Instance generations
A Reserved Instance fixes a family. AWS keeps shipping a cheaper one.
AWS retires instance families years before most Reserved Instance terms end. Whether your reservation can follow the newer family depends entirely on which service and RI type you bought.
First, the bet
A 1- or 3-year term locks in one family. AWS does not wait for it.
A Reserved Instance discounts one specific instance family for the length of its term. AWS keeps releasing newer, cheaper families on its own schedule. That schedule ignores your term's end date.
AWS lists its own previous-generation instances directly. A family on that list still runs and still bills, but it is no longer the family AWS wants you buying. A reservation bought against one of those families locks in a rate that a newer family would have beaten for the rest of the term.
Which families are already superseded
Fourteen EC2 families, six RDS, six ElastiCache.
This is the same roster Reserver's own recommender checks before it proposes a term, not a separate marketing list.
| Service | Superseded family | Current-generation replacement |
|---|---|---|
| EC2 | m4 | m7i, m8g |
| EC2 | m5 | m7i, m8g |
| EC2 | m6g | m8g |
| EC2 | m7g | m8g |
| EC2 | c4 | c7i, c8g |
| EC2 | c5 | c7i, c8g |
| EC2 | c6g | c8g |
| EC2 | c7g | c8g |
| EC2 | r4 | r7i, r8g |
| EC2 | r5 | r7i, r8g |
| EC2 | r6g | r8g |
| EC2 | r7g | r8g |
| EC2 | t2 | t4g |
| EC2 | t3 | t4g |
| RDS | m5 | m6i, m7g |
| RDS | m6g | m7g |
| RDS | r5 | r6i, r7g |
| RDS | r6g | r7g |
| RDS | t2 | t4g |
| RDS | t3 | t4g |
| ElastiCache | m5 | m7g |
| ElastiCache | m6g | m7g |
| ElastiCache | r5 | r7g |
| ElastiCache | r6g | r7g |
| ElastiCache | t2 | t4g |
| ElastiCache | t3 | t4g |
A family missing from this list is either already current-generation or not yet flagged. Reserver never guesses. An unknown or current family gets no upgrade suggestion at all. A false "this is old" costs more than a missed one.
The escape hatch depends on the service and the RI type
One RI type can follow the hardware. The rest cannot.
This is the asymmetry that decides how expensive an old-generation mistake gets.
| Reservation type | Can the reservation follow a newer family? |
|---|---|
| EC2 Convertible RI | ✓exchange for a different family, size, OS, or tenancy, at no fee, if the new RI's value is equal or greater |
| EC2 Standard RI | ~modification only, and modification cannot change family; the RI Marketplace is the only other way out |
| RDS reserved DB instance | ✗no modification, no exchange, no resale; the reservation rides out its term on the family you bought |
| ElastiCache reserved node | ✗the same as RDS: nothing to modify, nothing to exchange, nothing to resell |
Can the reservation follow a newer family?
- ✓EC2 Convertible RIexchange for a different family, size, OS, or tenancy, at no fee, if the new RI's value is equal or greater
- ~EC2 Standard RImodification only, and modification cannot change family; the RI Marketplace is the only other way out
- ✗RDS reserved DB instanceno modification, no exchange, no resale; the reservation rides out its term on the family you bought
- ✗ElastiCache reserved nodethe same as RDS: nothing to modify, nothing to exchange, nothing to resell
AWS's own exchange documentation is clear that only a Convertible RI qualifies. Modifying a Standard RI can change Availability Zone. On EC2 with size flexibility, it can also change size within the same family. It can never change the family itself.RDS reserved DB instances and ElastiCache reserved nodes allow no modification of family, no exchange, and no resale. A reservation bought against the wrong family there runs to term end regardless.
What being stuck actually costs
The rate you locked, held against the rate you missed.
Move a workload to the newer family and you still have to run something. Until the new family is itself reserved, that replacement runs on-demand while the old reservation keeps billing at its own rate, doing nothing useful.
A Standard RI or an un-exchanged Convertible RI pays this cost the same way a lapsed renewal does, at the illustrative rate the renewal playbook uses. An RDS or ElastiCache reservation pays it for as long as the term has left, since neither offers an exchange to end it early.
What to do instead
Flag the family. Do not reserve it again.
Skip the reservation on the old family, and reserve the family you are actually moving to instead.
Mark a family as one your team plans to leave, and Reserver stops recommending a new 1-year or 3-year term on it. The recommendation for the replacement family still appears, priced and ready, once that family is actually running. This is the live loop, today: a flag suppresses a bad recommendation before you buy it, not after.
How this scales with you
One team can hold the roster in their head. Thousands of reservations cannot.
A handful of reservations and one migration plan. Check the family roster by hand, once a year, and you will likely catch it.
Dozens of reservations spread across a few accounts. The upgrade flag becomes the shared memory the team checks instead of everyone's own notes.
Thousands of reservations across an AWS Organization. Nobody can hold that roster in their head, and a spreadsheet goes stale within a quarter.
Where Reserver fits
Reserver flags the family, today. It does not buy the replacement for you.
Reserver stamps every recommendation with its old-generation status and the families that replace it, and suppresses a recommendation on any family you flag. Buying the replacement reservation is still your call.
Old-generation detection and upgrade flagslive · M7
See features for how the flag fits the rest of the coverage map, and the renewal playbook for what else keeps a reservation from quietly costing more than it should.
Does a Reserved Instance track AWS's newer instance families automatically?
No. A Reserved Instance fixes the family you bought at purchase time, for the whole term. AWS keeps releasing newer, often cheaper families, but the commitment does not move on its own.
Can I exchange an old-generation Reserved Instance for a new one?
Only an EC2 Convertible RI can. AWS lets you exchange it for a different family, size, operating system, or tenancy, at no fee. The new RI's value must be equal to or greater than the one you exchange. An EC2 Standard RI can only be modified, never exchanged into a new family. RDS and ElastiCache reservations offer neither option.
What can I do with a Standard RI, RDS reservation, or ElastiCache reservation stuck on an old family?
For EC2 Standard, sell the remaining term on the RI Marketplace. RDS and ElastiCache offer no exit at all. The reservation runs to term end on the family you bought, whether or not you keep using it.
How does Reserver help with old-generation families?
Reserver marks a recommendation as old-generation and names the current-generation families that replace it. Flag a family as one you plan to leave, and Reserver stops proposing a new term on it.
Sources, fetched and confirmed on 2026-09-09: AWS's own list of previous-generation instances; exchanging a Convertible Reserved Instance; modifying a Reserved Instance; working with RDS reserved DB instances; and ElastiCache reserved nodes.
Stop overpaying for a steady fleet.
Reserver is in early access. Join the list and we'll onboard you as capacity opens up. Your first recommendation lands within minutes of connecting.