Alibaba Cloud targets full-stack AI dominance

Alibaba Cloud has fired the latest salvo in the global race for artificial intelligence (AI) supremacy with a slew of new AI models, agent development platforms and cloud infrastructure upgrades in a bid to become a full-stack AI service provider.

Unveiled at its annual Apsara conference in Hangzhou this week, the announcements, which included a RMB380bn ($53.4bn) investment in its AI and cloud infrastructure over the next three years, were capped by the release of Qwen3-Max, its most powerful large language model (LLM) to date with over one trillion parameters.

The company’s ambition is to create an ecosystem in which its AI models function as a new type of operating system and are incorporated into a wide range of devices, backed by a cloud platform optimised for every stage of AI development and deployment.

“We remain committed to open sourcing Qwen and shaping it into the ‘operating system of the AI era’, empowering developers around the world to build transformative AI applications,” said Eddie Wu, chairman and CEO of Alibaba Cloud Intelligence.

Key to the announcements is the new Qwen3 family of LLMs. Alibaba Cloud claimed that the performance of the flagship Qwen3-Max model is on par with top closed source models in key benchmarks such as SWE-Bench, which assesses an LLM’s ability to solve real-world software engineering problems.

The company has also enhanced its multimodal capabilities with Qwen3-VL, a vision-language model that can act as a visual agent, operating computer interfaces and generating code directly from visual designs.

Another addition is Qwen3-Omni, a native multimodal model that can handle text, image, audio and video inputs to deliver real-time, streaming spoken responses, making it ideal for hands-free interfaces in smart devices and vehicles.

Recognising that models alone are insufficient, Alibaba Cloud also made major moves to simplify the creation and deployment of AI agents for businesses.

Its AI development platform, Model Studio, now includes an agent development kit that lets professional developers create powerful AI agents in a high-code environment. Business users and people with limited programming skills can use a low-code agent development platform to build lightweight agents.

For more complex enterprise needs, the company rolled out major updates to AgentBay, its multimodal cloud operating environment, and Lingyang AgentOne, a one-stop platform designed to help organisations build and integrate production-ready agents across marketing, customer service and operations.

To power these advanced AI workloads, Alibaba Cloud has enhanced its cloud infrastructure, including a new vector bucket feature in its object storage service that unifies raw and vector data management, simplifying and lowering the cost of building retrieval augmented generation (RAG) applications.

The PolarDB database has also been enhanced with compute express link (CXL) technology, which reduces latency by 72% and increases memory scalability, making it suited for combined data and AI workloads.

In networking, Alibaba’s new HPN 8.0 architecture now provides 800Gbps of throughput, doubling its previous capacity to support ultra-large-scale model training. Additionally, the company’s cloud threat detection and response service now uses AI agents powered by Qwen to automate security operations and improve incident response.

The latest announcements signal a clear intent from Alibaba Cloud to compete aggressively with global hyperscalers such as Amazon Web Services, Microsoft Azure and Google Cloud, not only through competitive pricing, but also through a deeply integrated AI development and deployment platform to support the next wave of AI innovation.

Alibaba Cloud currently operates 91 availability zones in 29 cloud regions around the world. In the coming year, it expects to open its first datacentres in Brazil, France and the Netherlands, followed by additional facilities in Mexico, Japan, South Korea, Malaysia and Dubai. It is also establishing regional service centres in Indonesia and Germany to provide 24/7 multilingual customer support.