Robotics and edge AI put new pressure on computing infrastructure
Physical AI is forcing the technology industry to rethink the entire computing stack. Robots, autonomous systems and intelligent devices need economical inference, secure data access and infrastructure that works beyond conventional clouds. Rafay Systems is addressing those demands through orchestration that lets providers offer open models without dedicating entire GPU systems to individual…
The rapid advancement of robotics and edge AI is placing increasing pressure on computing infrastructure. Companies like Rafay Systems are addressing these demands through orchestration that enables providers to offer open models without committing entire GPU systems to individual customers, potentially reducing enterprise AI costs while safeguarding sensitive data.
Rafay Systems' CEO, Haseeb Budhani, explains that their serverless infrastructure and secure GPU slices allow them to monetize every second of their infrastructure. For enterprises, this results in a more cost-effective solution with better price points. Max Kan, tokenomics technical lead at SemiAnalysis LLC, highlights the importance of tokens in physical AI, noting that agentic applications demand significantly more tokens than simple chat interactions due to the need for reprocessing previous context.
The inference market is also splitting into specialized workloads, with prefill and decode requiring different computational needs. Positron AI Inc. is targeting these requirements with systems optimized for tokens per dollar and tokens per watt, using air-cooled designs suitable for enterprise data centers. As physical AI moves into various environments, infrastructure providers must consider power, security, and system control.
Axiado Corp. tackles these challenges through silicon-based platform management that dynamically adjusts cooling, frequency, and voltage based on workload, offloading these tasks from GPUs and CPUs. Compact models that can run locally are becoming increasingly important for physical AI applications, such as those developed by Liquid AI Inc. These customizable foundation models are designed for devices with limited power, memory, and connectivity, allowing specialized intelligence to reduce inference costs, enhance privacy, and deliver faster responses without relying solely on centralized data centers.
Written by urgent.news from SiliconANGLE's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
