A Post-FinOpsX Debrief from the Mavvrik Team
FinOpsX 2025 wasn’t just a conference—it was a confessional.
Between booth chats, hallway deep dives, and our now infamous personality badges (yes, you are a Spend Whisperer), we had over 300 conversations with FinOps, engineering, and finance leaders. And what we heard was clear:
People are overwhelmed. The financial model for infrastructure is shifting faster than the tools, teams, and processes can keep up. AI is exposing cracks that already existed—and creating a few new ones.
Here’s what we heard loud and clear:
- FinOps for AI ≠ AI for FinOps
The phrase “AI for FinOps” was everywhere. And sure, better anomaly detection and smarter tagging rules are helpful. But most teams we spoke to were more concerned about FinOps for AI:
“How do we govern the cost of AI infrastructure that’s totally unpredictable?”
This means tracking GPU usage, understanding model-level spend, aligning pricing with cost-to-serve. AI needs financial governance, not just clever prompts.
- AI Spend Is a Black Box—And That’s a Problem
Almost no one could answer confidently:
- How much are we spending on LLMs?
- What does this agent cost to run?
- How much GPU did that customer burn last month?
People hate reporting “% of total cloud spend” because it says nothing about value. The ask is shifting toward cost-to-serve, per feature, model, or outcome.
- Kubernetes Is Still a Visibility Nightmare
Yes, Kubernetes helps scale. No, most people can’t see what it’s costing them.
Many attendees shared frustrations about:
- Shared clusters with no attribution
- Internal platforms where “nobody owns the bill”
- Cost spikes they couldn’t trace
There’s still a huge gap in cost telemetry for K8s-based workloads—especially with AI in the mix.
- Hybrid Is Back. And It’s Complicated.
On-prem is no longer just legacy—it’s strategic. GPU scarcity and data gravity are pulling workloads into private cloud, colocation, and even HPC environments.
But cost visibility isn’t following. Most tools still assume a cloud-native world. People told us:
“We have workloads in 4 clouds and 2 data centers. We need one picture. Not 6 dashboards.”
- People Want a Hero. And a Hug.
The real surprise? The badges.
What started as a fun icebreaker—“Pick your FinOps persona!”—sparked more conversations than any pitch. Folks leaned in. Laughed. Shared war stories. Asked for photos. And most importantly: they opened up.
It wasn’t just about tools. It was about identity. Visibility. Control. The pressure of being the one person standing between AI ambition and budget reality.
Everyone is looking for a hero. Sometimes, they just need to know they’re not alone.
Final Takeaway:
FinOps is no longer just about cloud. It’s about control.
Control over AI. Over hybrid infra. Over financial truth.
And most of all—over the narrative.
We’re grateful to have had 300+ of those conversations. Let’s keep them going.
—
Want to know your FinOps alter ego?
Take the quiz
