AMD and Celestica have announced a new collaboration to build a rack‑scale artificial intelligence platform called Helios. The platform is designed to support large AI training and inference workloads in modern cloud and enterprise data centers.
The announcement reflects a growing trend in the data‑center industry where chip manufacturers and infrastructure providers work together to design integrated AI computing systems.
What the Helios AI Platform Is
Helios is a rack‑scale AI infrastructure architecture that integrates compute, networking, and system design into a unified platform optimized for artificial intelligence workloads.
The system is designed to support large GPU clusters required for training advanced AI models. In a rack‑scale architecture, an entire rack of servers works together as a single high‑performance computing system.
This approach improves performance, reduces latency between GPUs, and allows hyperscale cloud providers to scale AI infrastructure more efficiently.
Role of AMD and Celestica
In the partnership, AMD provides the core AI computing hardware while Celestica focuses on system integration and networking infrastructure.
Celestica is responsible for the engineering and manufacturing of high‑performance networking switches used inside the Helios rack architecture. These switches allow large numbers of GPUs to communicate with extremely high bandwidth.
The networking layer is a critical component for large AI clusters where thousands of GPUs must exchange data quickly during model training.
Hardware Architecture
The Helios platform integrates multiple AMD technologies to build a complete AI infrastructure stack.
Key components include:
- AMD Instinct MI450‑series GPUs for AI acceleration
- AMD EPYC Venice CPUs for compute processing
- AMD Pensando Vulcano networking interfaces
- ROCm open software ecosystem for AI workloads
These components allow the platform to run large‑scale machine learning models while maintaining high throughput and low communication latency.
Open Infrastructure Design
Helios follows open data‑center design standards inspired by industry initiatives such as the Open Compute Project. This approach allows data‑center operators to build flexible and scalable AI clusters without relying on fully proprietary systems.
The networking architecture also uses Ultra Accelerator Link over Ethernet (UALoE), a technology designed to provide high‑bandwidth connectivity between AI accelerators in large clusters.
Availability
AMD and Celestica expect the Helios platform to become available to customers in late 2026. The solution is primarily targeted at hyperscale cloud providers, research institutions, and enterprises deploying large AI infrastructure.
The launch highlights the increasing demand for specialized AI hardware as companies continue to develop larger machine‑learning models and generative AI systems.
Why It Matters
AI workloads are rapidly increasing the demand for high‑performance data‑center infrastructure. Traditional server architectures are often not optimized for large GPU clusters.
Rack‑scale systems like Helios address this challenge by integrating compute, networking, and infrastructure into a single architecture optimized for AI workloads.
For the AI infrastructure market, the AMD–Celestica partnership signals stronger competition in the race to build next‑generation AI data‑center platforms.
Sources
SiliconeUpdate.com is a technology news platform that publishes updates and informational content related to silicon technology, software, artificial intelligence, and emerging technologies.
All articles published on this platform are attributed to SiliconeUpdate.com instead of individual authors. Content is presented in a neutral, informational format without personal opinions.
—
Content Publishing
SiliconeUpdate.com publishes news and updates based on publicly available information, official announcements, and industry developments. The focus is on clarity, relevance, and timely reporting.
—
Editorial Control
All editorial decisions, updates, and content management are handled at the platform level. No individual human or AI identity is presented as the author of articles.
—
Contact
For editorial communication or general queries, contact:
Email: neemasharma@gmail.com
