Building Compute Foundations for the Physical Economy - with Drew Henry of ARM
AI in Infrastructure PodcastShow Notes
AI is moving into the physical economy, and enterprises are now confronting the infrastructure decisions required to support model‑driven systems in logistics, manufacturing, transportation, and other asset‑heavy environments. In this episode, Drew Henry, Executive Vice President for Physical AI at Arm Holdings, joins Emerj CEO Daniel Faggella to examine how leaders can redesign compute and hardware architectures for safety‑critical operations where power, reliability, and integration constraints are non‑negotiable. The discussion focuses on upgrading legacy automation, selecting the right compute for varied AI workloads, and using large‑scale simulation to validate changes before retooling physical systems. Emerj reaches over a million senior operators and AI decision‑makers each year — see how leading AI brands turn that attention into qualified pipeline at emerj.com/AD1
More from AI in Infrastructure Podcast
AI agents can place new demands on existing infrastructure, from database traffic to permissions for accessing and modifying information. In this episode, Alistair Speirs, General Manager of Global Platforms in Microsoft's Azure business, joins Daniel Faggella, Emerj CEO and Head of Research, to examine how organizations can adapt their systems while reducing administrative work for employees. He discusses processing expense receipts through familiar chat tools, governing agent access, and preserving model and deployment options as infrastructure costs and business requirements evolve.
Mobile networks were built for predictable human behavior, not the nonstop demands of agentic and physical AI. In this episode, Pallavi Mahajan, Chief Technology and AI Officer at Nokia, examines why AI workloads demand deterministic connectivity rather than traditional reliability benchmarks, and how networks are shifting from passive transport to active participants in AI decision-making. The conversation covers the shift from over-provisioning to AI-native network intelligence, the move toward distributed compute at the edge, and new capabilities like integrated sensing and real-time translation built directly into network infrastructure.
AI pilots that perform flawlessly in testing often stall the moment they reach production, and enterprise leaders are left wondering why.
Enterprises are past asking whether LLMs work; the harder question is how to run them at scale without becoming dependent on a single model, cloud, or chip vendor — and without running out of power or data-center capacity in the process.
The enterprise shift at the center of this episode is the growing gap between AI ambition and the underlying data and context infrastructure required to run systems reliably at scale.