The promise of cloud was compelling: pay only for what you use, scale up when you need it, scale down when you don't. For most businesses, the reality has been: pay for a lot of things you're not sure you need, scale up when traffic spikes, forget to scale down, and receive a bill that nobody can fully explain.

The problem isn't the cloud's pricing model — it's that the pricing model is unforgiving of inattention. On-premise hardware is a sunk cost you can ignore. Cloud resources bill continuously whether they're doing anything or not.

A Diagnostic Framework: Four Categories of Waste

Category 1: Idle resources

Resources running at minimal utilisation: EC2 instances below 10% CPU, RDS databases with near-zero connections, Elastic IPs not attached to running instances, EBS volumes attached to stopped instances. These are the easiest wins. A cloud audit typically finds 8–12% of spend in this category alone.

Category 2: Oversized resources

Instances provisioned for peak load and never rightsized after the peak passed. A database provisioned for a product launch that now runs at 15% of capacity. Spot the pattern by looking at 30-day utilisation averages, not the peak that justified the original sizing decision.

Category 3: Wrong pricing tier

On-demand is the right price for unpredictable workloads. For production databases, baseline application servers, and batch processing workloads that run on a schedule — reserved instances or savings plans are 40–60% cheaper with zero operational difference.

Category 4: Architectural inefficiency

This is the largest category and the hardest to fix quickly. Applications that can't scale horizontally. Storage architectures that don't use tiering. Synchronous designs that require always-on capacity for bursty workloads. These require architectural work, but the savings compound over the life of the application.

What a Systematic Reduction Looks Like

Our cloud cost engagements follow a consistent pattern: instrument first (tag everything, understand the bill at resource level), identify the top 10 cost items and their utilisation, address categories 1 and 2 immediately (typically 15–25% savings within 60 days), then develop a roadmap for categories 3 and 4 with clear ROI for each initiative.

The governance work — tagging policies, budget alerts, rightsizing review cycles — happens in parallel. Without it, savings erode within 12 months as new resources accumulate with the same patterns as the ones you just cleaned up.