Alibaba Cloud rolls out agentic cloud stack with AgentCore and Agent Context

Alibaba Cloud has launched new services for deploying and securing enterprise AI agents, alongside infrastructure upgrades it says cut AI storage costs by up to 69%.

Alibaba Cloud has unveiled an agentic cloud stack aimed at enterprises deploying AI agents at scale, announcing the services at its annual Apsara Conference in Hangzhou.

Alibaba said the upgrades span three layers, with AI Native Cloud handling large-scale model training and inference, Agent Native Cloud covering agent deployment and security, and a Context Engine supplying agents with real-time data and long-term memory.

At the centre of the agent layer is AgentCore, a new platform for building and managing AI agents across their lifecycle. Alibaba said it lets businesses control how humans and agents work together and monitor agent performance over time. Alibaba said security is embedded at the operating layer, giving organisations a single foundation to deploy different agents and connect them to core business systems.

It sits alongside the Agent Security Center, which Alibaba said provides lifecycle security and compliance management for enterprise agent applications, using real-time threat detection to keep agents within set boundaries.

Context and data cost claims

Agent Context, part of the Context Engine, connects a company's documents, business systems, chat records and multimodal data so agents can remember past tasks and share knowledge across teams. Alibaba said it reduces token usage by up to 67% in knowledge-intensive work such as customer service and AI coding. Tokens are the units of text a model processes and the basis on which most AI services are billed.

The company also upgraded OpenLake into a data lakehouse where a single copy of structured, unstructured, vector and streaming data serves multiple compute engines. Alibaba said this cuts total costs by 38% and query response times by 40% compared with traditional architectures.

On infrastructure, Alibaba said its Cloud Parallel File Storage system delivers throughput of around 100 terabytes per second and can cut enterprise AI storage costs by 69%. Its HPN 8.0 Pro networking architecture supports more than 130,000 ports at 800G speeds in a single cluster. Its Platform for AI completed post-training of Qwen models in five days.

The performance and cost figures are Alibaba's own. The company did not give pricing or regional availability for the new services.

Alibaba targets 20GW of data centre capacity

The rollout comes as Alibaba expands its compute footprint. CEO Eddie Wu said machine thinking currently accounts for less than 3% of human thinking and still has substantial room to grow.

"With this in mind, our target is that by 2032, the global data center capacity operated by Alibaba Cloud will surpass 20GW, fueling the industry's exponentially rising demand for AI," he said.

Alibaba also launched Qwen Intelligence, an agent platform that gives smartphone makers access to Qwen-powered agents capable of handling complex cross-app tasks.