AMD ‘헬리오스’ 애저 탑재…마이크로소프트, 엔비디아 대항마 택했다 – 테크42

Microsoft is integrating AMD’s Helios, the company’s first rack-scale AI system, into its Azure cloud platform. According to Tech42, this move follows similar adoptions by Meta, OpenAI, and Oracle, positioning the Helios system as a direct competitor to Nvidia’s dominant AI infrastructure.

AMD Helios Integration in Azure

The adoption of Helios by Microsoft marks a shift in the hardware composition of Azure’s AI clusters. Helios is designed as a rack-unit system, meaning it integrates the processors, networking, and power delivery into a single scalable unit rather than selling individual accelerators. Tech42 reports that Microsoft is utilizing this architecture to diversify its AI compute options away from a sole reliance on Nvidia hardware.

By deploying Helios, Microsoft joins a growing list of high-scale AI users. The system has already seen implementation by Meta, OpenAI, and Oracle, who are seeking alternatives to Nvidia’s GPU ecosystems for training and deploying large-scale AI models.

Technical Shift to Rack-Scale Architecture

Helios represents AMD’s transition from selling standalone GPUs to providing full-rack AI solutions. A rack-scale system optimizes the interconnects between multiple accelerators, reducing latency and improving data throughput across the entire cluster. This approach allows cloud providers like Microsoft to deploy massive compute blocks more efficiently than assembling individual server nodes.

The system is built to challenge Nvidia’s vertical integration, where Nvidia provides not only the H100 or B200 chips but also the NVLink interconnects and the software layers that manage them. AMD’s Helios aims to provide a comparable end-to-end hardware stack for the Azure environment.

Market Impact and Competitive Landscape

Microsoft’s decision to deploy Helios reflects a broader industry trend of “chip diversification.” Because the demand for AI compute exceeds the current supply of Nvidia chips, major cloud service providers are integrating alternative silicon to ensure operational continuity and price leverage.

The adoption by OpenAI is particularly significant given its deep partnership with Microsoft. If OpenAI’s workloads can run efficiently on AMD’s rack-scale systems, it validates Helios as a viable alternative for the most demanding generative AI tasks currently in the market.

According to the report from Tech42, the deployment of Helios in Azure is a strategic move to counter Nvidia’s market grip by offering a high-performance, scalable alternative that integrates directly into the cloud’s existing infrastructure.