In an AI period outlined by the explosive progress of clever purposes, each technological and financial paradigms are present process systemic transformation, cell networks should quickly evolve from mere “visitors pipe” into “AI engines” , and the transformation of telecom operators (Telcos) into know-how firms (Techcos) has change into pressing. Seizing these historic alternatives and challenges, ZTE—guided by its Connectivity + AI technique—has launched AIR MAX, a next-generation cell community answer that optimizes three efficiencies: vitality effectivity, labor effectivity, and funding effectivity. With an AI-native mindset, it re-architects the technological basis, operation mannequin, and enterprise worth of cell networks,enabling operators within the new period to remodel their operations and redefine worth creation. On this article, ZTE Senior Vice President Zhang Wanchun affords a targeted examination of how you can reshape the cell community paradigm.

Over the previous three years, AI computing efficiency has surpassed human specialists within the MMLU (Huge Multitask Language Understanding) benchmark, with computing prices dropping by an element of 100, laying the inspiration for large-scale AI purposes. The emergence of merchandise similar to Doubao AI Cellphone, OpenClaw, and Gemini Purchasing Assistant has additional validated the usability and productiveness potential of agent-based AI applied sciences, accelerating the speedy improvement of AI purposes. Within the AI period, underneath the brand new financial paradigm of “Bit + Token,” how you can seize the alternatives introduced by AI, rework cell networks from knowledge pipelines into AI engines, cut back vitality and labor prices, improve enterprise worth and return on funding, and construct a sustainable “Create Extra, Save Extra” progress flywheel has change into a core problem for operators.
AIR MAX framework: Three-tier full stack AI functionality
Constructed on a 10-block AI functionality stack, AIR MAX systematically upgrades the cell community framework for the AI period. It builds a three-layer functionality framework comprising AI-native infrastructure, L4 autonomous community operations, and monetization engines. By constantly driving upgrades in technical structure, manufacturing effectivity, and enterprise fashions, ZTE AIR MAX reshapes cell networks with a brand new paradigm known as ‘serve AI with AI’, comprehensively empowering operators in functionality evolution, service transformation, operational transition, and ecosystem reinvention for the AI period.
AI-native Infrastructure Layer – Technical Structure Improve: ZTE’s AI-native infrastructure layer is constructed upon two elementary ideas. The primary precept is that AI Native doesn’t equate to GPU-centric structure. Communication workloads are structurally various—some duties are extra effectively dealt with by conventional algorithms, some require specialised AI acceleration, and others demand general-purpose AI computing energy. Due to this fact, we undertake a heterogeneous computing structure. The second precept is that communication AI fashions have to be deeply optimized. Via strategies similar to light-weight mannequin design, mannequin compression, sparsification, and FP4/FP8 combined precision quantization, we compress fashions from the billion-parameter scale all the way down to the million-parameter scale and even smaller, making real-time deployment possible. Primarily based on these ideas, we construct a high-performance, environment friendly, and scalable AI-native infrastructure by 4 key capabilities: embedded AI Radio, heterogeneous computing, AI-driven large-scale MIMO coordination, and AI-powered high-precision sensing.
- Embedded-AI Radio: Combine AI deeply into wi-fi units, leveraging AI-based DPD and associated algorithms to cut back vitality consumption by 35%–40%; make the most of MIMO enhancement applied sciences similar to AI-powered beam-forming and AI-powered channel equalization to enhance spectral effectivity by 20%.
- Heterogeneous Computing: Primarily based on ASIC AI + xPU heterogeneous computing, we construct an environment friendly computing structure tailor-made for various workloads, together with deterministic communication logic processing, communication-specific AI acceleration, and computing for normal AI workloads. Trade-leading merchandise similar to AIREngine, AI UPF, and AI Dice ship environment friendly, versatile, and scalable computing energy for AI4Net and Net4AI. Amongst them, AIR Engine has been deployed in over 150,000 websites globally.
