AI-Optical Network unlocks the “Golden Triangle” for operators’ value increase

Technical Features

By Fang Hui, Senior Vice President of ZTE Corporation

The explosive growth of open-source large models like DeepSeek has lowered industry barriers, spurring increased computing and network demands, and bringing global AI development to a critical point. According to IDC, the global generative AI market is expected to grow at a 63.8% CAGR over the next five years, reaching $284.2 billion by 2028. Yet beneath this prosperity, AI deployment today still remains constrained by “computing silos, latency fluctuations, and unpredictable performance”. To address these challenges, ZTE’s “AI-Optical Network” strategy has come into being, aiming to achieve ubiquitous computing power, deterministic connections, and  personalized experiences.

Moreover, as computing power emerges as the fifth infrastructure pillar, AI-optical networks become critical to enabling intelligent, high-speed connectivity. They not only redefine the value of networks, but also rebuild industrial business models. Over the next decade, powered by AI-optical networks, telecom operators will shift from bandwidth sales to computing power operations—unlocking a trillion-dollar market.

REDEFINING THE NETWORK PARADIGM: BUILDING THE “GOLDEN TRIANGLE” WITH AIOPTICAL NETWORK

Network elasticity: 10G optical network boosting ubiquitous computing power The third wave of optical innovation, marked by the large-scale commercial deployment of 50G PON, heralds the dawn of the 10G access era. This breakthrough enables both home and enterprise access bandwidth to exceed 10 Gbps, powering applications such as 8K holographic display and XR cloud office. Through continuous innovation in FTTR and FTTM, ZTE is redefining the boundaries of connectivity for more immersive home entertainment experience and zero-error industrial production. Today, optical networks have gradually evolved into the “computing power nexus” that bridges the virtual and physical worlds.

Network intelligence: Changing the transmission pipeline into a decisionmaking brain.
AI is reshaping traditional networks. For example, ZTE Nebula Telecom Large Model is capable of fault prediction and self-healing.
It transforms the network from a transmission tool into an entity with sensing and decision-making capabilities. In scenarios like industrial quality inspection and remote healthcare, the integration of deterministic latency with realtime AI inference enables a seamless “perceive-decide-act” loop. This overcomes the “best-effort” constraints of traditional IP networks, providing a deterministic foundation for machine collaboration, autonomous driving, and other ultra-low-latency applications.

Ecosystem synergy: integration of optics, sensing, and computing An AI-driven all-optical architecture enables deep convergence of networks, computing, and sensing.
When home- and city-level compute nodes are interconnected through all-optical links, they form a distributed, hyper-converged brain. In this paradigm, boundaries blur: the optical network serves as the vascular system, computing becomes the brain, and sensing acts as the sensory interface. This synergy not only boosts resource efficiency but also drives innovative business models like “Compute Banks” and “Experience-as-a-Service,” enabling operators to shift from selling bandwidth to delivering personalized experiences.

REVITALIZING OPERATOR VALUE: ACTIVATING THE “GOLDEN TRIANGLE” FOR MONETIZATION

AI-Optical Campus Solution: unlocking new value for operators in the ToB market
•Smart manufacturing upgraded: The smart roaming algorithm boots AGV precision to ensure zero-lag transition. In one automotive plant, this boosted AGV transport frequency by 25% and reduced in-plant transport time of materials by 2.1 hours per day.
•High-density connectivity pressure relieved: In environments such as universities, airports, and convention centers, knowledge graph-based intelligent traffic scheduling effectively alleviates congestion caused by concurrent services. For instance, after implementation in a university’s multimedia classrooms, network-related complaints during peak usage periods dropped by 65%.
•Commercial security redefined: The “AI Safety Assistant”, combining fiber optics with machine vision, enables the automatic detection of over 20 types of risks, such as fire and intrusion, replacing 90% of manual inspections. This technology has helped a hotel chain reduce operating costs by 45% while achieving incident response times within seconds. Operators need to evolve from “connectivity providers” to “industry enablers”. AI optical campuses embed AI capabilities into every connection, extending them to factory workshops, street-side stores, and school classrooms, enabling in-depth synergy between networks and services. This also serves as a key leverage for operators to compete for dominance in the ToB market.

