By InfraZen Engineering · 2026-08-16 · 9 min read

The AWS native cost tools,
ranked by what they actually catch.

Before you buy a cost platform or hire anyone (including us), you should know that AWS ships eight cost tools that range from free to nearly free — and that they are not interchangeable. Each one sees a specific slice of your waste and is structurally blind to the rest. Vendors rarely explain this, because "the native tools cover more than you think" is a bad pitch. It happens to be true, so here it is, ranked by what each tool catches in practice.

Key takeaways

  • The CUR is the only source of truth; everything else is a lossy view of it. Enable it on day one — it has no memory of the past.
  • Anomaly Detection catches spikes, never drift — the slow 8%/month climb that actually ruins budgets sails straight past it.
  • Compute Optimizer knows utilization, not intent: it will recommend shrinking the headroom you deliberately keep for peak days.
  • No native tool remediates anything. The gap between "flagged" and "fixed" is where cloud waste actually lives.

The ranking, in one table

AWS native cost tools ranked by practical catch-rate, August 2026. Links go to AWS's own documentation.
Rank Tool What it actually catches What it structurally misses
1 Cost & Usage Report Everything, at line-item and resource-ID grain — the ledger every other tool summarizes Nothing, but it answers only the questions you engineer against it
2 Cost Anomaly Detection Spikes against learned baselines: the forgotten GPU box, the runaway Lambda loop Slow drift, anything that was already expensive on day one of its learning window
3 AWS Budgets Threshold breaches, forecast overruns; the only native tool with actions (it can cut IAM permissions) Anything under its thresholds; tells you that, not why
4 Cost Optimization Hub Aggregates idle, rightsizing and commitment recommendations org-wide, deduplicated and dollar-quantified Only sees what its feeder tools see; no architecture-level waste
5 Compute Optimizer Over-provisioned EC2/ASG/EBS/Lambda against observed utilization, incl. Graviton candidates Business context; memory (without agent metrics); your month-end peak
6 Cost Explorer Trends and what-changed questions a filter can answer; RI/SP purchase recommendations Resource-level truth; anything needing more than 13 months of history
7 S3 Storage Lens Bucket-level waste: incomplete multipart uploads, stale non-current versions, tiering candidates Everything outside S3; advanced metrics are the paid tier
8 Trusted Advisor Classic idle-resource checks: unattached EBS, idle load balancers, low-util instances Full checks need Business/Enterprise support; shallow next to the tools above

Why the ledger outranks every dashboard

The Cost & Usage Report wins because every other tool is a pre-baked summary of it. The question that actually finds money is never "what did EC2 cost?" — it's "which untagged resources in the staging account cost more than $500 last month, and whose IAM principal created them?" That's a five-line Athena query against the CUR and unanswerable anywhere else. The catch everyone learns too late: the report only accumulates from the day you enable it. Enabling it costs pennies in S3 storage; not having enabled it costs you the history every future analysis wants. It's the same reason we run this site's own analytics as Athena over raw logs — summaries expire, ledgers don't.

Where the free tools quietly end

Line the eight up and their shared blind spots draw a neat border around what "native" can do. Nothing remediates: every tool produces a finding; a human (or automation you build) produces a fix, and unactioned findings age into wallpaper — the Flexera data showing waste rising in 2026 is largely a monument to flagged-but-unfixed recommendations. Nothing sees architecture: no native tool will tell you your NAT gateway design is taxing every S3 call, that your chatty microservices are billing you cross-AZ, or that the whole batch tier belongs on Spot — the levers that made up most of the savings in our fintech cost case study. Nothing owns the process: tags decay, budgets go stale and anomaly monitors drift out of scope unless someone runs cost as a monthly practice — which is the actual subject of FinOps, not any tool.

The honest hierarchy: native tools for visibility and alarms, engineering for architecture-level fixes, practice for keeping both alive. If someone is selling you a platform, ask which of those three it replaces; the answer is usually "the first, more prettily." Our full lever ordering — commitments, rightsizing, traffic, storage, architecture — lives in AWS cost optimization, ranked by what pays, and the cloud waste calculator will give you a first estimate of what's recoverable without creating an account anywhere.

The enablement order for a new team

Day one: activate cost-allocation tags, set two or three Budgets with alerts wired to a channel humans read, and turn on Anomaly Detection with per-service monitors (coarse monitors average real spikes into silence). Same week: enable the CUR to S3 — you don't need Athena queries yet, you need the history to start existing. First month: read Compute Optimizer and the Cost Optimization Hub with your load calendar open, and act on the top five items only. That 90-minute setup catches the embarrassing failure modes for roughly $0, which is exactly why we make clients do it before any paid engagement starts — consultants who skip the free layer are billing you for alarm installation, not engineering.


Related: AWS cost optimization guide · Cloud billing & FinOps consulting · DevOps & cloud cost statistics 2026 · RI & Savings Plan calculator · Cloud bill audit checklist

IZ
InfraZen Engineering
DevOps, SRE & Cloud Consulting · Bangalore, India

Written by the InfraZen engineering team, consultants who run production DevOps, SRE and cloud platforms for fintech, SaaS and enterprise clients across the US, UK, Middle East and APAC. Posts are reviewed for technical accuracy by Bangalore-based SREs before publication.

About InfraZen → Talk to an engineer → LinkedIn →

Frequently asked questions

What is the difference between Cost Explorer and the Cost and Usage Report?

Cost Explorer is the dashboard: pre-aggregated views, 13 months of history, good for questions a filter can answer. The Cost and Usage Report (now delivered via Data Exports) is the raw ledger: every line item, every SKU, resource IDs, amortization detail, delivered to S3 for Athena or your warehouse. Cost Explorer answers 'what went up?'; the CUR answers 'which exact resource, tagged by whom, at what effective rate?'

Why didn't AWS Cost Anomaly Detection catch my bill spike?

Three usual reasons. First, it learns from your history, so a cost that grew steadily rather than spiking never looks anomalous. Second, monitors scoped too broadly average the spike away: a $2K jump inside a $200K service line is noise to a coarse monitor. Third, it detects after spend lands in the data, so a weekend GPU mistake can run for many hours before the alert. It's a smoke detector, not a sprinkler.

Is Compute Optimizer free, and can you trust its recommendations?

The default version is free, with a paid tier for enhanced infrastructure metrics that looks back further. Trust its direction, verify its context: it sees CloudWatch utilization, not your business calendar. It will happily recommend downsizing the instance you keep headroom on for month-end peaks, and by default it does not see memory unless you ship the agent metrics. Treat it as a shortlist generator, with a human holding the load context.

Which AWS cost tool should a small team enable first?

In order: activate cost allocation tags and set two or three AWS Budgets with alerts the same afternoon, turn on Cost Anomaly Detection with per-service monitors, then set up the Cost and Usage Report to S3 even if nobody queries it yet, because it only accumulates history from the day you enable it. Cost Explorer needs no setup. Everything else can wait until you have a real question.