Worldwide AI spending is on track to reach 2.59 trillion dollars in 2026, with infrastructure, chips, servers and networking, now the single largest spending category at just over 46 percent of that total and global AI infrastructure spend already past 98 billion dollars. Most of that load still runs in public cloud, around 67 percent of enterprises, with hybrid and fully on-premises deployments each taking a meaningfully smaller share, a split with real consequences for cost predictability. This session breaks down what is actually driving infrastructure spend growth for teams running generative AI at scale, where GPU FinOps discipline is paying off today, and what the deployment mix looks like for teams trying to keep that bill predictable. What this session will cover: - What is actually driving infrastructure spend growth beyond the headline story that AI is expensive - The real trade-offs between public cloud, hybrid and on-premises deployment for generative AI workloads at scale - Where GPU FinOps practices are delivering measurable savings today, not just in theory - What infrastructure decisions are worth locking in now versus revisiting once pricing shifts again