To support this shift, Huawei showcased an integrated Service-Network-Compute strategy that blends service-layer intelligence, AI-optimized networking, and high-performance storage and compute . The goal is to let operators capture value from both traditional traffic and AI workloads.

Key product unveilings included:

• 10 AI-Optical Network (AI-ON) products: A suite of all-optical network solutions designed to build AI-centric target networks for widespread AI service delivery .
• Live-network AI inference acceleration: In partnership with China Mobile Hubei, Huawei completed China's first carrier live-network validation of its AI Inference Acceleration Solution, powered by OceanStor A800 storage. The solution delivered up to a 372% improvement in token throughput for long-sequence AI inference workloads .
• Multi-agent collaboration platforms: At the service layer, Huawei is helping carriers deploy specialized AI agents for calling, experience monetization, and home broadband .

Yang Chaobin, a Huawei executive, proposed the formula "5G-A x AI" as the strategic upgrade path for carriers, enabling them to monetize not just traffic and network experience, but AI services themselves .

ZTE's AiR Core Goes Agentic

ZTE presented its AI-native network core, AIR Core, which now supports agentic AI capabilities. This embeds intelligence directly into the core network to enable autonomous operations and AI workload delivery .

The GSMA's Open Telco AI Initiative Gains Momentum

First launched at MWC Barcelona on March 2, 2026, with AT&T and AMD as founding supporters, the Open Telco AI initiative delivers open foundations for what the GSMA calls "Telco-Grade AI" — including shared datasets, open-telco models, and a community portal for industry collaboration .

At MWC Shanghai, the initiative continued to be a central topic. The GSMA hosted an AI Leaders Roundtable and noted that Open Telco AI has grown to 46 member organizations in 35 countries . South Korean AI data evaluation company Selectstar also joined as an official data evaluation partner .

The initiative's importance is underscored by the fact that network operations represent 34% of operator OpEx, yet only 16% of GenAI deployments target this area .

The Cost Competitiveness of Chinese AI Models

A key backdrop to the MWC discussions was the rapidly shifting economics of AI. DeepSeek released its V4 model in April 2026, a 1.6-trillion-parameter Mixture-of-Experts model trained entirely on domestic Huawei Ascend and Cambricon chips with zero Nvidia CUDA dependency . It is described as cost-competitive with top US closed-source models from OpenAI and Google DeepMind .

More broadly, Chinese AI models in Q2 2026 generally offer per-token prices three to ten times below US frontier models . For example, Step 3.5 Flash ships at $0.10/$0.30 per million tokens — about 25× cheaper than GPT-4o with comparable math reasoning . Chinese labs now hold four of the top five positions in open-weight AI leaderboards .

This pricing pressure is forcing global operators to rethink their AI infrastructure partnerships and cost models.

Nokia's Agentic AI Push

Nokia made several major announcements at MWC Shanghai and the parallel DTW26 event:

• Google Gemini integration: Nokia announced it will integrate Google's Gemini models into the Nokia Assurance Center (its network monitoring and management platform) to bring AI agents into telecom network operations. Six specialized agents are planned, with two (Router and Event Triage) already operational .
• Autonomous networks portfolio upgrade: Nokia upgraded its autonomous networks portfolio with new agentic AI capabilities, including an Autonomous Networks Agent Library, the Autonomous Networks Suite, and MantaRay SMO — all aimed at improving network performance, reliability, and operational efficiency at scale .
• Fixed-network agentic AI: Nokia also launched agentic AI specifically for home and broadband networks, built on insights from 600+ million deployed broadband lines .
• AI-RAN with NVIDIA: Nokia continued its strategic AI-RAN partnership with NVIDIA, announcing new customer integrations with T-Mobile, Indosat, and SoftBank Corp., and successful functional tests of its anyRAN software on NVIDIA's GPU-accelerated AI-RAN platform .

The Big Picture: A Cross-Industry Consensus

Taken together, the announcements at MWC Shanghai 2026 reveal a strong consensus across the industry:

• Monetization must move beyond connectivity. Whether it's Huawei's token economy, ZTE's agentic core, or Nokia's AI slicing, the core message is that operators must capture value from AI workloads, not just data transport.
• Open collaboration is essential. The GSMA's Open Telco AI initiative, with its focus on shared datasets and models, reflects a recognition that no single operator can solve telco-grade AI alone.
• Cost pressure from Chinese models is reshaping the landscape. The availability of powerful, low-cost AI models from DeepSeek and others is creating both opportunities and competitive threats for telecom operators globally.
• Agentic AI is the next frontier. Both Huawei and Nokia are investing heavily in autonomous AI agents that can manage networks, handle customer experience, and deliver new services without human intervention.

For telecom operators, the message from MWC Shanghai 2026 was clear: the AI revolution is here, and the window for building a new revenue model — one based on tokens, agents, and intelligent infrastructure — is opening now.

