Cost Optimization at Scale: A Knowledge-Graph Approach
Why flat-inventory cost tools plateau — and how relational analysis unlocks deeper savings.
Cloud cost optimization is often treated as a billing exercise — tagging resources, purchasing reserved capacity, and deleting obviously unused instances. But the scale of cloud waste suggests these practices are not enough. Flexera's 2025 State of the Cloud Report found that organizations waste 27 to 32 percent of their cloud budgets, with public cloud spending projected to reach $1.03 trillion in 2026. Harness's "FinOps in Focus" report estimated $44.5 billion in infrastructure cloud waste for 2025 from underutilized resources alone — and cloud budgets are already exceeding limits by 17 percent on average.
Why traditional cost tools plateau
Traditional cost optimization tools analyze billing data and resource utilization metrics. They identify an EC2 instance running at 5 percent CPU and recommend a downsize. They flag an unattached EBS volume and suggest deletion. According to DataStackHub's analysis, idle or stopped resources account for 10 to 15 percent of monthly cloud invoices, over-provisioned compute adds 10 to 12 percent, and orphaned storage artifacts add another 3 to 6 percent.
These are valuable findings — but they represent the shallow end of cost optimization. The deeper savings lie in understanding relationships: the load balancer serving traffic to a single instance that could be replaced by a simpler architecture, the three teams running nearly identical staging environments that could share one, the data pipeline copying terabytes between regions because of a historical architectural decision that no longer applies. Harness found that 52 percent of engineering leaders say the disconnect between FinOps and development teams is a primary cause of waste, and 55 percent of developers say purchasing commitments are based on guesswork.
Relational cost intelligence
A knowledge-graph approach to cost optimization models the full dependency chain: from compute instances to the services they run, the teams that own them, the data they process, and the business functions they support. This relational context enables a class of optimization that flat-inventory tools cannot achieve.
Consider a common scenario: an organization has 15 microservices, each with its own RDS instance, running in both staging and production across two regions. A flat-inventory tool sees 60 database instances and evaluates each individually. A knowledge graph sees that 8 of these databases serve read-only query patterns that could be consolidated onto a shared read replica, that 4 staging databases are oversized replicas of production, and that cross-region replication for 3 services is unnecessary because those services have no users in the second region. Only 43 percent of developers have access to real-time data on idle cloud resources — the graph provides that visibility by default.
How Hermeez finds deeper savings
Hermeez's Cost Agent continuously traverses the infrastructure graph to identify optimization opportunities across six dimensions: idle resources, right-sizing, architectural redundancy, reserved capacity gaps, data transfer inefficiency, and storage lifecycle misalignment. Each recommendation includes a projected savings figure, an impact assessment based on the dependency graph, and a confidence score reflecting the agent's certainty in the finding.
The dependency-aware impact assessment is critical. Recommending that an engineer delete an "unused" resource that is actually a disaster recovery failover target is worse than not making the recommendation at all. DataStackHub's research shows that rightsizing and auto-scaling alone cut compute waste by 25 to 35 percent, automated shutdown schedules reduce waste by 20 to 25 percent in the first 90 days, and commitment management lowers run-rate by 20 to 37 percent. The graph ensures these recommendations are safe to act on.
From cost cutting to cost intelligence
Gartner found that 54 percent of infrastructure and operations leaders cite cost optimization as the top goal for adopting AI. The FinOps Foundation's State of FinOps 2025 report showed that 50 percent of practitioners rank workload optimization and waste reduction as their number one priority, with investment in FinOps tools and automation up 20 percent year-over-year.
The most valuable shift that graph-based cost optimization enables is from reactive cost cutting to proactive cost intelligence. Instead of reviewing last month's bill and looking for waste, teams can evaluate the cost implications of architectural decisions before they are implemented. Gartner notes that organizations with no optimization plan overspend by up to 70 percent. The same knowledge graph that powers cost analysis simultaneously serves security and compliance — it does not just find waste, it reveals the structural patterns that create waste, enabling teams to address root causes rather than symptoms.