Speaking Topic: Cloud FinOps & Cost Optimization
Why most cloud cost audits miss the real bill, and what a forensic FinOps review actually finds.
I open this talk with a number that usually unsettles the room: in forensic audits of scaling SaaS platforms, I routinely find companies spending 22 to 28 percent of their entire cloud budget on proprietary telemetry platforms, often paying more for monitoring than for the production databases those tools are watching.
The AWS bill that nearly killed a startup before its Series B is the case study I return to most. It was not a traffic spike. It was an observability and infrastructure architecture that nobody had audited since the company was a tenth of its current size.
For every dollar deployed into specialized AI infrastructure, including high-density GPU clusters, custom ASICs, and high-bandwidth memory, enterprise software revenue has to generate a corresponding operational gross margin, or the infrastructure is simply burning runway with better branding.
Key Takeaways
- 01Observability tooling can silently become a bigger line item than production infrastructure.
- 02Cloud cost audits need to happen on a schedule tied to growth, not reactively.
- 03AI infrastructure spend needs the same margin discipline as any other capital allocation.
Frequently Asked
What is the most commonly missed cloud cost in a FinOps audit?
Proprietary observability and telemetry platforms, which can reach 22 to 28 percent of total cloud spend, often exceeding the cost of the production databases they monitor.
From the Q&A
“Isn't observability spend supposed to be high for a mature platform?”
Some increase is expected, but 22 to 28 percent of total cloud spend going to telemetry alone is a sign of unaudited defaults, not maturity.
“What's the fastest way to find hidden cloud cost?”
Reconstruct spend against actual usage, not list pricing, starting with observability and monitoring tools, which are the most consistently overlooked line item.