● Covered by 4 sources · 6 reportsMedium impact4 neutral
Microsoft to Deploy AMD's Helios AI Accelerator on Azure
Key points
- Microsoft will use AMD's Helios AI system on Azure.
- Helios competes with Nvidia's AI server offerings.
- The deployment is aimed at handling AI workloads.
- The integration supports Azure AI services.
- Launch highlights rise in AI computational needs.
- Helios uses Radeon Instinct MI455X and Epyc Venice processors.
- Helios is AMD's first rack-scale AI accelerator system.
- Microsoft will use Helios for frontier model inference.
- AMD will begin shipping Helios to customers later this year.
- AMD and Cerebras will partner to combine EPYC processors in Helios with Cerebras' Wafer-Scale Engine solutions.
- The MI455X GPU has 320 billion transistors and uses HBM4 memory.
- The Helios rack-scale architecture integrates 72 MI455X GPUs.
- Microsoft will use Helios for internal and external use.
- Helios will underpin managed compute for enterprise customers through Microsoft Foundry.
- Microsoft announced its commitment to Helios on Monday.
- Microsoft will add two new computing instances run on AMD's Venice CPUs.
- Helios is the first rival to Nvidia's Grace Blackwell and Vera Rubin systems.
- AMD and Cerebras expect their platform to deliver up to 5X higher tokens per second per watt.
- The MI455X GPU uses CDNA 5 architecture.
- The MI455X GPU is built using advanced packaging technologies.
- The MI455X GPU has four Accelerator Complex Dies (XCDs) stacked on a Fabric and Cache Die (FCD).
- The MI455X GPU uses TSMC’s CoWoS-L technology.
- The MI455X GPU uses TSMC 2N gate-all-around (GAA) process technology on the XCDs.
- AMD also introduced the Venice-X CPU for data centers.
- AMD Chair and CEO Dr. Lisa Su promoted Helios at the Advancing AI conference.
- Helios is built to train and run the most demanding frontier models at massive scale.
- Helios will be deployed by leading AI companies at gigawatt-scale.
- Microsoft will deploy Helios in volume.
- Microsoft CEO Satya Nadella announced the commitment in a press release.
- Helios is the largest system AMD has built so far.
- The MI455X replaces the Instinct MI355X.
- Helios scales to 72 GPUs in a single pod, up from 8 GPUs for previous MI355X systems.
- Helios features a new fault-tolerant scale-up networking architecture based on UALink over Ethernet (UAEoL).
- A single MI455X contains 256 Work Group Processors (WGPs) across 8 Accelerator Complex Dies (XCDs).
- The MI455X has a max engine clock of 2.4GHz.
- The MI455X enables peak compute figures up to 40.26 PFLOP for OCP MXFP4.
- The MI455X has 12 stacks of HBM4 each on a 2048 bit bus for a total of 192 channels.
Overview
Microsoft has announced plans to incorporate AMD's Helios rack-scale AI accelerator into its Azure cloud infrastructure. This deployment aims to enhance AI computing capabilities for Microsoft's internal use and Azure's external customer base.
The Helios system, developed by AMD, is positioned as a competitor to Nvidia's AI server solutions. It will enable Azure customers to handle complex AI workloads and power frontier-model inference applications.
Background
AMD's Helios is the company's first rack AI system. Microsoft joins other industry giants like Meta, OpenAI, and Oracle in utilizing AMD's technology amid increasing demands for AI computational resources.
The system is set to be used in Azure's data centers, reflecting broader trends in the tech industry towards more powerful AI infrastructure solutions.
Why It Matters
The collaboration between Microsoft and AMD is significant as it brings a new player into the competitive AI infrastructure market, challenging Nvidia's dominance.
The increased AI compute capacity will benefit Azure customers, particularly those in AI labs and enterprises needing advanced AI capabilities for training and inference workloads.
Takeaways
The integration of AMD's Helios into Azure not only bolsters Microsoft's cloud offering but also signals a shift in the AI infrastructure landscape, with AMD emerging as a viable contender against Nvidia. The partnership underscores the rapid growth and demand in AI computational power.
Updates
- Microsoft will deploy Helios in volume.
- Microsoft CEO Satya Nadella announced the commitment in a press release.
- Helios is the largest system AMD has built so far.
- The MI455X replaces the Instinct MI355X.
