TL;DR: Multi-account AWS is the default beyond "two engineers and a side project," and AWS Organizations + Consolidated Billing is the free structure that makes it pay: volume-tier pricing aggregates across accounts, and a Reserved Instance or Savings Plan bought in one account auto-shares to every account in the Org by default. Going from one big account to a 5-account Org costs $0 and typically unlocks 5–15% of the total bill — the highest-ROI cost move in AWS. The catch is discipline: keep the management account near-empty and never let accounts silently opt out of sharing.
The numbers
- Volume tiers aggregate: S3 drops price at 50 TB and 500 TB — two accounts at 40 TB each pay first-tier rates separately, but summed to 80 TB in one Org, everything above 50 TB gets the cheaper tier. Data-transfer-out tiers similarly.
- RI/SP sharing is the default — a single Compute Savings Plan in the management account can cover EC2 across every member account, sized against Org-wide usage, not per-account.
- Structural change costs $0 but unlocks 5–15% in volume-tier + shared-commitment savings.
- Field examples: a 60-person startup captured ~$3,200/mo from S3 tier aggregation + a suddenly-Org-wide SP the first month, then ~$5,000/mo more by centralizing SP purchases (~$98K/yr); a 200-account enterprise found 34 accounts opted out of sharing and recovered ~$28K/mo by re-enabling it.
Do this
- Create the Org in "All features" mode (not consolidated-billing-only — you want SCPs and account management too), even with just one workload account today.
- Keep the management account near-empty — it exists to pay the bill and own Org config; workloads in it are a security and recovery liability. Plan to drain an existing main account over 3–12 months.
- Centralize Savings Plans purchases in the management (or a dedicated "Savings") account, sized from Cost Explorer's SP recommendation with the linked-account filter set to all accounts — never buy fragmented per-account commitments.
- Audit RI/SP discount sharing periodically — Billing Preferences per account; treat any opt-out as a documented, annually-reviewed exception.
- Group accounts into OUs matching the business (prod, non-prod, sandbox, security, logging) so SCPs and backup plans attach cleanly.
Gotchas
- Volume savings are automatic but invisible — no "you saved $X from tier pricing" line; it shows only as a bill lower than the sum of list prices.
- Per-account sharing opt-out breaks the Org-wide commitment math — the leading cause of "we bought an SP but production still shows on-demand."
- The management account can never leave its own Org and some actions don't apply to it — plan accordingly.
- Cross-account workload migration is painful — S3 buckets can't be transferred (copy), RDS needs snapshot-share, KMS keys can't move at all; account closure is a 90-day process.
- Overly-restrictive SCPs can deny even root in member accounts — test on a non-critical OU first.
Skip this if
- You genuinely have one tiny workload and no second account on the horizon — but even then, set the Org up now; the right time was the first account, the second-best is today. The commitment that delivers the bulk of Org-level savings is Compute Savings Plans (and Reserved Instances for non-EC2); pair the structure with Budget Actions for enforcement and Cost Categories to align billing views to your OU tree.