Advanced Micro Devices (AMD) has introduced Helios, its first rack-scale artificial intelligence system, marking its most significant move yet to compete with Nvidia in the rapidly expanding AI infrastructure market. The announcement comes with Microsoft confirming that it will deploy the Helios platform across its Azure cloud infrastructure, strengthening AMD's position among hyperscale cloud providers investing heavily in AI compute.
The Helios platform represents AMD's shift from supplying standalone processors to delivering integrated AI systems that combine graphics processing units, central processing units, networking and software into a unified architecture. The company said the platform is designed to support production-scale AI workloads, including inference, reasoning, search, agentic AI and data processing, areas that are becoming increasingly important as enterprises move beyond AI experimentation.
Microsoft said Azure will integrate AMD's Helios rack-scale solution alongside sixth-generation EPYC "Venice" processors and Instinct MI455X accelerators to power a new generation of cloud services. The company plans to use the infrastructure for frontier AI model inference while also introducing new virtual machine offerings for high-performance computing, chip design and large-scale AI applications. Scott Guthrie, Executive Vice President of Cloud and AI at Microsoft, said Azure's strategy is to provide customers with a heterogeneous infrastructure that combines Microsoft's own silicon with technologies from industry partners to optimise performance, cost and energy efficiency.
The partnership gives AMD another high-profile customer as it seeks to narrow the gap with Nvidia, which continues to dominate the AI accelerator market. Microsoft joins Meta, OpenAI and Oracle among the early adopters of Helios, with commercial shipments expected to begin later this year. Industry analysts view Microsoft's endorsement as an important validation of AMD's broader AI infrastructure strategy, particularly as cloud providers seek to diversify their hardware supply chains amid rising demand for AI computing capacity.
AMD Chief Executive Lisa Su has positioned Helios as the company's answer to the industry's transition towards rack-scale AI systems. Rather than focusing solely on individual AI chips, hyperscalers are increasingly adopting integrated compute platforms that combine accelerators, CPUs, networking and software to improve efficiency and simplify deployment across large data centres. Helios has been built around this approach, allowing customers to deploy AI infrastructure at scale through a single integrated platform.
The launch comes as competition in AI infrastructure intensifies. Nvidia remains the dominant player in AI data centre hardware, but companies including AMD are expanding their portfolios to address growing enterprise demand for training and inference infrastructure. AI workloads have evolved beyond model development to include agentic AI, retrieval, search, simulation and enterprise automation, creating demand for more specialised computing systems.
The announcement also lifted investor sentiment, with AMD shares rising following news of Microsoft's expanded partnership. Analysts said the collaboration demonstrates growing confidence in AMD's ability to compete in the AI infrastructure market and could strengthen its position as enterprises and cloud providers continue to invest in large-scale AI deployments.
With Helios, AMD is broadening its ambitions beyond semiconductor manufacturing to become a provider of complete AI infrastructure. As cloud providers expand their AI capabilities and enterprises scale production deployments, the competition between AMD and Nvidia is expected to increasingly centre on integrated AI systems rather than individual chips alone.