AMD challenges Nvidia's monopoly with the new Helios integrated AI system

American tech giant AMD is taking a serious step in the battle for leadership in the artificial intelligence market. The company introduced Helios – a new integrated computing system designed specifically for data centers dealing with the training and management of large-scale AI models.

Unlike individual components, Helios is a comprehensive platform that combines four key AMD technologies: AI graphics processing units (GPUs), central processing units (CPUs), networking solutions, and specialized software. The goal is to offer a more efficient and cheaper alternative to Nvidia's solutions, which currently control nearly 95% of the AI accelerator market.

One of the first major customers of Helios will be Microsoft, which plans to integrate the system into its Azure cloud platform to support advanced AI services. Other industry giants such as Meta, OpenAI, Oracle, and Tata Consultancy Services (TCS) are already showing interest in the new technology.

Technically, Helios relies on the new generation of MI400 series graphics accelerators in combination with powerful EPYC processors. The system is structured through 18 computing modules, with each module including four AI GPUs and one CPU. AMD is betting on cost optimization, aiming to reduce the resources required to process each individual request to the AI models.

Although Nvidia maintains a massive advantage through its CUDA software ecosystem, AMD hopes that the combination of hardware and software in Helios will attract companies looking to diversify their suppliers. For AMD, data centers are already a key driver of growth, and the company predicts that from 2027, AI solutions will generate tens of billions of dollars in revenue.