AMD Expands AI Software Strategy With ROSS Orchestration Platform

AMD is rolling out ROSS as part of its broader agentic AI software initiative, designed to handle complex workflows that involve models, tools, databases, and multiple execution steps. ROSS complements other AMD software projects like PACE and ROCm.AI, each addressing different layers of the company's AI strategy across its CPU, accelerator, and graphics hardware. The company emphasizes that agentic workloads require system-level orchestration and CPU compute alongside accelerator capabilities.

AMD's ROSS platform represents a significant evolution in how the company approaches artificial intelligence infrastructure. Rather than treating AI execution as a simple model-inference problem, AMD recognizes that modern AI agents require orchestration across multiple layers—managing interactions between language models, external tools, databases, and decision-making logic. This reflects a fundamental shift in AI application architecture, where the software that coordinates different components has become equally important as the underlying neural networks.

The company's broader AI software ecosystem demonstrates this layered approach. PACE handles orchestration of complex workflows, ROCm.AI provides developer tools and optimization for AMD hardware, and ROCm itself manages the low-level GPU operations. This multi-component strategy acknowledges that agentic systems place different computational demands on hardware, with CPUs managing workflow coordination and data processing alongside accelerators that handle model inference.