Connect with us

AI

Huawei’s AI-Centric Network Pitches Tokens Over Bytes at MWC Shanghai

Huawei unveiled its AI-centric target network at MWC Shanghai 2026, urging carriers to shift from selling bytes to selling AI tokens. Analysts are skeptical.

Published

on

Huawei used MWC Shanghai 2026 to pitch an AI-centric network and ask carriers to start selling AI tokens alongside data bytes. David Wang, Huawei’s Deputy Chairman of the Board and Rotating Chairman, told the opening keynote on June 24 that the next decade of mobile communications will run through networks rebuilt around intelligence. Huawei’s official position frames the architecture as the platform carriers need to charge for token production, transmission, and consumption at scale.

China’s three major carriers are already running pilots on the new architecture. Independent analysts told Fierce Network that counting tokens is a particularly bad way to sell AI access. Suman Kanuganti at Personal AI disagrees and expects the first US telco token plan inside six to nine months. The bet is on. So is the rebuttal.

Wang’s Token Era Vision and Six Imperatives for the Decade Ahead

The opening framing by Wang was his sixth-generation pitch. “With each generation, we have pushed the limits of spectral efficiency and performance,” Wang said in his keynote. “Network architecture has gradually flattened, with new application scenarios and services emerging left and right.”

Tokens are the fundamental unit of output in large language models and AI services, in Huawei’s framing. Wang cited Huawei projections showing the number of AI agents operating worldwide is expected to exceed 100 billion AI agents by 2030 and could reach the trillions by 2040. In the most densely connected urban environments, agent density is projected to surpass 10 million per square kilometer. That is the traffic base Huawei wants carriers to be paid for, and it sits well above the kind of usage current data plans price against.

Wang framed his pitch around Huawei’s official token-monetization announcement and a list of six imperatives for the industry’s next decade, each one a choice carriers and regulators must make, in Huawei’s telling. Items two, four, and five map directly onto Huawei’s own commercial roadmap: the AI-centric target network, the U6 GHz push, and the standards work inside 3GPP. Items one, three, and six form the strategic envelope around them.

  1. Developing new services and capabilities for future mobile communications systems.
  2. Integrating AI with mobile communications to build three distinct layers of intelligence.
  3. Building network architecture for integrated satellite-ground communications.
  4. Advocating for sustainable and future-oriented spectrum planning and allocation.
  5. Clearly defining the specifications of AI-native core networks.
  6. Exploring new business models and application scenarios for mobile services.

The Three-Layer Network Huawei Wants Carriers to Build

Huawei unveiled the AI-centric target network at the event alongside China’s three major carriers. The architecture has three integrated layers, each rebuilt to serve AI workloads as the primary design target. The bottom layer is a basic communications network, designed for real-time interaction with guaranteed connectivity, high uplink and downlink, and guaranteed quality of service. The middle layer is a computing network that shifts from mere traffic transport to network-wide compute scheduling, with connecting to the network equivalent to accessing compute.

The top layer is AI computing infrastructure, anchored by Huawei’s SuperPoDs and designed for high-efficiency token production. The stack supports open-source development and integration with third-party systems, per Huawei’s MWC Shanghai 2026 microsite. This layer is where operators can charge directly for token production, transmission, and consumption. “Connecting to the network being equivalent to accessing compute” is the line Huawei used at the booth and in the keynote to describe the shift.

Where a traditional carrier network is optimized to push the most bytes per second to the most users at the lowest cost, the AI-centric model assumes some of the most valuable bits are the AI tokens generated, transmitted, and consumed by agents running on behalf of users. Carriers that do not rebuild risk being reduced to dumb pipes carrying token traffic for someone else’s monetization layer. Huawei made that case at the booth and on the keynote stage in equal measure.

Layer What it does Source-quoted shift
Basic communications network Real-time interaction with guaranteed connectivity “from traffic-centric networking to networking for real-time interaction”
Computing network Network-wide compute scheduling and supply “connecting to the network being equivalent to accessing compute”
AI computing infrastructure High-performance, efficient token production “support for open-source and open ecosystems”

Why the Uplink Is Now the Bottleneck, and U6 GHz the Fix

Mobile networks have historically been built for downlink-heavy consumers watching video and browsing the web. AI inverts that. Uplink is now the bottleneck, and Huawei is selling the fix.

