Alibaba Cloud has unveiled its innovative CUBE DC 5.0 architecture, designed to meet the rising demand for AI infrastructure with improved performance and faster build times. Announced at the Apsara conference, the modular design promises a 50% reduction in deployment times compared to traditional methods. The architecture includes advanced technologies like a wind-liquid hybrid cooling system and an all-direct current power distribution, boosting both sustainability and efficiency.
While the cooling system’s details remain scarce, it likely involves a combination of cold plate cooling and airflow technologies to keep data centers running optimally. According to Alibaba Cloud CEO Eddie Wu, this initiative is part of the company’s broader mission to establish a global benchmark for AI infrastructure. “We’re not just meeting AI demand, we’re setting the standard for sustainability,” Wu said.
In addition to the CUBE DC 5.0, Alibaba introduced improvements to its Enterprise Elastic Compute Service (ECS), now in its ninth generation, which is 30% faster in recommendation engine tasks and offers a 17% boost in database operations. The company also launched its “Open Lake” data utility, integrating various big data engines for generative AI applications, and introduced the DMS: OneMeta+OneOps, capable of managing metadata across 40 sources.
The conference also saw the debut of the Qwen 2.5 multimodal AI models, offering up to 72 billion parameters and enhanced capabilities in math, coding, and natural language processing. Another highlight was the Tongyi Wanxiang text-to-video AI model, which can generate high-quality videos in diverse visual styles, from realistic to 3D animation, using both Chinese and English prompts.
With these developments, Alibaba Cloud aims to revolutionize how data centers are built and how AI infrastructures are deployed, making its solutions faster, smarter, and more sustainable than ever before.