To support this shift, Huawei showcased an integrated Service-Network-Compute strategy that blends service-layer intelligence, AI-optimized networking, and high-performance storage and compute . The goal is to let operators capture value from both traditional traffic and AI workloads.

Key product unveilings included:

• 10 AI-Optical Network (AI-ON) products: A suite of all-optical network solutions designed to build AI-centric target networks for widespread AI service delivery .
• Live-network AI inference acceleration: In partnership with China Mobile Hubei, Huawei completed China's first carrier live-network validation of its AI Inference Acceleration Solution, powered by OceanStor A800 storage. The solution delivered up to a 372% improvement in token throughput for long-sequence AI inference workloads .
• Multi-agent collaboration platforms: At the service layer, Huawei is helping carriers deploy specialized AI agents for calling, experience monetization, and home broadband .

Yang Chaobin, a Huawei executive, proposed the formula "5G-A x AI" as the strategic upgrade path for carriers, enabling them to monetize not just traffic and network experience, but AI services themselves .

ZTE's AiR Core Goes Agentic

ZTE presented its AI-native network core, AIR Core, which now supports agentic AI capabilities. This embeds intelligence directly into the core network to enable autonomous operations and AI workload delivery .

The GSMA's Open Telco AI Initiative Gains Momentum

First launched at MWC Barcelona on March 2, 2026, with AT&T and AMD as founding supporters, the Open Telco AI initiative delivers open foundations for what the GSMA calls "Telco-Grade AI" — including shared datasets, open-telco models, and a community portal for industry collaboration .

At MWC Shanghai, the initiative continued to be a central topic. The GSMA hosted an AI Leaders Roundtable and noted that Open Telco AI has grown to 46 member organizations in 35 countries . South Korean AI data evaluation company Selectstar also joined as an official data evaluation partner .

The initiative's importance is underscored by the fact that network operations represent 34% of operator OpEx, yet only 16% of GenAI deployments target this area .

The Cost Competitiveness of Chinese AI Models

A key backdrop to the MWC discussions was the rapidly shifting economics of AI. DeepSeek released its V4 model in April 2026, a 1.6-trillion-parameter Mixture-of-Experts model trained entirely on domestic Huawei Ascend and Cambricon chips with zero Nvidia CUDA dependency . It is described as cost-competitive with top US closed-source models from OpenAI and Google DeepMind .

More broadly, Chinese AI models in Q2 2026 generally offer per-token prices three to ten times below US frontier models . For example, Step 3.5 Flash ships at $0.10/$0.30 per million tokens — about 25× cheaper than GPT-4o with comparable math reasoning . Chinese labs now hold four of the top five positions in open-weight AI leaderboards .

This pricing pressure is forcing global operators to rethink their AI infrastructure partnerships and cost models.

Nokia's Agentic AI Push

Nokia made several major announcements at MWC Shanghai and the parallel DTW26 event:

• Google Gemini integration: Nokia announced it will integrate Google's Gemini models into the Nokia Assurance Center (its network monitoring and management platform) to bring AI agents into telecom network operations. Six specialized agents are planned, with two (Router and Event Triage) already operational .
• Autonomous networks portfolio upgrade: Nokia upgraded its autonomous networks portfolio with new agentic AI capabilities, including an Autonomous Networks Agent Library, the Autonomous Networks Suite, and MantaRay SMO — all aimed at improving network performance, reliability, and operational efficiency at scale .
• Fixed-network agentic AI: Nokia also launched agentic AI specifically for home and broadband networks, built on insights from 600+ million deployed broadband lines .
• AI-RAN with NVIDIA: Nokia continued its strategic AI-RAN partnership with NVIDIA, announcing new customer integrations with T-Mobile, Indosat, and SoftBank Corp., and successful functional tests of its anyRAN software on NVIDIA's GPU-accelerated AI-RAN platform .

The Big Picture: A Cross-Industry Consensus

Taken together, the announcements at MWC Shanghai 2026 reveal a strong consensus across the industry:

• Monetization must move beyond connectivity. Whether it's Huawei's token economy, ZTE's agentic core, or Nokia's AI slicing, the core message is that operators must capture value from AI workloads, not just data transport.
• Open collaboration is essential. The GSMA's Open Telco AI initiative, with its focus on shared datasets and models, reflects a recognition that no single operator can solve telco-grade AI alone.
• Cost pressure from Chinese models is reshaping the landscape. The availability of powerful, low-cost AI models from DeepSeek and others is creating both opportunities and competitive threats for telecom operators globally.
• Agentic AI is the next frontier. Both Huawei and Nokia are investing heavily in autonomous AI agents that can manage networks, handle customer experience, and deliver new services without human intervention.

For telecom operators, the message from MWC Shanghai 2026 was clear: the AI revolution is here, and the window for building a new revenue model — one based on tokens, agents, and intelligent infrastructure — is opening now.