So it turns out the best cost optimization tool isn't a fancy SaaS dashboard. It's a properly timed, credible threat.
We've been running a non-trivial footprint on Azure for about three years. The bill had that slow, steady creep—extra storage tiers here, a few more managed services there. The usual entropy. Our FinOps reports were solid, but we'd plateaued on technical savings.
Last renewal cycle, I built a PoC of our core application stack on GCP using Terraform. Nothing crazy, just the critical path:
```hcl
# Not our actual config, but you get the idea.
module "gcp_alternative" {
source = "./modules/ugly-port"
region = "europe-west3"
node_pool_count = 4 # 2 fewer than Azure, thanks to their bigger machine types
}
```
Ran a load test, crunched the TCO numbers (including the migraine cost of migration), and presented it to our account manager with a simple message: "We're not unhappy, but we're not married to the platform. This runs 15% cheaper over here. Help us close the gap."
Key moves:
* **Data is ammunition:** The PoC wasn't a bluff. We *could* move. The terraform code proved it.
* **Timing is everything:** Did this 90 days before contract renewal. Not a day earlier.
* **Target the right line items:** Focused on committed compute spend. They'll never budge on egress or marketplace fees.
Result? A 10% discount on committed use for the next year. Not a credit, an actual rate reduction.
The lesson? Cloud providers are landlords. They'll only fix the leaky roof if you show them a signed lease from the building next door. Your leverage is your exit plan.
Pager duty survivor.
NightOps
I've followed a nearly identical playbook with AWS, but the devil is in the detail of the TCO calculation. Your 15% cheaper figure is critical - you must bake in the operational transition costs or your leverage evaporates.
Where I'd add a caveat is on the longevity of the discount. A one-time 10% concession on a committed spend is good, but it often just resets the clock. The creep resumes next cycle. The more valuable move is to negotiate not just a discount, but a new, lower pricing tier or committed-use model that alters the baseline. That's what we pushed for and secured after presenting similar migration data.
Also, having the Terraform code is good, but the real credibility comes from the load test results. Being able to show identical or better p99 latency on the rival platform is what makes the threat non-refutable.
> "We're not unhappy, but we're not married to the platform."
That's the right tone. Never show desperation.
Your 15% figure is the key. Anything under 10% and they'll just call the bluff. Account teams have internal thresholds for what they can approve without escalation. You need to cross one.
Did you get the discount applied to your committed spend contract, or just as a one-time credit? The latter is a band-aid.
Data over opinions