TMTPOST — Huawei Cloud officially launched its next-generation model training and inference platform, ModelArts Next, at its INSPIRE conference on Friday, introducing a unified infrastructure framework to accelerate the corporate deployment of autonomous artificial intelligence agents. The updated Model-as-a-Service (MaaS) architecture deploys four core technical capabilities: automated reinforcement learning services, hardware-isolated confidential inference, dynamic model routing, and a multi-model software matrix. To minimize commercial execution costs, the platform's intelligent routing engine dynamically shifts enterprise workloads across alternative computing strategies—prioritizing either cost, performance, or balanced profiles—which drops average API invocation overhead by 20%. The infrastructure launch reflects an aggressive structural pivot within the Chinese Mainland’s cloud computing sector, where hyperscalers are transitioning from simple raw compute provision toward specialized, secure artificial intelligence execution networks. By embedding hardware-level cryptographic trust zones that prevent data exfiltration alongside day-zero integration for external flagship models like DeepSeek and Kimi, cloud operators are building high-margin, enterprise-ready software foundations tailored for high-sensitivity commercial sectors like digital finance.