Introduction

Forrester, a global research and advisory firm, released "The Forrester Wave™: AI Infrastructure Solutions, Q1 2024". Alibaba Cloud, with its advanced product performance, got the second highest score in the current offering category.

This report is an authoritative assessment of global AI infrastructure, evaluating the top 12 technology companies in the field as identified by Forrester. The companies were scored across 19 criteria spanning three dimensions: current offering, strategy, and market performance. Alibaba Cloud received the highest score possible of 5 in the criteria of Workloads-Data, Tools-Management, Tools-Development, Deployment-Fault Tolerance, and Deployment-Efficiency within the current offering category.

"AI infrastructure is designed to satiate AI’s need for three core AI workloads: data preparation, model training, and model inferencing." The Forrester report states. Alibaba Cloud has built a rich array of cloud AI services encompassing the entire AI development process, featuring both robust AI computing power and convenient platforms and tools. Additionally, the cost-effectiveness and ease of use of its products are particularly notable.

Currently, Alibaba Cloud has formed the most comprehensive AI infrastructure products and services in China: At the IaaS layer, the Lingjun Smart Computing Cluster utilizes the HPN 7.0 network architecture, supporting a scalable size of up to 100000 cards, while the storage system CPFS provides up to 2TB/s read throughput and 30 million IOPS to meet exponentially growing AI demands. At the PaaS layer, the AI platform PAI has a distributed training acceleration ratio of up to 96%, capable of running large-scale tasks for over five weeks stably, and its full suite of big data products have been upgraded for AI scenarios, offering computing at the 10PB level, end-to-end data governance, and vector search capabilities.At the MaaS layer, ModelScope community gathers over 3,000 high-quality open-source models and thousands of datasets, and Alibaba Cloud's Model Studio platform helps enterprises quickly create custom models and AI applications.

Alibaba Cloud's AI infrastructure also boasts an excellent product experience, supporting a serverless billing model that can be activated with one click, allocated on-demand, and operated with minimal oversight. This flexibility in invoking AI computing resources in large-scale training and inference scenarios significantly reduces customer usage costs.

Currently, Alibaba Cloud has become the public base for large-scale models in China. A large number of enterprises and institutions, including Baichuan Intelligent, Zhipu AI, Zero One Infinite, Kunlun Universe, VIVO, Fudan University, and Giant Network, train large models on Alibaba Cloud and provide services to the outside world through Alibaba Cloud. For example, Baichuan Intelligent, which released seven models last year, not only efficiently completed the training tasks of large-scale models through deep cooperation with Alibaba Cloud but also effectively reduced the cost of model inference.

Forrester believes that AI infrastructure solutions customers should look for providers that maximize the performance of core AI workloads, offer a management layer to optimize cost and tame complexity, and match enterprise’s commitment to AI. Alibaba is a good fit for existing Alibaba customers and for new customers that need to run cloud-scale AI workloads in China. Alibaba reference customers appreciate its ability to scale AI infrastructure to handle huge AI workloads and the breadth of other services to deploy AI applications.