AMD and Microsoft Expand AI Collaboration, Helios Directly Challenges NVIDIA

AMD (AMD-US) announced on Monday (20th) that it will expand its long-term strategic partnership with Microsoft (MSFT-US). Microsoft Azure will be the first cloud platform to deploy the next-generation AI rack system 'Helios' as the core computing infrastructure for large-scale artificial intelligence (AI) model inference, Azure AI services, and enterprise customer applications. This marks the first adoption of Helios by Microsoft, joining major customers such as Meta (META-US), OpenAI, and Oracle (ORCL-US), signaling AMD’s formal challenge to NVIDIA (NVDA-US), which has long dominated the AI rack market.

AMD stated that Helios will begin shipping in the second half of this year, with customers including Microsoft deploying it progressively. However, the two companies did not disclose the financial value or procurement scale of the collaboration.

Microsoft’s integration of Helios will comprehensively upgrade Azure’s AI infrastructure.

According to AMD’s press release, Helios integrates the Instinct MI455X GPU, the sixth-generation EPYC 'Venice' CPU, Pensando networking chips, and the ROCm open-source software platform, supporting large-scale AI training and inference workloads. Microsoft will leverage Helios to support cutting-edge AI model inference, Azure AI services, and enterprise customer applications. Additionally, Microsoft will launch two new virtual machines (VMs) powered by the Venice processor, targeting workloads such as Agentic AI, data processing pipelines, and semiconductor design.

Furthermore, both companies will deepen the integration of AMD Pensando DPUs with Azure Boost technology to enhance Azure’s cloud networking performance and large-scale computing capabilities.

Satya Nadella, CEO of Microsoft, stated, 'Enterprise customers need infrastructure that can handle diverse AI workloads including training, inference, data processing, and search. Through our collaboration with AMD, Azure will deliver higher performance, greater scale, and more diverse AI computing options.'

This announcement continues the companies’ long-standing partnership. AMD chips have long powered Microsoft’s Surface laptops and Xbox gaming consoles. In 2023, Microsoft was the first to adopt AMD’s MI300X GPU, and now with the integration of Helios, the collaboration is deepening further.

Helios is AMD’s first rack-scale AI system and its first direct challenge to NVIDIA’s Grace Blackwell and Vera Rubin platforms. The system integrates GPU, CPU, networking, and software to deliver a complete AI computing platform, rather than a single chip product.

Forrest Norrod, head of AMD’s Data Center Group, said the goal in developing Helios was to achieve the lowest total cost of ownership (TCO) and the lowest inference cost per token. Customer feedback indicates that Helios has already met this goal.

Currently, NVIDIA controls over 95% of the data center GPU market, while AMD holds around 4.5% market share. However, research firm Futurum Group believes that if Helios achieves volume production, AMD could increase its market share to 20%–25%, representing a multi-hundred-billion-dollar market opportunity.

Analysts note that in addition to Meta, OpenAI, and Oracle, companies such as Tata Consultancy Services (TCS) in India and SpaceX’s AI division under Elon Musk (SPCX-US) have also adopted AMD’s Instinct GPUs. Meta plans to deploy up to 6GW of AMD GPUs long-term, with the first 1GW to be installed in Helios racks.

However, the market believes AMD’s real challenge remains its software ecosystem. Neil Shah, analyst at Counterpoint Research, stated, 'Helios’ GPU and CPU hardware performance can now rival NVIDIA’s, but CUDA still has a more mature developer ecosystem and software advantage—this is the key area AMD must overcome.'

AMD CEO Lisa Su said, 'Over the past few years, we have consistently invested across GPU, CPU, networking, and software to build a complete AI platform. With Helios now entering mass production and shipment, and major cloud providers like Microsoft adopting it, AMD is poised to expand its footprint in the AI data center space and close the gap with NVIDIA.'