China Mobile Unveils National AI Computing Network Plan

China Mobile Bets Big on National AI Infrastructure

China Mobile, the world's largest mobile carrier by subscribers, has unveiled ambitious plans to build a nationwide integrated computing power network and launch trillion-level token service packages for AI developers and enterprises. Chairman Chen Zhongyue outlined the strategy at the China Mobile 2026 Mobile Cloud Conference on May 8, signaling a major push by China's state-backed telecom giant into the AI infrastructure race.

The announcement comes as China accelerates efforts to build domestic AI capabilities independent of Western technology, and positions China Mobile as a central player in what could become one of the world's largest unified AI computing ecosystems.

Key Takeaways

• China Mobile will build a national integrated computing power network offering on-demand, accessible AI compute services
• The company launched MoMA, a model service platform integrating over 300 mainstream AI models
• Trillion-level token service experience packages will be made available to developers and customers
• Gigawatt-scale data center campuses are planned, alongside accelerated AI data center construction
• MoMA supports China Mobile's proprietary 'Jiutian' foundation model alongside DeepSeek, Qwen, Doubao, Kimi, and GLM
• A pioneering token-based centralized operations model has been introduced — a first in China's telecom industry

Building a Unified Computing Power Network

China Mobile's vision centers on creating a seamless, nationwide computing infrastructure that provides what Chen described as 'instant access and universally affordable' AI computing services. This is not merely a data center expansion — it represents a fundamental rethinking of how computing resources are distributed and consumed across China's vast geography.

The plan includes several concrete infrastructure initiatives. The company will strengthen all-optical high-speed direct connections between hub nodes and high-demand regions, creating a backbone network optimized for AI workloads. This fiber-optic backbone is essential for reducing latency in distributed AI training and inference tasks.

Perhaps most striking is the commitment to building gigawatt-scale data center campuses. To put this in perspective, a single gigawatt-scale facility would rival the largest data centers currently operated by U.S. hyperscalers like Microsoft, Google, and Amazon. These facilities would be purpose-built for AI workloads, incorporating the latest cooling and power management technologies.

The initiative also includes upgrading AI cloud computing services and fostering what China Mobile calls a 'diversified, heterogeneous domestic computing ecosystem.' This language is significant — it signals a deliberate effort to build AI infrastructure around domestically produced chips and hardware, reducing reliance on NVIDIA and other Western semiconductor suppliers that face U.S. export restrictions.

MoMA Platform Aggregates 300+ AI Models

Alongside the infrastructure announcement, China Mobile launched its MoMA (Mobile Model-as-a-Service) platform, a comprehensive model service hub that aggregates over 300 AI models under a single, unified API gateway. This positions MoMA as a potential competitor to platforms like Amazon Bedrock, Microsoft Azure AI, and Google Vertex AI in the Chinese market.

The platform currently integrates several of China's most prominent AI models:

• Jiutian — China Mobile's proprietary foundation model
• DeepSeek — the open-source model that shocked global markets earlier this year
• Qwen (Tongyi Qianwen) — Alibaba's flagship large language model
• Doubao — ByteDance's AI model powering its consumer applications
• Kimi — Moonshot AI's long-context model popular among Chinese users
• GLM — Zhipu AI's bilingual language model series

MoMA's unified API approach means developers can switch between models or use multiple models simultaneously without rebuilding their applications. This model-agnostic design mirrors the strategy adopted by Western cloud providers but adds a distinctive feature: a centralized token operations model that China Mobile claims is a first in the industry.

The token centralization approach allows enterprises to purchase token packages that can be used across any model on the platform, rather than managing separate billing relationships with each model provider. This dramatically simplifies procurement and cost management for enterprise AI deployments.

Trillion-Token Packages Signal Aggressive Pricing Strategy

The offer of trillion-level token service experience packages represents an aggressive move to drive adoption and build market share. While specific pricing details were not disclosed at the conference, the scale of the offering suggests China Mobile is prepared to subsidize early adoption to establish its platform as the default AI infrastructure layer for Chinese enterprises.

