live networks, the result of a profound transformation driven by the deep integration of AI and optical networks.

When nearly 30% of the energy consumption of global data centers is used for data transmission between chips, and when processing just 1TB raw data for AI training may require transferring up to 10TB intermediate results, a fundamental conflict becomes increasingly prominent: the huge gap between the exponential growth of computing power and the bottleneck of data transmission.

Light is becoming the core force to bridge this gap. We are standing at a historic turning point moving from all-optical networks to Al x Optical networks, and ultimately into a new era of connected intelligence.

Industry Turning Point: Super Cycle of Optical Networks Triggered by AI

The global optical network market is undergoing a structural transformation. The research institute Omdia has clearly stated that AI traffic will grow rapidly at a compound annual growth rate of 60%, and is estimated to pass conventional traffic and become majority of network traffic in 2031. The AI infrastructure build super cycle is catalyzing an optical network super cycle. The explosion of Generative AI not only brings about a surge in computing demands, but also reshapes the data flow mode. The traditional data center architecture dominated by north-south traffic is being fundamentally disrupted by the intensive east-west traffic patterns required for AI training and inference. This transformation forces optical networks to provide unprecedented bandwidth, near-zero latency, and deterministic service quality.

Market trends indicate that AI will shift from a special design consideration to a new benchmark that shapes the entire data center and network architecture in 2026. From finance to healthcare, from automobile manufacturing to higher education, various industries are deploying GPU clusters on a large scale, bringing the demanding requirements of AI to every network environment. This is driving the evolution of networks from “connecting information” to “connecting intelligence”, positioning them as the core infrastructure supporting future man-machine coordination and group intelligence.

Evolution Path: from FTTH to FTTR to FTTA

Looking back on the evolution of optical networks, we see a clear evolution path from “inclusive connectivity” to “revenue growth” and then to “intelligence empowerment”.

The first phase is the “fiberization stage” with large-scale deployment of Fiber to the Home (FTTH) as the core. It is primarily driven by technological innovations that radically optimize the Total Cost of Ownership (TCO) of network construction and operation. For example, accurate diagnosis solutions such as “Fiber Fingerprint” can significantly shorten the fault location time and reduce O&M expenses. The lightweight design of devices can help operators rapidly expand their user base and effectively control their CAPEX. The core mission of the network in this phase is to establish more economical and efficient basic broadband connections for the entire society.

The second phase is the “monetization stage” marked by service innovations such as Fiber to the Room (FTTR). Its focus shifts from internal cost reduction to external revenue creation. By packaging network capabilities into scenario-specific solutions, operators build a three-tier monetization system of speed, experience, and service, transforming high-quality connectivity directly into business growth. For example, offering guaranteed network slicing services for high-value scenarios such as gaming and business enables operators to effectively increase ARPU, marking a strategic upgrade from “pipeline” to “platform”. This not only drives revenue growth for operators but also delivers perceptible, differentiated and ultimate experience for end users.

Now, we are standing at the threshold of the third phase: Fiber to the Agent (FTTA), characterized by a fundamental restructuring of network architecture and operation mode centered around AI Agents. This is not mere tool enhancement. Rather, through deep integration of AI and optical networks, the networks are endowed with native intelligence, including perception, analysis, decision-making, and self-optimization. For example, optical transmission equipment with AI computing power can achieve fault prediction and self-healing, transforming the O&M mode from “passive response” to “proactive prevention”. More importantly, intelligent networks will transcend traditional connectivity boundaries and become the intelligent cornerstone of the new service ecosystem, including smart home, industrial digitalization, and even computing power networks, creating a new era of network autonomy and unlimited service expansion.

Global Practice: Three Breakthroughs Leading All-Optical Innovation

Facing the intelligent era, ZTE is building the core competitiveness of intelligent connectivity based on the deep integration of AI and all-optical networks, and has achieved groundbreaking innovations in three key fields.