- Helios scales to 72 GPUs in a single pod, up from 8 GPUs for previous MI355X systems.
- Helios features a new fault-tolerant scale-up networking architecture based on UALink over Ethernet (UAEoL).
- A single MI455X contains 256 Work Group Processors (WGPs) across 8 Accelerator Complex Dies (XCDs).
- The MI455X has a max engine clock of 2.4GHz.
- The MI455X enables peak compute figures up to 40.26 PFLOP for OCP MXFP4.
- The MI455X has 12 stacks of HBM4 each on a 2048 bit bus for a total of 192 channels.
- Microsoft will use Helios for internal and external use.
- Helios will underpin managed compute for enterprise customers through Microsoft Foundry.
- Microsoft announced its commitment to Helios on Monday.
- Microsoft will add two new computing instances run on AMD's Venice CPUs.
- Helios is the first rival to Nvidia's Grace Blackwell and Vera Rubin systems.
- AMD and Cerebras expect their platform to deliver up to 5X higher tokens per second per watt.
- The MI455X GPU uses CDNA 5 architecture.
- The MI455X GPU is built using advanced packaging technologies.
- The MI455X GPU has four Accelerator Complex Dies (XCDs) stacked on a Fabric and Cache Die (FCD).
- The MI455X GPU uses TSMC’s CoWoS-L technology.
- The MI455X GPU uses TSMC 2N gate-all-around (GAA) process technology on the XCDs.
- AMD also introduced the Venice-X CPU for data centers.
- AMD Chair and CEO Dr. Lisa Su promoted Helios at the Advancing AI conference.
- Helios is built to train and run the most demanding frontier models at massive scale.
- Helios will be deployed by leading AI companies at gigawatt-scale.
- Helios uses Radeon Instinct MI455X and Epyc Venice processors.
- Helios is AMD's first rack-scale AI accelerator system.
- Microsoft will use Helios for frontier model inference.
- AMD will begin shipping Helios to customers later this year.
- AMD and Cerebras will partner to combine EPYC processors in Helios with Cerebras' Wafer-Scale Engine solutions.
- The MI455X GPU has 320 billion transistors and uses HBM4 memory.
- The Helios rack-scale architecture integrates 72 MI455X GPUs.
✨ This summary was generated by AI from the outlets' reporting listed below. It is not independently verified and may contain errors — check the original sources. How BrevFeed works →
The daily brief
One email each morning: the day's tech stories, clustered across outlets and summarized. No account needed.
One email a day. Unsubscribe in one click, any time.
Today's brief
Spend a few minutes, get the whole day. Every topic's top stories in one hands-free rundown — listen, watch, or read the transcript.
▶ Play today's brief
New every morning, and the back catalogue is archived by date.
How outlets covered it
AMD introduced the Instinct MI455X GPU, an upgrade to its AI accelerator lineup, and the accompanying Helios rack-scale solution. The MI455X is AMD's first GPU designed for rack-scale AI, featuring the new CDNA5 architecture with significant improvements in compute performance, memory bandwidth, and capacity, which matters for large-scale AI model training and inference.
AMD introduced Helios, a new AI rack-scale system designed for large-scale AI model training and deployment, along with the Venice-X CPU for data centers. This release positions AMD to directly compete with Nvidia in the high-performance AI hardware market, with major customers like Microsoft and Anthropic already planning deployments.
AMD introduced its Instinct MI455X GPU and the Helios rack-scale architecture, which integrates 72 MI455X GPUs into a single system. This release marks AMD's most competitive offering against Nvidia in the data center AI compute market, featuring advanced chiplet design and architectural changes in CDNA 5.
AMD and Cerebras Systems announced a partnership to develop an AI inference platform that integrates AMD's EPYC processors and Instinct GPUs with Cerebras' Wafer-Scale Engine (WSE) solutions. This collaboration aims to optimize AI inference workloads by disaggregating tasks, with AMD handling prompt processing and large context windows, and Cerebras WSE managing memory-bandwidth-intensive token generation.
AMD has launched its first rack AI system, Helios, designed to compete with Nvidia's offerings. Microsoft is now using the Helios system in its data centers, highlighting the increasing demand for AI compute capacity among major tech companies.
Microsoft and AMD will deploy AMD's Helios rack-scale AI accelerator for AI workloads on Azure. This partnership aims to enhance AI compute capabilities for Azure customers, directly competing with Nvidia.