A Long-Standing Trust in AWS
I’ve been an AWS customer for over 15 years and have always appreciated the platform’s capabilities and billing transparency. However, I recently encountered a situation involving a Reserved Instance (RI) that remained visible and appeared fully valid in the EC2 console for over two years, while silently failing to apply any discount the entire time.
I attempted to resolve this privately through AWS Support, including an escalation to a Support Manager, but ultimately AWS declined to address the historical overcharge. I am sharing this so other AWS users are aware of how this can happen and how subtle the indicators can be.
A Reserved Instance That Appeared Fully Valid
In April 2023, I purchased a No Upfront Reserved Instance in the eu-west-1 (Ireland) region:
- Instance type and number: t3a.nano x 23
- Operating System: Linux
- Tenancy: Shared
- Payment option: No Upfront
- Expiration date shown: April 12, 2026
Because it was a No Upfront RI, there was no payment due at purchase, and I had no reason to expect any billing issues. The RI appeared normally in the EC2 console, and the Start and End Dates were clearly visible. Everything suggested that it was active and functioning.
During this entire period, I consistently ran 23 t3a.nano instances — the type this reservation was meant to cover.
What I Eventually Discovered
In late 2025, while checking something unrelated, I noticed a small “payment-failed” label associated with the RI. Importantly, this indicator is shown in a column positioned at the far right edge of the table — not visible on a typical laptop or desktop screen without horizontal scrolling.
There were:
- No warning banners
- No console alerts
- No emails
- No billing notifications
- No indication in Cost Explorer that something was wrong
The reservation continued to show a normal expiration date in 2026, just like any valid RI. The failure state was effectively hidden unless one happened to scroll far to the right, and for a No Upfront RI, a “payment failure” is not something a customer would logically expect.
The Financial Impact
Once I realized what had happened, I calculated the cost difference:
- On-demand price for t3a.nano: $0.0051 per hour
- RI price (No Upfront): $0.0022 per hour
- Difference: $0.0029 per hour
Over 23 instances running continuously for more than 31 months, this added up to roughly:
$1512 in unintended overcharges.
For a small business, this is not a trivial amount.
My Attempt to Resolve This Privately
I opened a support case in September 2025 and provided:
- A screen recording showing the RI exactly as displayed
- The list of instances that should have been covered
- A detailed explanation of the UI issue and resulting overcharge
After slow progress and several weeks of silence, the case was escalated to a Support Manager. I proposed a resolution that was both fair to me and easy for AWS to implement:
- Terminate the failed RI
- Apply a credit equal to the historical overcharge
- Use that credit directly to purchase a new 3-year All Upfront RI on my behalf
This would have avoided any need for retroactive billing adjustments and resulted in a new long-term RI commitment from me.
AWS Support Management declined the proposal, citing internal policies and stating they could only consider adjustments for the most recent three invoices.
Why I’m Sharing This
This situation highlighted two important issues that I believe AWS customers should be aware of:
- A failed No Upfront RI can remain visible with a normal expiration date for years, giving no indication that it’s not applying.
- The only sign of a failure may be hidden in an off-screen column with no notifications or alerts of any kind.
This combination makes it extremely easy for a customer to believe their RI is active when it is not — especially for No Upfront reservations, where a payment failure is counterintuitive and unexpected.
Closing Thoughts
I continue to use AWS and appreciate the platform, but this experience showed that even long-standing features like Reserved Instances can fail silently, and the console UI may not clearly reflect the true billing state.
I hope AWS takes this feedback to improve the visibility and notification mechanisms for RI failures so other customers do not unknowingly overpay for extended periods. If AWS revisits this issue or provides a different resolution in the future, I will update this post.