🔥36氪•Stalecollected in 6m
Shanghai Telecom Launches First Token-Based AI Billing Service
💡First telecom-integrated Token billing service: a new way to scale AI app monetization.
⚡ 30-Second TL;DR
What Changed
Shanghai Telecom is the first operator to launch a Token-based billing package.
Why It Matters
This model could standardize AI compute pricing for the mass market and lower the barrier for developers to integrate LLMs into consumer-facing applications.
What To Do Next
Evaluate if your application can leverage telecom-integrated billing to reduce friction for your end-users.
Who should care:Developers & AI Engineers
Key Points
- •Shanghai Telecom is the first operator to launch a Token-based billing package.
- •Pricing is set at 1 RMB for 250,000 tokens (based on Mixtral-K2.5 model).
- •The service supports direct payment via mobile phone bills for easy access.
🧠 Deep Insight
Web-grounded analysis with 18 cited sources.
🔑 Enhanced Key Takeaways
- •The service provides users with access to over 30 mainstream large-scale models, including both text and multimodal AI, through standard API interfaces.
- •Shanghai Telecom offers flexible pricing options, including pay-as-you-go and discounted bulk purchases, and provides a one-month free trial of 25 million tokens for mobile users.
- •This initiative is part of a broader strategic transformation by China's major telecom operators to shift from traditional data traffic monetization to 'Token operations' to address the 'dumb pipe' dilemma in the AI era.
- •The pricing strategy is positioned aggressively, with operator token plans reportedly priced around 18% lower than leading cloud service providers' API call rates.
- •The underlying model referenced for the 250,000 token benchmark is likely Kimi-K2.5 from Moonshot AI (a probable correction from 'Mixtral-K2.5' in the original article), a 1-trillion parameter multimodal Mixture-of-Experts model released in January 2026.
🛠️ Technical Deep Dive
- The service provides access to over 30 mainstream large-scale models (text and multimodal) via standard API interfaces.
- The 250,000 token benchmark is based on the Kimi-K2.5 model (likely a correction from "Mixtral-K2.5" in the original article), developed by Moonshot AI.
- Kimi-K2.5 is a 1-trillion parameter Mixture-of-Experts (MoE) multimodal model with 32 billion active parameters per token.
- It was pre-trained on approximately 15 trillion mixed visual and text tokens and features native multimodality, Agent Swarm technology (coordinating up to 100 specialized AI agents), and a 256K context window.
- Shanghai Telecom's parent company, China Telecom, is building an intelligent cloud ecosystem centered on its XiRang integrated intelligent computing service platform, which includes Infrastructure as a Service (IaaS), Platform as a Service (PaaS), Data as a Service (DaaS), Model as a Service (MaaS), and Application as a Service (SaaS) layers.
- China Telecom operates over 6,000 edge data centers and 900 large-scale data centers, and by December 2025, reported self-owned intelligent compute capacity of 54 EFLOPS, with total scheduled and operated compute exceeding 85 EFLOPS through its XiRang platform.
- The company is deploying 40,000 high-power racks in Shanghai for intelligent computing and supercomputing, with a focus on high-end GPUs.
🔮 Future ImplicationsAI analysis grounded in cited sources
Telecom operators will significantly disrupt the AI API market by leveraging their existing infrastructure and billing relationships.
Their ability to offer competitive pricing, direct mobile bill payment, and integrate AI services into existing communication packages provides a unique advantage over traditional cloud providers.
The 'token operations' model will become a standard for monetizing AI compute and models, especially in regions with strong telecom infrastructure.
This shift addresses the 'dumb pipe' dilemma for telcos and aligns with tokens as the core workload metric for AI, providing a tangible economic unit for AI consumption.
China's telecom operators will become dominant players in providing AI inference capacity.
They are making significant investments in AI computing infrastructure, including large-scale data centers and edge nodes, and are strategically shifting procurement towards inference-type equipment.
⏳ Timeline
2016
China Telecom released its Transformation 3.0 strategy and CTNet2025 network architecture reconstruction plan, aiming to embed AI.
2019
China Telecom released its AI Development White Paper 1.0 during MWC Shanghai, outlining plans for a universal AI capability platform.
2024-10
Shanghai Telecom's Smart Cloud Shanghai project was awarded the Gigabit City Innovation Award by the World Broadband Association (WBBA).
2025-03
China Telecom showcased its XiRang integrated intelligent computing service platform and '1+1+1+M+N' AI system at MWC 2025.
2026-03
China Telecom announced plans to put AI tokens at the core of its business, signaling a transformation to token value-based operations.
2026-05-16
Shanghai Telecom officially launched its token computing power service.
📎 Sources (18)
Factual claims are grounded in the sources below. Forward-looking analysis is AI-generated interpretation.
- Google Search Source
- Google Search Source
- Google Search Source
- Google Search Source
- Google Search Source
- Google Search Source
- Google Search Source
- Google Search Source
- Google Search Source
- Google Search Source
- Google Search Source
- Google Search Source
- Google Search Source
- Google Search Source
- Google Search Source
- Google Search Source
- Google Search Source
- Google Search Source
📰
Weekly AI Recap
Read this week's curated digest of top AI events →
👉Related Updates
AI-curated news aggregator. All content rights belong to original publishers.
Original source: 36氪 ↗