OpenAI
2026-06-25
Huawei Pushes Token-Based Billing at MWC Shanghai 2026: Shifting Carrier Monetization from Bytes to AI Inference Value
Summary
Key Takeaways
At MWC Shanghai 2026, Huawei unveiled its "Advancing All Intelligence" strategy, urging carriers to shift from byte-based to token-based billing. With AI agents and LLMs becoming dominant network consumers, Huawei argues that the natural billing unit should be AI tokens generated and delivered, not bytes transferred.
Huawei and China Mobile Hubei validated the AI Inference Acceleration Solution based on OceanStor A800, Ascend A3 SuperPoD, and Unified Cache Manager, achieving a 372% token throughput improvement for long-sequence inference (code generation, multi-turn dialogue).
On the radio side, Huawei highlighted the Upper-6 GHz (U6 GHz) band for AI agent traffic, noting that AI wearables like smart glasses require 20 Mbps sustained uplink—a requirement 5G was not optimized for. Over 20 countries have designated U6 GHz for IMT, with 2026 as its commercial debut year, starting in the Middle East.
Huawei's AI-centric target network architecture gives carriers visibility into compute workloads, enabling token-based monetization. Global 5G-A users have exceeded 100 million, and Huawei is collaborating with carriers in China, Middle East, and Hong Kong to monetize 5G-A via high-uplink services and experience-based pricing.
Why It Matters
Huawei's token billing proposal is a defensive move against Western AI platforms (OpenAI, Google, Meta) and rival vendors (Ericsson, Nokia). By shifting billing from bytes to tokens, Huawei aims to lock carriers into its proprietary hardware ecosystem (OceanStor A800, Ascend A3 SuperPoD) and reposition them as AI compute gatekeepers, stripping AI platforms of the ability to bypass carriers via direct bandwidth purchase.
However, Huawei downplays key limitations: the 372% throughput gain is specific to long-sequence inference; tail latency and PFC/ECN bottlenecks for short, real-time AI tasks remain unaddressed. U6 GHz requires massive capex for coverage, a cost trap Huawei frames as an opportunity. Token monetization is impossible in Five Eyes markets due to Huawei bans, making this a regional lock-in strategy for China, Middle East, and Africa, not a global architecture.
PRO Decision
【Vendors】Competitors like Ericsson and Nokia should counter Huawei's token billing by highlighting its proprietary hardware lock-in (OceanStor A800, Ascend A3 SuperPoD) and promoting open-standard alternatives (O-RAN, generic x86/GPU) that offer interoperability and lower TCO. Partner with AI platforms (OpenAI, Google) for joint token billing pilots to prove feasibility without vendor lock-in.
【Enterprises】CIOs and architects must perform zero-trust audits on carrier token billing contracts: demand that pricing be tied to AI inference quality (latency, accuracy) not just token count. Evaluate whether the carrier network supports multi-cloud and hybrid architectures to avoid Huawei lock-in. Prioritize AI inference services with open APIs and standard protocols to retain portability.
【Investors】Look beyond Huawei's PR: the company is building a proprietary AI infrastructure monopoly in non-Five Eyes markets using geopolitical isolation. Monitor 5G-A equipment orders in China, Middle East, Africa but beware of political risk and lock-in. Investing in open networking and white-box vendors (e.g., Nokia's MantaRay AutoPilot) may offer more long-term resilience.
Get 3-5 key AI infrastructure signals weekly →
💬 Comments (0)