ZTE at MWC Shanghai 2026: Agentic AI Is Rewriting What Mobile Networks Need to Do

Mobile networks were built to move bits. ZTE’s argument at this week’s MWC Shanghai is that AI agents have changed what the network is actually for.

At MWC Shanghai 2026, ZTE Corporation’s Chief Development Officer Cui Li delivered a keynote titled “Unlocking Value and Embracing Uncertainty in the AI Era”, framing the company’s strategic shift under the slogan “All in AI, AI for All.” The core thesis: the explosive growth of autonomous AI agents is not just adding more traffic to mobile networks, it is changing what networks fundamentally need to deliver — shifting the requirement from high-bandwidth connectivity to what ZTE calls “reachable intelligence,” the ability to process AI workloads locally, at low latency, close to the point where the agent is actually operating.

The GTI Summit at MWC Shanghai, where ZTE CDO Cui Li and 6G architect Guo Qi both delivered keynotes under the theme “Mobile AI Powering 6G Future,” positioned ZTE as one of the vendors most aggressively framing 6G not as a faster 5G but as the first mobile generation designed from the ground up for an AI-native world.

What Happened

According to ZTE’s official press release, Cui Li’s keynote framed 2026 as a moment of structural uncertainty: AI is advancing faster than any single “one-size-fits-all” model can serve, and that speed of change means the companies and network operators that build for flexibility rather than optimizing for today’s dominant configuration will be better positioned for whatever comes next. ZTE’s strategic response is to deeply embed AI-native capabilities across its products at every layer — in base stations, in core network software, in handsets, and in its GigaMIMO 6G antenna prototype — rather than treating AI as an overlay added to existing infrastructure.

At the GTI Summit at MWC Shanghai 2026, ZTE’s 6G planning architect Guo Qi presented on the convergence of mobile AI with 6G architecture, examining how the shift to an agent-centric internet fundamentally changes the performance requirements mobile networks must satisfy. The key distinction ZTE is drawing: 5G was optimized for human users generating data that travels to cloud servers and back. 6G, in ZTE’s framing, needs to be optimized for AI agents that need continuous, low-latency access to compute and data that may be distributed across edge nodes, satellites, and ground infrastructure simultaneously. That is a significantly more demanding and architecturally different requirement than anything 5G was designed to serve.

The Mechanism: From Connectivity to Reachable Intelligence

ZTE’s argument about “reachable intelligence” rests on a claim about AI agent latency requirements that deserves examination rather than assumption. A human user making a search query can tolerate 200-400 milliseconds of round-trip latency without noticing. An AI agent making real-time decisions — an autonomous vehicle adjusting trajectory, a factory robot responding to a machine fault, an AI-assisted surgeon tool confirming a positioning calculation — may require end-to-end response times measured in single-digit milliseconds, with processing happening at the nearest edge node rather than routing to a distant data center and back. That latency requirement cannot be met by routing computation to a central cloud, regardless of how fast the network connection is, because the physics of signal propagation introduce irreducible delays that are incompatible with the response times agentic applications in latency-sensitive domains actually need.

GigaMIMO is ZTE’s most visible physical demonstration of how to meet those requirements at the radio layer. The prototype uses over 2,000 antenna elements operating in the U6 GHz band, managed by AI algorithms that continuously adapt beam patterns to user locations and channel conditions, achieving claimed capacity improvements of up to 10 times over current 5G massive MIMO configurations. The antenna count matters because AI-driven beam management becomes meaningfully more capable at this scale: the system can direct precise beams to individual devices with enough granularity to serve many simultaneous users in a dense environment without the interference problems that limited earlier massive MIMO approaches.

The Backstory

ZTE’s 6G-AI convergence positioning at MWC Shanghai is not an isolated announcement — it reflects years of research and standardization work. The ITU-R formally finalized the Minimum Technical Performance Requirements for IMT-2030 (6G) in February 2026, marking the transition from 6G as a vision to 6G as an implementation programme with defined performance targets. ZTE has been actively participating in that standardization process and has previously published a detailed white paper on AI-native 6G architecture, arguing that AI should be designed into the network standard from the outset rather than layered on afterward as it was with 5G.

