Beijing to Layout Token Factories, Aiming to Add 50,000P Intelligent Computing Power in H2 北京：将布局建设Token工厂，力争下半年新增智能算力5万P

Summary

In the first half of the year, the value-added growth rate of Beijing’s core digital economy industries reached 9.8%, significantly driving GDP growth. In the second half of the year, plans include establishing “Token factories” and distribution platforms, along with formulating Token economy development policies to foster multi-sector innovation. Leveraging the Open Source Chip Research Institute, a full-stack autonomous technology system based on “RISC-V + AI OS” will be built to create an ecosystem spanning from chips to intelligent agents. The plan also includes adding 50,000 PetaFLOPS of intelligent computing power in the second half of the year, bringing the total annual capacity to exceed 130,000 PetaFLOPS, thereby refining the “super node + industry node” architecture.

Deep Analysis

TL;DR

• In the first half of the year, the value-added growth rate of Beijing’s core digital economy industries reached 9.8%, significantly driving GDP growth.
• In the second half of the year, plans include establishing “Token factories” and distribution platforms, along with formulating Token economy development policies to foster multi-sector innovation.
• Leveraging the Open Source Chip Research Institute, a full-stack autonomous technology system based on “RISC-V + AI OS” will be built to create an ecosystem spanning from chips to intelligent agents.
• The plan includes adding 50,000 PetaFLOPS of intelligent computing power in the second half of the year, bringing the total annual capacity to exceed 130,000 PetaFLOPS, thereby refining the “super node + industry node” architecture.

Why It Matters

This article reveals the latest strategic layout of local governments regarding AI infrastructure, particularly the inaugural proposal of the “Token factory” concept. This marks a new industrialization stage where data elements converge with computing power. For industry professionals focused on regional AI policy trends, computing supply chains, and autonomous, controllable technology routes, this serves as a significant barometer.

Technical Analysis

• Token Economy Infrastructure: Beijing has explicitly outlined plans to establish Token factories and distribution platforms around the key stages of Token production, distribution, and application. This implies that the standardized production and circulation of high-quality Token data for future AI training and inference will become an independent infrastructure component.
• Full-Stack Autonomous Technology System: By leveraging the Open Source Chip Research Institute and the Tongming Lake Information Innovation Center, an ecosystem based on “RISC-V + AI OS” will be constructed. This approach aims to connect the entire chain from underlying chip instruction sets and operating systems to upper-layer agent applications, emphasizing a combination of technological autonomy and open-source openness.
• Expansion of Computing Supply: A dual-layer support system comprising “super nodes + industry nodes” will be adopted. The goal is to add 50,000 PetaFLOPS of intelligent computing power in the second half of the year, pushing the annual total beyond 130,000 PetaFLOPS. Here, “P” typically refers to PFLOPS (petaflops, or quadrillions of floating-point operations per second), reflecting continued investment in large-scale parallel computing capabilities.

Industry Implications

• Accelerated Assetization of Data Elements: The introduction of “Token factories” signifies that data is no longer merely raw material but becomes high-value assets processed into standardized formats ready for model training. Relevant enterprises should focus on commercial opportunities in data cleaning, labeling, and Tokenization services.
• Diversification of Domestic Computing Ecosystems: The integration of RISC-V with AI OS offers a new pathway to reduce reliance on single architectures. Investors and developers should closely monitor developments in AI acceleration chips and supporting software stacks based on the RISC-V architecture.
• Decentralization and Specialization of Computing Resources: The “super node + industry node” model indicates that computing supply will extend from centralized supercomputing centers to industry-specific nodes. Vertical sectors (such as industry and healthcare) may require low-latency, highly specialized computing support closer to their specific scenarios.

TL;DR

• 北京上半年数字经济核心产业增加值增速达9.8%，对GDP增长带动显著。
• 下半年将布局建设“Token工厂”及分发平台，制定Token经济发展政策以推动多领域创新。
• 依托开源芯片研究院打造“RISC-V+AI OS”全栈自主技术体系，构建从芯片到智能体的生态。
• 计划下半年新增智能算力5万P，使年内算力总规模突破13万P，完善“超级节点+行业节点”体系。

为什么值得看

本文揭示了地方政府在AI基础设施层面的最新战略布局，特别是首次提出“Token工厂”概念，标志着数据要素与算力结合进入产业化新阶段。对于关注区域AI政策导向、算力供应链以及自主可控技术路线的行业人士而言，具有重要的风向标意义。

技术解析

• Token经济基础设施：北京明确提出围绕Token的生产、分发和应用关键环节，布局建设Token工厂和分发平台。这暗示了未来AI训练与推理中，高质量Token数据的标准化生产与流通将成为独立的基础设施环节。
• 全栈自主技术体系：通过开源芯片研究院和通明湖信创中心，构建“RISC-V+AI OS”生态。这一方案旨在打通从底层芯片指令集、操作系统到上层智能体应用的全链路，强调技术自主可控与开源开放的结合。
• 算力供给扩容：采用“超级节点+行业节点”的双层支撑体系。目标是在下半年新增5万P智能算力，使全年总量突破13万P。这里的“P”通常指PFLOPS（每秒千万亿次浮点运算），体现了对大规模并行计算能力的持续投入。

行业启示

• 数据要素资产化加速：“Token工厂”的提出意味着数据不再仅仅是原材料，而是经过标准化处理、可直接用于模型训练的高价值资产。相关企业应关注数据清洗、标注及Token化服务的商业化机会。
• 国产算力生态多元化：RISC-V与AI OS的结合为摆脱单一架构依赖提供了新路径。投资者和开发者应密切关注基于RISC-V架构的AI加速芯片及配套软件栈的发展动态。
• 算力资源下沉与专业化：“超级节点+行业节点”模式表明算力供给将从集中式超算中心向行业专用节点延伸。垂直行业（如工业、医疗）可能需要更贴近场景的低延迟、高专用性算力支持。