AI FinOps: Why GPUs are expensive (and often wasted)

About this episode

The episode explores the hidden costs of AI infrastructure, particularly focusing on the expense and underutilization of GPUs.

Send us Fan Mail AI is everywhere, but the real cost often stays hidden. In this episode of DevOps Sauna, Pinja and Stefan dive into AI FinOps and explore why GPUs are expensive and why so much AI infrastructure ends up underutilized. They unpack hidden costs like idle hardware, power consumption, networking, and data movement, and explain why traditional Cloud FinOps doesn’t fully apply to AI workloads. They also discuss how Kubernetes and Dynamic Resource Allocation (DRA) can help improve G...

People in this episode

Hosts: Pinja, Stefan

Topics covered

• AI FinOps
• GPUs
• Cloud FinOps
• Kubernetes
• Dynamic Resource Allocation
• AI infrastructure
• cost management

Keywords

• AI
• FinOps
• GPUs
• infrastructure
• Cloud FinOps
• Kubernetes
• resource allocation

Mentioned in this episode

Organizations: Eficode

More episodes of DevOps Sauna from Eficode

• The most dangerous team is a silent one: Psychological safety in high-performing teams · June 5, 2026 · 27 min
• Vibe coding meets UX: Can AI build great user experiences? · May 22, 2026 · 24 min
• AI vs humans? Why people still come first in the age of AI · May 8, 2026 · 30 min
• Security isn’t a cost—It’s your biggest growth engine · April 24, 2026 · 26 min
• You built observability… But are you actually using it? · April 10, 2026 · 26 min
• AI agents are coming for your job… Or creating a better one? · April 9, 2026 · 28 min