Huawei is positioning itself at the center of a fundamental shift in how telecommunications networks generate value, moving beyond traditional byte-based traffic monetization to embrace token-based AI services. At MWC Shanghai 2026, the company unveiled an integrated service-network-compute architecture designed to help carriers capture revenue from both conventional connectivity and emerging AI inference workloads. This strategy arrives as Chinese operators accelerate domestic procurement to reduce reliance on foreign technology, creating a self-reinforcing cycle that strengthens Huawei’s ecosystem while global competitors face mounting restrictions.
The developments matter because they illustrate how geopolitical fragmentation is reshaping technology markets. As export controls limit access to advanced Western chips, Chinese carriers and their suppliers are building parallel infrastructures optimized for AI agents and long-context inference. Huawei’s announcements show concrete technical progress in KV cache management and experience-layer monetization, yet they also coincide with enforcement actions that highlight the persistent risks of cross-border supply chains.
AI-Centric Target Networks Reshape Carrier Infrastructure
David Wang, Huawei’s Deputy Chairman, outlined six imperatives for the next decade of mobile communications during his MWC keynote, emphasizing the need for AI-native architectures that flatten traditional hierarchies while integrating satellite-ground systems. The company’s AI-centric target network prioritizes a shift from traffic-centric to intelligence-centric operations, enabling carriers to optimize both spectral efficiency and compute allocation dynamically.
This approach directly addresses the performance bottlenecks that emerge when AI models scale to handle multi-turn dialogues and code generation tasks. By embedding intelligence across radio access, transport, and core layers, Huawei aims to let operators translate network capabilities into differentiated services rather than competing solely on coverage or speed. The strategy aligns with broader industry movement toward intent-based networking, where service-level objectives drive automated resource allocation instead of manual configuration.
Token Monetization Emerges as a New Revenue Vector
Alongside Chinese carriers, Huawei demonstrated 5G-A high-uplink innovations paired with token monetization frameworks that allow operators to charge for AI inference capacity. The approach treats tokens—units of AI computation—as billable resources alongside conventional data traffic and premium quality-of-service experiences. Early trials showed carriers launching event-specific packages, such as a Thai operator’s football livestreaming offering that reached 420,000 users within three months of a one-week development cycle.
The technical foundation rests on Huawei’s Unified Cache Manager (UCM), which extends KV cache beyond on-chip and DRAM limits into petabyte-scale external storage. Validation with China Mobile Hubei using the vLLM-Ascend framework delivered token throughput gains of up to 372 percent on GLM-5.1 models at 128K sequence lengths, alongside 51–93 percent improvements in time-to-first-token. These metrics matter because long-context AI workloads rapidly exhaust conventional memory hierarchies, making external hierarchical caching a prerequisite for economically viable agentic services at carrier scale.
Agentic Operations Drive End-to-End Experience Upgrades
Eric Yang, President of Huawei Carrier Business, introduced the concept of agentic operations at the 11th Operations Transformation Forum, framing it as a scaling law for telecom revenue: Scale Out to reach more users and agents, Scale Up for personalized service models, and Scale Fast for rapid service launches. The framework maps four intelligent upgrades across the “Buy-Use-Assure-Retain” journey, replacing static offerings with intent-aware recommendations and proactive experience assurance.
A Hong Kong carrier trial during a 110,000-attendee concert improved high-value customer experience ratings by 33 percent through real-time network adjustments. Such outcomes suggest that agentic systems can convert network telemetry into actionable service differentiation, moving carriers from connectivity providers toward experience platforms. The model also expands addressable markets from 9 billion human subscribers to include hundreds of billions of AI agents, each potentially generating token-based revenue streams.
Domestic Procurement Accelerates Huawei Ecosystem Growth
China Telecom’s award of a multibillion-yuan server contract for 40,000 high-performance units underscores how state-driven self-reliance policies are channeling demand toward Huawei’s Kunpeng ecosystem. Although Huawei itself did not appear on the final supplier list, the six winning bidders for the Arm-based portion are all publicly tied to its architecture, allowing indirect market capture without triggering additional scrutiny.
TrendForce data projects that domestic suppliers led by Huawei and Cambricon will reach 56 percent of China’s AI server market in 2026, up from 46 percent the prior year, while foreign vendors fall to 21 percent. This shift reflects both policy incentives and practical constraints: internet companies are rapidly deploying locally designed ASICs and accelerators to maintain AI training and inference roadmaps. The resulting ecosystem lock-in strengthens Huawei’s position in servers, storage, and networking even as it faces continued exclusion from certain Western markets.
Persistent Legal and Regulatory Headwinds
Countervailing pressures remain visible. Bosch agreed to a $36 million penalty and disgorgement of profits after its non-U.S. subsidiaries sold $72 million in export-controlled MEMS sensors and automotive software to Huawei between 2020 and 2024. Meanwhile, a Danish court ordered the state to compensate TDC NET approximately $12 million after authorities mandated removal of Huawei DWDM equipment from its fiber network, ruling the intervention constituted expropriation under the constitution.
These cases illustrate the uneven compliance landscape. Voluntary self-disclosure yielded leniency for Bosch, yet the underlying export rules continue to constrain component flows. In Europe, security-driven equipment bans carry direct fiscal costs for operators, creating friction that slows network modernization even as carriers seek AI-ready infrastructure.
The convergence of these threads points to a bifurcated global telecom and AI supply chain in which Chinese carriers and their domestic partners accelerate token-centric and agentic capabilities while navigating ongoing restrictions. Success will hinge on whether Huawei’s integrated architectures can deliver measurable monetization advantages fast enough to offset external constraints and sustain the momentum now visible in domestic procurement and live-network validations.