AI-OPTICAL INFRASTRUCTURE: ACCELERATING COMPUTING POWER TRADING
•Precision access: With a word-leading AI-native router achieving 95% accuracy in service awareness, ZTE’s AI-Optical Network meets users’ diversified computing power access needs, drives innovation in computing power access, and allows enterprises to adopt computing-network convergence packages for better service experience.
•Agile use: The industry’s first CPE that integrates sensing, communication, and computing together, supports 1,000+ inference application scenarios across 30+ industries. It reduces inference latency by a factor of 5 and doubles parallel inference capabilities at the same cost, redefining the billing model for inference services.
•Efficient training: The industry’s highest-density 800GE intelligent computing switch, equipped with large buffers to handle burst traffic and innovative congestion control technologies such as iGLB and ENCC, improves computing efficiency by 15% and unlocks the “hidden capacity” of GPUs.
•Ultra-Fast Interconnect: The industry’s first 800G/1.6T C+L integrated full-band OTN doubles single-fiber terabit capacity compared to traditional solutions, providing a “compute acceleration channel” for intelligent computing centers.
The essence of computing power trading is latency economics. Operators’ millisecond-level deterministic latency zones, built upon the AI-optical network, will become a core metric for pricing compute resources.

AI-OPTICAL HOME: AN EMOTIONAL ENGINE FOR USERS’ ARPU GROWTH
•A family-based vDC for new service models: Home gateways leverage local computing power to precisely sense user needs and dynamically optimize bandwidth strategies. For instance, identifying children’s study hours to automatically block game traffic has boosted education platform usage by 23%, opening up a new “scenario-based subscription” charging model for operators.
•AI Wi-Fi 7 for enhanced user experience: Omnidirectional antennas plus self-developed chipsets enable intelligent roaming and seamless whole-home 3000 Mbps experience. After the deployment, users can enjoy lag-free 4K/8K VR services.
•Strong emotional interaction for deep connection: Based on the collaboration between a home-based knowledge base and an end-cloud model, a tailored smart assistant is designed to enhance people’s daily living. For the elderly, AI screens deliver services such as schedule reminders through intuitive voice interactions, effectively easing challenges associated with memory decline. For children, home companion robots are equipped with advanced emotion recognition and bionic interaction systems, serving as intelligent and adorable companions. These robots excel at understanding children’s emotions through playful interactions and empathetic dialogue analysis.
AI home services will drive ARPU growth. Leveraging the AI home service ecosystem, operators enhance personalized service experiences such as home guardianship, emotional interaction, and scenario-based subscriptions, thereby building a “one-time deployment, continuous revenue” model.

FUTURE OUTLOOK: BRIDGING THE DIGITAL DIVIDE AND ACCELERATING AI INCLUSIVITY
The global AI market is booming at a CAGR of over 80% (according to IDC forecasts), yet uneven computing power distribution across regions, industries, and groups has created a “new digital divide” hindering societal advancement. AI-Optical Network will bridge this gap to speed up AI inclusivity based on:
•Open architecture: Eliminating technological barriers to ensure equal access to the computing power resource pool for any organization.
•Industry standards: Breaking down information silos and enabling efficient scheduling and synergy of computing, data, and models, to ensure equal capabilities between the edge and core.
•AI inclusivity scenarios: Powering applications such as telemedicine, personalized education, smart farming, and agile production through low-latency and highly reliable computing to improve livelihoods.
The ultimate mission of AI-optical networks goes far beyond technological iteration. It is to boost AI inclusivity by building an open ecosystem where everything is computable and all computing is interconnected.