- AI-Pushed Huge MIMO Collaboration: Modern AI-driven GigaMIMO large-scale antenna array collaboration answer, which has been verified in subject trials with China Cell. Clever centralized and distributed Huge MIMO achieves a cell throughput of 192 Gbps and constant consumer expertise, with edge-user charges improved by 6 instances.
- AI-Pushed Excessive-Precision Sensing: Empower built-in sensing and communication (ISAC) with AI to attain over 95% accuracy in drone detection, with each false alarm and missed detection charges beneath 5%.
L4 Autonomous Operation layer-Productiveness Improve: Aiming on the general purpose of upgrading productiveness from single-domain L4 to end-to-end L4, the autonomous operations layer consists of three core parts: a low-hallucination 1+N mannequin matrix, the industrial-grade dependable Co-Sight 2.0 agent manufacturing unit, and the CoClaw digital worker for end-to-end autonomous activity closure—enabling structured and measurable L4 capabilities.
- 1+N Mannequin Matrix: ‘1’ refers back to the general-purpose mannequin basis for the communications area, enabling exact understanding of enterprise situations; ‘N’represents a set of small-scale, structured fashions devoted to specialised area reasoning. The mannequin achieves a 92% accuracy charge in autonomous decision-making—industry-leading—and 95% factual consistency in outputs, approaching near-zero hallucination. It has been deployed in 20 high-value autonomous situations.
- Co-Sight 2.0 Agent Manufacturing unit: Agent improvement has advanced from “hand-crafted customization” to a “standardized assembly-line” course of, enabling the manufacturing of reliable, auditable, and high-quality brokers inside minutes.Co-Sight brokers ranked No.1 in each April and September 2025 on the GAIA benchmark.
- Co-Claw Digital Staff: Permits environment friendly human-machine collaboration and end-to-end collaboration amongst a number of brokers, supporting diminished and even unmanned operation.
Within the TMF Autonomous Community Degree evaluation, ZTE has collectively achieved industry-leading leads to situations together with RAN vitality effectivity optimization, RAN community fault administration, and repair assurance of particular person companies. AIR MAX will proceed to drive the evolution from area L4 to end-to-end L4.
Monetization Engine-Enterprise mannequin improve: With the overarching imaginative and prescient of empowering operators to reshape their enterprise paradigms and ecosystem positioning, ZTE constantly expands worth boundaries in three key areas: expertise monetization, AI service monetization, and AI ecosystem enlargement—accelerating the operator’s evolution from a connectivity supplier to an AI service supplier and finally to a secured AI ecosystem hub.
AI Ecosystem Enlargement: Construct a safe and environment friendly clever agent service system, supporting capabilities similar to eSIM-like authentication, on-demand subnet orchestration, teminal-edge-cloud collaborative computing, and proactive service supply, driving open enlargement and worth sharing inside the AI ecosystem.
Expertise Monetization: By combining “first-class cabin-level companies” with “scenario-based value-added packages,” ARPU is elevated by over 20%, driving the strategic shift from “traffic-centric” to “experience-centric” operations, redefining connectivity worth and accelerating industrial returns.
AI Service Monetization: Leveraging wide-area deterministic community connectivity, low-latency and cost-efficient heterogeneous distributed computing (flexibly enabled by AIR RAN, AI UPF, AI Dice, and so forth.), and big real-time knowledge era and provide, we set up a high-value closed-loop AI functionality to empower reasonably priced and inclusive Token companies.
Future outlook: From reshaping connectivity to transcending connectivity
AIR MAX isn’t just a technical improve, however a paradigm shift within the improvement philosophy of cell networks. Evolving from “reshaping connectivity” to “transcending connectivity,” AIR MAX will proceed to speed up the excellent transformation of telecommunication service suppliers into know-how firms, making cell networks the core infrastructure of an clever society.
Shifting ahead, ZTE will work hand in hand with world operators and ecosystem companions, leveraging AIR MAX because the core enabler, to collectively construct an AI-native future community and inject robust impetus into the event of the digital financial system.