AI glasses used for real-time translation and exhibition viewing already require sustained uplink speeds of 20 Mbps, per Huawei’s keynote. Wang set new benchmarks for operators to plan around: 1 Gbps peak uplink capacity and 20 Mbps as a universal uplink baseline. Multimodal agents that see, hear, and act in real time need to upload as much data as they download. The numbers are tight against what most current 5G mid-band spectrum can carry.

Operators will need to secure continuous bandwidth of 200 MHz to 400 MHz or more per cell to deliver five times downlink and ten times uplink at low latency and high reliability. That sits well beyond what current mid-band 5G allocations typically allow. Upper-6 GHz is the band Huawei is now pushing, which Wang called the next-generation golden frequency band for AI-era mobile networks. More than 20 countries and regions have explicitly designated U6 GHz for International Mobile Telecommunications (IMT), covering nearly 80% of the world’s population. In 2026, which marks the commercial debut of U6 GHz, the Middle East is expected to deploy the world’s first commercial 5G-A network running on U6 GHz, with carriers in Hong Kong and Macao of China also initiating commercial deployment.

  • 20 Mbps uplink: minimum required for AI-glasses real-time multimodal interaction
  • 1 Gbps peak uplink: new planning benchmark Wang set at MWC Shanghai
  • 20 Mbps universal baseline: target for operators going into token monetization
  • 200 MHz to 400 MHz: continuous bandwidth carriers need per cell to hit those targets
  • More than 20 countries and regions: have explicitly designated U6 GHz for IMT, covering nearly 80% of the world’s population

The Subscriber Numbers Behind the Pitch

The pitch is backed by a real installed base. The number of 5G-A users worldwide has exceeded 100 million, per Huawei’s official release. China alone has surpassed 110 million 5G-Advanced users, per Wang’s keynote. Global 5G subscribers have crossed 3.1 billion, a separate figure Wang cited to size the wider market carriers are about to renegotiate pricing on.

2026 is also the year U6 GHz goes commercial. The Middle East is expected to deploy the world’s first commercial 5G-A network on U6 GHz. Hong Kong and Macao carriers will initiate commercial deployment on the same band.

Market Milestone Source as of MWC Shanghai 2026
Global 5G-A users More than 100 million Huawei official release
China 5G-A users Surpassed 110 million Wang keynote
Global 5G subscribers Crossed 3.1 billion Wang keynote
Hong Kong and Macao Commercial U6 GHz deployment beginning 2026 Huawei official release
Middle East First commercial 5G-A on U6 GHz in 2026 Huawei official release

Live Deployments Already Running on the New Stack

In Singapore, Singtel’s deployment of agentic AI in customer operations saw its AI assistant handle more than 70,000 customer interactions in six weeks. 70% of routine queries were resolved without human involvement, freeing employees to focus on higher-value work. Singtel plans to extend the deployment to voice channels.

On the industrial side, China Mobile’s 5G-Advanced network is already streaming real-time thermal imagery from firefighting robots on the frontline of emergency response. Cloud-based AI systems use that feed to support fire source localisation, remote piloting, and autonomous execution in complex scenarios. The robots are commercial units, not pilots, per the keynote.

On the consumer side, Huawei’s Celia assistant is recording three billion daily activations, a figure the company tied to a 4.5-fold increase in agent distribution. In logistics, AI agent networks have demonstrated the ability to cut shipment delays by 15% by replanning delivery routes in real time during weather disruptions. At the 5G-A monetization forum’s milestone results on June 25, China Mobile Research Institute, Huawei, and GSMA Intelligence jointly introduced the Connection Agent and the “China Mobile Bixing Agent Platform” Intent Openness Gateway, a piece of the agent-to-network plumbing Huawei wants to standardize.

  • Singtel: 70,000 customer interactions handled in six weeks; 70% of routine queries resolved without human involvement
  • China Mobile firefighting robots: real-time thermal imagery streamed over 5G-Advanced for fire source localisation and remote piloting
  • Huawei Celia: three billion daily activations, a 4.5-fold increase in agent distribution
  • Logistics AI agent networks: 15% shipment delay reduction by real-time route replanning

Why Counting Tokens May Not Sell Like Data Plans

The skeptic case is real. Roger Entner, founder and principal analyst at Recon Analytics, told Fierce Network he does not think a straight by-the-token metering system would readily find consumer acceptance, noting that Recon Analytics’ data shows more than a quarter of mobile customers are already using AI on a daily basis and many AI companies are selling bucket plans for access. Suman Kanuganti, founder of Personal AI, sees it the other way and predicts the first US telco token plan will land inside six to nine months, citing publicized work with AT&T, T-Mobile, Verizon, and Comcast. The sharper critique came from the analyst case against metering AI access by token, advanced by Dean Bubley, founder of Disruptive Analysis.

