Rack-scale systems combine multiple processors into a single high-performance computing unit and are primarily used in data centres to train and deploy large AI models. Nvidia has long dominated this segment with its Grace Blackwell and Vera Rubin platforms, but AMD is now directly targeting that market with Helios.

According to industry reports, Helios outperforms Nvidia’s Vera Rubin system across several performance metrics, potentially giving AMD a stronger foothold in the AI infrastructure race.

Major AI companies back the new platform

Although Helios was first unveiled in 2025 and showcased again at CES 2026, AMD confirmed that the system will begin shipping later this year.

Several leading AI companies have already committed to deploying the platform. Microsoft will integrate Helios into its Azure cloud infrastructure, while Anthropic has entered a strategic partnership with AMD to deploy up to two gigawatts of GPUs using the new rack-scale system. OpenAI, Meta and Oracle are also among the announced customers.

New Venice-X processor also announced

Alongside Helios, AMD introduced its upcoming Venice-X CPU, a processor aimed at data centres and high-performance computing workloads. The new chip is scheduled to launch in 2027 and is expected to complement AMD’s expanding AI hardware portfolio.

AMD expects AI chip market to reach $1.4 trillion

During her keynote, Lisa Su said the rapid growth of agentic AI is dramatically increasing computing demand, as AI agents perform multiple reasoning steps, access tools and repeatedly process information before completing tasks.

AMD now expects the global AI accelerator market to reach approximately $1.4 trillion by 2030, making it comparable in size to today’s entire semiconductor industry.

Su added that GPUs will continue to account for the majority of that market because AI algorithms are still evolving rapidly, making programmable GPU architectures the preferred choice for future AI workloads.