In AI x Optical infrastructure, we are committed to building a solid foundation that enables smooth evolution and cross-domain intelligence for the industry. On the one hand, our latest 50G PON three-generation time-division coexistence solution has successfully solved the global problem of legacy terminal compatibility. It provides the industry’s smoothest 10-gigabit upgrade path in more than 150 trials, dramatically lowering the deployment threshold and cost of building “10-gigabit cities”. On the other hand, our innovative Scale-Across intelligent computing solution, based on cutting-edge technologies such as 1.6T OTN and hollow-core fibers, interconnects geographically distributed data centers at a high speed and builds an “AI computing power factory” capable of cross-domain scheduling, achieving efficient pooling and globally optimal configuration of computing power resources. These two breakthroughs have jointly established an ultra-broadband, lossless, and scalable all-optical intelligent foundation to support the digital economy and intelligent era.

In AI x Optical campus, we drive all-optical networks deep into core industry scenarios, creating benchmark value. In Baqing County, Tibet, at an altitude of over 4,000 m, we built the first all-optical smart hospital in the snow-covered plateau. The innovative high-integration FTTR-B equipment improved both the network transport capacity and O&M efficiency in extreme environments. This practice not only proves the outstanding robustness of our all-optical solution in harsh scenarios, but also sets a model of social responsibility that empowers life and health and bridge the digital divide with reliable connectivity.

In AI x Optical home, we have redefined the standards for the experience of high-quality smart life. The innovative “distributed Wi-Fi all-optical network solution based on AI tuning” integrates the deep learning neural network model and all-optical architecture for the first time. It can identify and avoid interference in real time, and schedule resources dynamically, achieving seamless coordination between “wired all-optical” and “wireless intelligent” networks. It delivers ultimate roaming experience with “zero perception and zero lag” for users, truly unleashing the value of ultra-gigabit bandwidth in every corner.

These three breakthroughs, from consolidating infrastructure, and empowering vertical industries to lighting up smart life, systematically demonstrate our comprehensive capabilities and firm commitment to driving the intelligent transformation of the society through the dual-engine momentum of “AI and all-optical”. 

Future Trends: Four Pillars for Connected Intelligence

In the future, the development of optical networks will evolve from technology-driven to scenario-driven and intelligence-driven. Its evolution will unfold around four core pillars, ultimately supporting a new era of Connected Intelligence.

All-domain hyper-converged architecture: Future networks will break the physical boundaries of space, air, ground, and sea, and achieve deep integration. The combination of AIDC-based terrestrial optical networks and space-based satellite laser communication technology will build a three-dimensional space-air-ground integrated computing network, to provide seamless continuous connectivity and computing power services for all-domain intelligence.

Intrinsically deterministic connectivity: For key scenarios such as industrial Internet and remote precise control, the networks need to upgrade from “best-effort” to “deterministic”. Through AI-based global scheduling and end-to-end hard slicing technologies, optical networks will provide strict deterministic guarantee of bandwidth, latency, and jitter. When proactively combined with quantum encryption and other security technologies, they can deliver trusted connectivity with both ultra-low latency and ultra-high security. This marks a fundamental leap for networks from “best-effort” to “mission-assured”.

AI-Native autonomous operation: AI will evolve from an external tool to the “digital brain” and intrinsic attribute of networks. From digital twin-based automatic planning to intent-driven simplified deployment, and to proactive self-healing of predictive maintenance, the intelligent closed loop throughout the full lifecycle will drive a fundamental transformation in the O&M model, and the role of network O&M engineers will change from “firefighters” to “policy definers.”

Green and sustainable development: As global attention on carbon emissions intensifies, network energy efficiency will become a key competitive differentiator. By introducing CPO/OIO chip-level optical interconnection, intelligent shutdown, dynamic power adjustment, and advanced cooling technologies, next-generation optical networks are expected to achieve a tenfold increase in capacity while maintaining or lowering the energy consumption per bit.

Conclusion: AI x Optical for Connected Intelligence

The evolution of networks is a three-part journey, from cost optimization to value creation, and ultimately to native intelligence. We are now standing at the threshold of the third phase. Facing the AI-triggered super cycle, optical networks are evolving from a connectivity tool to the cornerstone of the intelligent era.

When optical networks evolve from pipelines carrying “traffic” to neural networks carrying “intelligence”, when network O&M shifts from relying on expert experience to self-optimization driven by AI agents, when connectivity services mature from standardized bandwidth products to differentiated service experience, we are truly entering the era of connected intelligence.