There might be ways to sell or bundle access to AI assistants, but I think counting tokens is a particularly bad way to do it. Trying to sell AI applications on token numbers is like trying to sell content by the pixel. Nobody understands it.

Dean Bubley, founder of Disruptive Analysis, made the case in an interview with Fierce Network. The consumer-billing layer is where token monetization either finds paying customers or stalls. The split ran through the same MWC Shanghai floor.

What Lands in the Next Six Months

The first concrete date on Huawei’s roadmap is August 2026. GSMA, China Mobile, and Huawei jointly unveiled the 5G-A High-Speed Railway Network Acceleration Service at MWC Shanghai. The package is built on a “1+3+5” framework and is scheduled for commercial launch in China that month.

The Middle East expects to deploy the world’s first commercial 5G-A network running on U6 GHz in 2026. Hong Kong and Macao carriers will initiate commercial U6 GHz deployment this year as well. Huawei has said it will collaborate with carriers in Guangdong, Shanghai, Hebei, and other Chinese locations in 2026 to reengineer B2C and B2H services with AI. The announcements ran across the same Hall N1 stage.

Huawei is committed to shaping 3GPP Release 21, the standards framework that will govern mobile communications from 2030 to 2040. Release 21 is expected to launch in March 2027. Standards and spectrum decisions sit on a tighter clock than the AI roadmaps vendors are publishing alongside them. Wang told the keynote that the decisions operators and governments make in the next 12 to 18 months will determine who leads and who follows in the global race toward intelligent connectivity. The token-meter question runs through the same window.

Frequently Asked Questions

What is the Huawei AI-centric network?

Huawei’s AI-centric target network is a three-layer architecture unveiled at MWC Shanghai 2026 alongside China’s three major carriers. The basic communications layer is rebuilt for real-time interaction. The computing layer is positioned so that connecting to the network is equivalent to accessing compute. The AI computing infrastructure layer, anchored by Huawei’s SuperPoDs, is designed for high-efficiency token production.

What is token monetization?

Token monetization is Huawei’s framing for charging carriers’ customers for the production, transmission, and consumption of AI tokens, the fundamental output unit of large language models and AI services. It extends the existing mobile business model, where carriers charge for data bytes, into a layer where the network also bills for the AI inference its customers run.

When will 5G-A networks running on U6 GHz launch commercially?

2026 is the commercial debut year for U6 GHz, per Huawei. The Middle East is expected to deploy the world’s first commercial 5G-A network running on U6 GHz in 2026. Carriers in Hong Kong and Macao of China will also initiate commercial U6 GHz deployment this year.

How many 5G-A users are there right now?

Huawei’s official release says more than 100 million 5G-A users are active worldwide as of MWC Shanghai 2026. David Wang’s keynote added that China’s 5G-Advanced user base alone has surpassed 110 million. Global 5G subscribers, across 5G and 5G-A, have crossed 3.1 billion.

Who is skeptical of Huawei’s token pitch?

Independent analysts Roger Entner of Recon Analytics and Dean Bubley of Disruptive Analysis told Fierce Network that counting tokens is a difficult way to sell AI access to consumers. Bubley compared token-metered pricing to selling content by the pixel and warned that consumers will not understand what they are buying. Suman Kanuganti, founder of Personal AI, disagrees and expects the first US telco token plan inside six to nine months.

Logan Pierce is a writer and web publisher with over seven years of experience covering consumer technology. He has published work on independent tech blogs and freelance bylines covering Android devices, privacy focused software, and budget gadgets. Logan founded Oton Technology to publish clear, no nonsense tech news and reviews based on real hands on testing. He has personally tested and reviewed dozens of mid range and budget Android phones, written extensively about app privacy, and built and managed multiple WordPress publications over the past decade. Logan holds a bachelor's degree in English and studied digital marketing at a certificate level.

Continue Reading
Click to comment

Leave a Reply

Your email address will not be published. Required fields are marked *

Trending