The $100k already hiding in your cloud bill
Most SaaS founders think they need to find an extra $100k — a bigger round, a new pricing tier, one more growth hire. At any real scale, that money is usually already there. You’re just paying it to AWS.
Spend on the constraint, not the input
When a team gets a budget windfall, the reflex is to spend it on the thing that feels like progress: more leads, more sales, more marketing. But pouring money into the top of a leaky funnel just makes the leak more expensive. The highest-return move is almost always to fix the binding constraint — and for a lot of profitable, traffic-heavy SaaS, the binding constraint isn’t growth. It’s margin.
Margin is the unglamorous one nobody gets excited to spend on, and that’s exactly why it’s underpriced. A marketing spend is gone in 90 days. Recovered margin compounds every month, drops straight to the bottom line, and lifts your enterprise value — SaaS is valued on efficiency now, not just growth. So before you go find $100k, it’s worth checking whether you’re already paying it. It tends to hide in four places.
Where the wasted $100k hides
1. Egress — sending your own data out
The single most overlooked line. AWS charges roughly $0.05–0.09 per GB just to send your own data out — to your users, between regions, across availability zones. You already paid to store those bytes; egress charges you again every time you read them. For anyone serving media, files, backups, or shuffling data between services, this is frequently the biggest line on the bill — and it grows with your success, not your feature set. The same bytes on Cloudflare R2 or Backblaze B2 cost $0 to egress.
This one’s worth its own teardown — including how to find it on your own invoice in five minutes — so I wrote it up separately: The meter you can’t see.
2. NAT Gateway and cross-AZ traffic
A NAT Gateway bills an hourly charge plus ~$0.045 per GB processed. Route your private-subnet traffic through it at volume and a few thousand dollars a month quietly disappear into a service that does nothing interesting. Cross-AZ traffic is the same trick at a smaller unit price but large volume — you pay per gigabyte just to move data between availability zones inside the same region.
3. Idle and over-provisioned compute
Instances sized for a launch-day spike that never recurs. Staging and demo environments nobody turned off. Dev databases running 24/7. Reserved capacity that lapsed back to on-demand. Across most accounts this is commonly 20–30% of the compute bill — pure waste, recoverable with right-sizing and a schedule, no migration required.
4. Managed-service markup
Convenience has a price, and it’s often 3–5x the self-run cost. Managed Postgres (RDS/Aurora), observability (Datadog, CloudWatch), and friends are wonderful until the invoice lands. Some are worth every cent. Others are one flat-rate equivalent away from a much smaller number. The point isn’t “never use managed services” — it’s knowing which ones you’re overpaying for.
Find your $100k in an afternoon
You don’t need a consultant or a tool to start. The fix takes an afternoon, not a fundraise:
- Pull last month’s invoice (or a Cost Explorer / billing export).
- Sort by cost, descending.
- Look at Data Transfer and NAT Gateway first — the lines people forget exist, and almost always higher than anyone guesses.
- Then scan for compute that’s bigger or always-on without a reason, and managed services priced like a small team.
Add up what’s recoverable. For a lot of teams, the “$100k” they were about to go raise or earn is sitting right there — no new revenue required, and every dollar of it lands on the bottom line.
The honest caveat
None of this means rip out AWS. Some of those lines are earning their keep, and moving the wrong workload costs more than it saves once you count the engineering time. The goal isn’t to leave the cloud on principle — it’s to separate the spend that buys you something from the spend that’s just a meter you forgot was running.
If you’d rather not dig alone, send me your bill — I’ll break out the four lines above and tell you honestly how much is actually recoverable and whether it’s worth moving. And if your bill is already lean, I’ll tell you that too.