AWS Cost Optimization Hub

TL;DR: For years, AWS cost recommendations lived in six consoles — Compute Optimizer for EC2 rightsizing, Trusted Advisor for idle resources, Cost Explorer for RI/SP advice. Cost Optimization Hub (Nov 2023) is AWS's first real attempt to put them all on one sortable screen with dollar amounts, scoped across your whole Organization. It's an aggregator, not a generator: it surfaces the underlying services' recommendations, with all their caveats intact, and it prioritizes the work but doesn't do it — you still need an external process to turn findings into tickets.

The numbers

• Free — you pay only for the underlying services (Compute Optimizer is free; full Trusted Advisor cost checks need Business+ Support, ~$100/mo min).
• Each recommendation carries estimated monthly savings, effort (low/med/high), risk, resource tags, and account ID — the killer view is sort by savings descending for a prioritized to-do list.
• Setup: opt in (management account for Org scope), enable Compute Optimizer Org-wide, Cost Explorer with ~14 days history; wait 24–48 hours.
• Field examples: a 200-account Org built a Lambda to push Hub findings to Jira and closed ~$80K/mo on a rolling basis (mostly SP coverage gaps + gp2→gp3); a 3-account startup found $1,400/mo, realized $1,100/mo ($13K/yr) in ~10 hours, on Basic Support.

Do this

1. Enable it (30 min) from the management account at Organization scope, verify Compute Optimizer is enrolled Org-wide, then wait 24–48h.
2. Sort by estimated savings descending, filter to Effort: Low — usually idle-resource deletions, Compute Savings Plan buys, and gp2→gp3 conversions you can act on this week.
3. Group by account and tag to route each recommendation to the team that owns the resource.
4. Run a 30-min weekly triage — mark in-flight items, open tickets for the rest, note disagreements (often burst workloads Compute Optimizer misreads); a healthy state is "top 5 are in-progress or intentionally deferred."
5. Deep-dive in the source service before acting on a specific recommendation — the Hub is the dashboard, Compute Optimizer et al. are the workshop.

Gotchas

• It aggregates, it doesn't sanitize — an EC2 rightsizing rec inherits Compute Optimizer's blind spots (e.g. it assumes CPU-bound without the CloudWatch Agent reporting memory, risking a dangerous downsize on memory-bound workloads).
• No built-in workflow — it shows, it doesn't ticket; you need a Lambda or manual process to route findings, and refreshes lag 24–48h (a fix from yesterday may still show).
• Trusted Advisor coverage is partial on Basic Support — Compute Optimizer recs come through regardless, but idle-resource cleanup checks are gated behind Business+.
• Estimates are optimistic — SP savings assume current spend continues forever.
• Each member account needs the source services enabled — Org scope alone shows "no data" for accounts where Compute Optimizer/Trusted Advisor are off.

Skip this if

• You only run one service and already watch its native optimizer — the aggregation buys little at tiny scale (though it's still free to enable).
• You want the detail behind a recommendation, not the prioritized list — go straight to Compute Optimizer and Trusted Advisor — Cost Optimization Checks; purchase the Savings Plans it recommends in AWS Cost Explorer, and route findings by team with Cost Allocation Tags.