Key Takeaways
- AMD has partnered with Celestica to create Helios, a rack-scale AI infrastructure platform
- Celestica will lead R&D, engineering, and production of scale-up networking switches for the system
- The networking switches will link upcoming AMD Instinct MI450 Series GPUs for hyperscale AI deployments
- Helios leverages Open Compute Project Open-Rack-Wide standards and Ultra Accelerator Link over Ethernet technology
- Platform launch planned for late 2026; AMD shares increased approximately 1% in premarket trading Monday
Advanced Micro Devices has joined forces with Celestica (CLS) to develop Helios, a rack-scale platform engineered for demanding AI training and inference applications. The system addresses the infrastructure needs of cloud providers, enterprise data centers, and research facilities running large-scale AI operations.
Celestica will oversee the research, development, and production of specialized scale-up networking switches integral to the Helios architecture. The design follows the Open Compute Project Open-Rack-Wide specification, a standardized approach increasingly adopted by hyperscale operators.
Advanced Micro Devices, Inc., AMD
These networking switches incorporate silicon designed to facilitate high-bandwidth connections among AMD’s forthcoming Instinct MI450 Series GPUs. The system employs Ultra Accelerator Link over Ethernet technology to maintain rapid communication between GPU clusters during compute-intensive operations.
Rack-scale AI platforms treat the complete rack as a unified computing element rather than a collection of separate servers. This approach integrates GPUs, high-performance networking infrastructure, and liquid cooling systems into a cohesive platform optimized for training large language models at enterprise scale.
“Helios represents a new blueprint for AI infrastructure,” said Forrest Norrod, AMD’s executive vice president and general manager of Data Center Solutions. He said it enables customers to deploy AI with the performance, efficiency, and flexibility needed for next-generation workloads.
Steven Dorwart, senior vice president at Celestica, emphasized that large-scale AI deployment demands infrastructure solutions capable of rapid and reliable delivery. Celestica brings its established capabilities in data center architecture, engineering expertise, and supply chain management to the Helios collaboration.
AMD’s Market Position
AMD commands a market capitalization near $315 billion and has delivered a 92% return over the trailing twelve months, driven by surging demand for AI-focused infrastructure. Helios marks another step in AMD’s expansion within the data center GPU segment.
UBS maintains a Buy rating on AMD with a $310 price target, highlighting revenue expansion opportunities extending through 2027. The firm suggests AMD may secure a third major hyperscaler client for its data center products, with Microsoft identified as a probable contender.
Wolfe Research assigns an Outperform rating to AMD, emphasizing the company’s server traction and its AI accelerator development pipeline as primary catalysts.
Additional Strategic Initiatives
Along with Helios, AMD has finalized a multi-year licensing arrangement with Adeia Inc., securing rights to Adeia’s semiconductor intellectual property and resolving all pending legal disputes between the organizations.
Avalon GloboCare has gained admission to AMD’s AI Developer Program, providing the company with access to AMD’s development tools and AI resources.
AMD shares advanced roughly 1% in premarket trading Monday after the Helios announcement. Celestica stock rose approximately 3% during the same period.
Customer deployments of AMD Helios are planned for late 2026.