This strategy parallels what has been happening in the U.S. market, where companies like OpenAI, Google, and Anthropic have engaged in fierce price competition. OpenAI, for example, has repeatedly slashed API prices — in some cases by more than 75% — to maintain market dominance. China Mobile's approach of bundling massive token allocations into experience packages takes this price war to a new level.

For context, 1 trillion tokens represents an enormous amount of AI processing capacity. A typical enterprise chatbot deployment might consume several billion tokens per month, meaning trillion-token packages could sustain significant AI operations for extended periods. The 'experience' framing suggests these may function as trial packages designed to onboard new customers.

Strategic Context: China's AI Infrastructure Race

China Mobile's announcements must be understood within the broader context of China's national AI strategy. Following the release of DeepSeek's R1 model in January 2025, which demonstrated that competitive AI performance could be achieved with fewer high-end chips, Chinese tech companies and state-backed enterprises have accelerated their AI infrastructure investments.

Several factors make China Mobile uniquely positioned in this landscape:

• Network reach: With over 980 million mobile subscribers and extensive fiber-optic infrastructure, China Mobile has unmatched distribution capabilities
• Government backing: As a state-owned enterprise, China Mobile has access to preferential financing, land, and regulatory support
• Edge computing: Its existing base station network provides natural edge computing nodes for AI inference
• Enterprise relationships: Deep existing relationships with government and enterprise customers provide immediate go-to-market channels

The company's push into AI computing also aligns with China's 'East Data, West Computing' national project, which aims to route computing demand from data-rich eastern cities to energy-abundant western regions. China Mobile's all-optical network upgrades directly support this initiative by ensuring low-latency connections between geographically distributed data centers.

What This Means for the Global AI Market

For Western companies and investors, China Mobile's aggressive entry into AI infrastructure carries several implications. First, it confirms that the global AI computing buildout is accelerating on both sides of the Pacific, with state-backed Chinese enterprises now matching — and potentially exceeding — the capital expenditure commitments of U.S. hyperscalers.

Second, the MoMA platform's model-agnostic approach could accelerate the commoditization of large language models in the Chinese market. By making it trivially easy to switch between competing models, China Mobile reduces the moat that any individual model provider can build, potentially compressing margins across the industry.

Third, the emphasis on domestic computing ecosystems signals continued divergence between Chinese and Western AI technology stacks. As China builds AI infrastructure around domestically produced chips from companies like Huawei (Ascend series) and Cambricon, the technical standards and optimization approaches may increasingly diverge from the NVIDIA CUDA ecosystem that dominates Western AI.

For developers building AI applications targeting the Chinese market, MoMA's unified API could become an important integration point. The platform's support for multiple models through a single gateway reduces the complexity of multi-model deployments and could establish new standards for how AI services are consumed in enterprise environments.

Looking Ahead: From Telecom Giant to AI Powerhouse

China Mobile's transformation from a traditional telecom operator to an AI infrastructure provider reflects a broader trend among global carriers seeking new revenue streams. In the West, companies like AT&T and Deutsche Telekom have explored similar pivots, though none have matched the scale of China Mobile's ambitions.

The company outlined 2 major initiatives for the near term. The 'Computing Power New Momentum Action' will deepen cooperation in resource construction, computing interconnection, and integrated services. Meanwhile, the 'Intelligent New Space Plan' will focus on computing networks, AI agents, and industrial internet applications.

These plans target 'hundreds of millions of customers and developers,' according to Chen's remarks — a scale that, if achieved, would make China Mobile one of the world's largest AI service providers by user count. Whether the company can execute on this vision while navigating U.S. chip export controls and building a competitive domestic technology stack remains the central question.

What is clear is that the AI infrastructure race is no longer solely a competition among Silicon Valley giants. With announcements like these, China's state-backed enterprises are staking their claim to a significant share of the global AI computing market — and they are bringing telecom-scale resources and distribution to the fight.