At the 2025 Alibaba Cloud Analyst Summit and Apsara Summit in Hangzhou, Alibaba Cloud reinforced its position as an AI‑native cloud provider, delivering a full‑stack AI architecture and expanding its global ecosystem. The announcements also outlined a plan for Artificial Super Intelligence (ASI). Though ASI may be a distant goal, Alibaba Cloud has adopted a practical approach to AI-native cloud, focusing on AI-optimized infrastructure, model platforms, global expansion, and open-source innovation.
FullStack AI: Building An AI-Native Cloud
Forrester describes an AI-native cloud as a system in which AI is incorporated as an architectural principle throughout infrastructure, platforms, and applications. Alibaba Cloud’s announcements at Apsara 2025 align with this vision by integrating AI into every layer of its stack. The company introduced a new datacenter fabric, compute services, model training platforms, and MaaS offerings to support largescale AI workloads. Here are the main announcements across:
Compute infrastructure. Alibaba unveiled Panjiu AI Infra 2.0 and HPN 8.0, which provide ultra-low latency and high reliability for massive GPU clusters, enabling trillion parameter model training and inference at scale. The compute layer gained enhancements with Elastic GPU Service (EGS) for bare metal GPU instances, ACK with GPU node auto healing, and Alibaba Cloud Linux 3 optimized for containerized AI workloads. Storage improvements include CPFS with RDMA acceleration and OSS bandwidth upgrades for high throughput data pipelines.
Data and AI development. PAI/Model Studio now integrates NVIDIA’s Physical AI stack to support robotics, simulation, and digital twin scenarios. Alibaba also introduced vector and multimodal database capabilities in AnalyticDB and OceanBase to enable retrieval augmented generation and domain specific fine- The MaaS portfolio expanded with Qwen3‑Max, a trillion‑parameter LLM, Wan 2.5 for image and video generation, and Fun speech models. Alibaba’s Qwen models are fueling agentic AI adoption in China’s financial sector, led by ICBC’s “Zhiyong” system. The Bailian platform added agent development kits, governance features, and AI Guardrails for compliance. And ModelScope’s open-source AI models have been downloaded over 600 million times.
Securing AI-native cloud. Alibaba Cloud is implementing a comprehensive security strategy for the agentic AI era, introducing a machine identity management framework with IDaaS and KMS for secure resource access. Its multi-agent system automates threat detection and incident response, while Qwen models enhance identity verification to protect against deepfake fraud. This dual approach secures both AI agents and operations in an AI-native cloud environment.
Ecosystem And Globalization: Open Source And AI Partnerships
Chinese cloud vendors are accelerating globalization through ecosystem strategies that combine open‑source leadership, strategic partnerships, and regional expansion. The announcements during the event highlight this approach, with a focus on developer adoption, embodied AI innovation, and compliance‑driven localization. These initiatives aim to strengthen Alibaba Cloud’s competitive position in international markets while mitigating supply‑chain and regulatory risks. Major moves include:
Aggressive global expansion. Alibaba Cloud announced new regions in Brazil, France, and the Netherlands, along with additional service centers in Indonesia and Germany, expanding its footprint beyond 91 availability zones across 29 regions.
Strengthened global partnerships. Alibaba Cloud deepened its partnership with NVIDIA by integrating the Physical AI software stack into PAI, enabling developers to build robotics, autonomous systems, and digital twin applications in a cloud‑native environment.
Continuous open-source commitment. Alibaba Cloud continued its open-source momentum with Qwen and Wan models released on ModelScope and Hugging Face, driving community adoption and enterprise portability across global markets.
Proven industrial collaboration. These collaborations continue to showcase real‑world impact, including AstraZeneca for pharmacovigilance and Shiseido for AI‑driven security. And Japanese partners like GladCube and FLUX also help with the fine‑tuning Qwen for local finance and marketing use cases.
Recommendations: A Pragmatic Road To AGI and ASI
While ASI is still too far away, and AGI is not close either, enterprises should take a structured approach to adopting AI‑native cloud capabilities while balancing innovation with operational resilience. Alibaba Cloud’s roadmap offers opportunities, but success depends on aligning architecture, governance, and ecosystem engagement with business priorities.:
Architect for AI‑native capabilities across the full stack. Enterprises should map workloads to AI‑optimized infrastructure, including high‑performance networking for distributed training, RDMA‑enabled storage for checkpointing, and vector databases for RAG. This approach ensures scalability and cost efficiency while reducing integration complexity. Treat AI as a core design principle rather than a bolt‑on feature to future‑proof your architecture.
Adopt agent‑centric and open‑source strategies for flexibility. Use agent frameworks to design modular applications that can orchestrate multiple models and tools without heavy refactoring. Combine this with open‑source models for portability and cost control, while leveraging managed services to simplify operations. This dual approach balances innovation speed with governance and sovereignty requirements.
Embed AI governance into your roadmap. Implement AI guardrails to meet multi‑jurisdictional regulations, and Forrester’s AI governance framework can help. At the same time, prepare for embodied AI by incorporating simulation and digital twin techniques into industrial workflows, if robotics is a core business focus.
If you’d like to dive deeper set up an inquiry or guidance session with Charlie Dai (AI-native cloud, agentic AI, and humanoids), and Meng Liu (enterprise fraud management, identity verification, fintech, security and risk in financial services) for a conversation.
