Report: China's AI Inference Data Volume First Exceeds Training Data Volume
CCTV reported on November 2 that a report from the China Center for Information Industry Development, released October 31, shows China's AI inference data volume has for the first time exceeded training data volume, marking the shift from model-training R&D to scaled industrial application. By June 2026 average daily token calls surpassed 175 trillion, over 1,000 times the early-2024 level, and top-model iteration cycles compressed from months to weeks or days. China has built more than 500 excellent-grade smart factories across 31 provinces covering over 80% of manufacturing categories, with domestic compute and large models deployed in 20 industries across 1,000-plus application scenarios, enabling personalized customization and full-lifecycle services. The report said cross-industry fusion of high-end and digital manufacturing now contributes over half of the growth in above-scale industrial value added.
Source: CCTV News / Huanqiu, Nov 2, 2026 | 2026-11-02
- Apple Partners With Alibaba to Embed Qwen Into China Apple Intelligence
- AgiBot Ships Embodied Model Genie G2 Max; Origin Robot's World Model Tops Benchmark
- Domestic AI Software and Hardware Breakthroughs: SJTU Rebuilds Industrial Code With MiniMax, X7 Ultra Chip, JD Robots
- Ant's LingBot Open Source Tops 2.2M Downloads; Compute Tokens Debut as a Double 11 Product Category
- iFlytek's Liu Qingfeng: Keep Innovating on an Autonomous, Secure Tech Base
- People's Daily Overseas: 'China Computes, and It Works' on the Global Stage
- 2026 GAIC Conference Opens: SOE AI Scenario Co-Creation Plan and Physical-AI Benchmarks Unveiled
- China's First Open-Source Themed Pavilion Opens in Wuzhen for World Internet Conference
- DeepSeek Open-Sources DeepInfer: 'AI Directs AI' Reshapes the Industry Stack
- From 'World Supermarket' to 'Digital-Trade Hub': The Yiwu Vitality Blueprint