The company’s SAGIN (Space-Air-Ground Integrated Networks) framing connects to a separate strand of 6G standardization that has attracted significant attention: the question of how to provide continuous AI connectivity to users and systems that move between terrestrial mobile networks, airborne relay platforms, and satellite coverage, without the handover disruptions that currently make satellite-to-cellular transitions unreliable for latency-sensitive applications. For agentic AI specifically, continuous connectivity across all three layers matters because an agent that loses connectivity mid-task is not just degraded — it may fail in ways that cause real-world harm if the task involves physical systems. That’s the same infrastructure concern driving NTT DOCOMO’s deployment of Nokia’s MantaRay AutoPilot on public cloud — autonomous AI network management requires the network itself to be more adaptive and self-correcting than anything today’s manually configured base-station infrastructure provides.

Reactions

ZTE has positioned itself as a conceptual bridge between the telecommunications infrastructure world and the AI applications world, arguing that the decisions network architects make in the 6G design phase will constrain or enable an entire generation of AI agent applications. Whether that framing resonates with network operators — who are simultaneously being pitched by every major telecom vendor on their respective 6G and AI-native visions — will depend more on ZTE’s ability to demonstrate working systems at commercial scale than on the architectural arguments the company is making at MWC Shanghai. GigaMIMO is a prototype, not a commercial product; SAGIN remains largely at the standardization and research stage.

The Dispute: Standards Lead, Markets Follow

The most consistent critique of 6G-AI convergence announcements from ZTE and its competitors is that they describe an end state rather than a credible near-term deployment timeline. 6G commercial service is not expected before 2030 in even the most optimistic scenarios, and the IMT-2030 standards process is still in its early implementation phase. ZTE’s arguments about what 6G must do for AI agents are architecturally coherent, but they will not shape a deployed commercial network for at least four years. In the meantime, the latency and edge compute requirements that ZTE’s ‘reachable intelligence’ framing addresses will be served — or not — by 5G Advanced networks and private edge deployments, not by 6G.

There is also a geopolitical dimension ZTE’s public messaging does not directly address. ZTE is a Chinese vendor operating in a global market where US export controls, Five Eyes cybersecurity concerns, and EU vendor diversity requirements have all shaped which companies build which countries’ critical telecommunications infrastructure. The Five Eyes’ joint advisory this week warning about AI-enabled cyber threats did not name specific vendors, but the broader context in which it was issued — the same week China’s LineShine topped the supercomputing rankings on all-domestic chips — is the same context in which ZTE is presenting its 6G-AI vision. Every technical argument ZTE makes about network architecture is also, implicitly, an argument about who builds the infrastructure that hosts agentic AI at global scale.

What Happens Next

ZTE’s GigaMIMO and SAGIN demonstrations at MWC Shanghai set the agenda for its 6G commercial roadmap over the next several years, with formal standardization milestones in 2026-2027 followed by technical trial deployments in 2028-2029. The more immediate near-term signal will be whether ZTE’s 5G Advanced products — particularly the AI-driven base station software and edge computing platforms it is already deploying in China with China Mobile, China Telecom, and China Unicom — generate the kind of verifiable performance data that could support the broader ‘agent-centric network’ architecture claims. Those deployments, not the 6G prototypes, are the near-term proof of concept for whether AI-native network management actually delivers at scale.

Why It Matters

ZTE’s MWC Shanghai presentation is significant because it articulates a specific, testable claim about what 6G must do that other vendors’ visions often leave vague: that the latency and spatial distribution requirements of agentic AI cannot be met by routing computation to centralized cloud infrastructure, and that the network layer itself must become part of the AI compute stack rather than merely a carrier of bits that AI applications consume. That claim has direct implications for how much of the world’s AI processing will happen at the edge versus in centralized data centers, and for how the investment flowing into large-scale AI data center infrastructure globally will need to be supplemented by distributed edge infrastructure to serve the latency-sensitive agentic applications that 5G Advanced can partially address but 6G is being designed to handle systematically. Whether ZTE’s vision of 6G as AI infrastructure becomes the dominant architectural paradigm or one voice among many will depend on how quickly its GigaMIMO and SAGIN systems move from Shanghai demonstration to deployable commercial product.

Sources

ZTE Corporation (PRNewswire); GTI Summit MWC Shanghai 2026; RCR Wireless News; BriefGlance.