Rethinking AI Infrastructure for the Age of Reasoning

For years, AI infrastructure has been built around a simple assumption: bigger models require bigger hardware. But as AI enters the era of reasoning, agentic systems, and enterprise deployment, scaling compute alone is no longer enough. The next generation of AI will require infrastructure designed for efficiency, adaptability, and sustainable performance.

In this keynote, Jeehoon Kang explores how the demands of modern inference are reshaping AI hardware architecture. From balancing performance with power efficiency to enabling real-world enterprise deployment, he'll examine why the future belongs to purpose-built AI accelerators designed for the workloads that matter most.

Drawing from FuriosaAI's experience building next-generation inference processors, Jeehoon will share his perspective on where AI infrastructure is headed and the future that FuriosaAI is building.