Purchase quotas
AWS caps how many Reserved Instances you can buy.
Reserving 300 instances in one Region does not happen in one month. Here's EC2's monthly purchase quota, the On-Demand limit trap it sets, and how RDS and ElastiCache cap reservations differently.
First, the EC2 quota
AWS limits how many new Reserved Instances you can buy in a month, not just how many you can hold.
AWS's own EC2 limits page sets a default of 20 new regional Reserved Instances per Region per month, and 20 new zonal Reserved Instances per Availability Zone per month. Both quotas run in parallel, so a Region with three Availability Zones has a default of 80 new Reserved Instances a month.
Only a running instance counts toward the quota. AWS excludes pending, stopping, stopped, and hibernated states. The quota is adjustable through Service Quotas, but the request itself takes time, so ask before a large purchase, not during one.
Instances, not orders
One purchase can use up the whole month's quota by itself.
AWS counts the quota against instances, not against how many times you clicked buy. One Reserved Instance configuration bought with an instance count of 10 counts as 10 against the monthly quota, the same as making that purchase 10 separate times. A single large order can clear most of a Region's default 20 in one step.
The On-Demand limit trap
A regional Reserved Instance cannot raise what you can run. A zonal one can.
AWS states you cannot exceed your running On-Demand Instance limit by purchasing regional Reserved Instances. The default On-Demand Instance limit is 20. Run 20 On-Demand Instances today, buy 20 more as regional Reserved Instances, and the discount applies to the 20 you already run. Buy past that, and there is nothing left to launch against the extra Reserved Instances.
A regional Reserved Instance
cannot raise your running On-Demand Instance limit. Buy 20 regional Reserved Instances while already running 20 On-Demand Instances, and you cannot launch anything new to use them.
A zonal Reserved Instance
reserves capacity as well as a discount, so it can exceed your running On-Demand Instance limit. It is the one purchase type that raises what you can actually run.
AWS's own advice is to request the On-Demand limit increase before the Reserved Instance purchase, not after. A small account is the one most likely to hit this: it starts closest to the default On-Demand limit, and a single regional purchase can cross it.
Three services, three quota shapes
EC2 caps a monthly flow. RDS caps a standing stock. ElastiCache publishes neither.
Reserver scans reservations on all three services, and AWS caps each of them differently.
| Service | Quota shape | Default limit |
|---|---|---|
| EC2 | Monthly purchase flow | 20 regional + 20 per zonecounts instances, not orders; running instances only |
| RDS | Standing stock | 40 per Region20 in ap-south-1; how many you hold, not how many you buy this month |
| ElastiCache | Not published | no reserved-node quotathe 300-node Region quota applies only to non-reserved nodes |
Quota shape
- EC2Monthly purchase flow
- RDSStanding stock
- ElastiCacheNot published
Default limit
- EC220 regional + 20 per zonecounts instances, not orders; running instances only
- RDS40 per Region20 in ap-south-1; how many you hold, not how many you buy this month
- ElastiCacheno reserved-node quotathe 300-node Region quota applies only to non-reserved nodes
AWS's RDS quotas page names a "Reserved DB instances" quota directly: the maximum number of reserved DB instances allowed in the account, in the current Region, 40 by default and 20 in ap-south-1. It caps how many you hold at once, with no separate monthly purchase limit.
AWS's ElastiCache quotas page lists no reserved-node quota at all. Its "Nodes per Region" quota of 300 states plainly that it "applies to your non-reserved nodes within the given Region", so a reserved node sits outside that count too. AWS publishes no cap on ElastiCache reservations, in either direction.
Where this meets the renewal ladder
A clustered expiry month is also a clustered purchase month.
The renewal ladder guide gives one reason to spread expiry dates: a big renewal backlog slips past the people reviewing it. The EC2 purchase quota gives a second, independent reason.
Sixty EC2 reservations expiring in one Region the same March need sixty regional repurchases that same March. The default quota stops a same-day re-buy at 20. A ladder that already spreads expiry dates across the year spreads the purchases needed to replace them too, so no single month runs into the quota by accident.
How this scales with you
The default quota is the same size for every account. What it takes to hit it is not.
A handful of reservations never gets near 20 in a month. The quota that matters here is the On-Demand Instance limit. Check it before the first regional purchase, not after.
Dozens of reservations across a few Availability Zones can hit the zonal quota one zone at a time, even while the Region total looks fine. A purchase plan needs a per-zone count, not just a per-Region one.
Reserving hundreds of instances in one Region in one month runs straight into the default 20-plus-20 quota. Spreading the buy across months, or requesting a Service Quotas increase ahead of the purchase, both work. Waiting until the purchase fails does not.
Where Reserver fits
Reserver counts what you already hold, and what's still queued to renew.
Reserver scans your RDS, ElastiCache, and EC2 reservations alongside what's running, so a purchase plan starts from an exact reservation count, not a guess. It does not request a quota increase or place a purchase for you.
Check your own expiry shape with the free Expiry Check tool before a large renewal month arrives, and read the renewal ladder guide for the fix. See the queued purchases guide for the one case, an EC2 regional Reserved Instance or a Savings Plan, where a purchase can wait for a future date and still land against that future month's quota.
How many Reserved Instances can I buy in a month?
For EC2, AWS's default quota is 20 new regional Reserved Instances per Region per month, plus 20 new zonal Reserved Instances per Availability Zone per month. A Region with three Availability Zones has an 80-Reserved-Instance default, AWS's own worked example.
Does buying one Reserved Instance with an instance count of 10 use up 10 of my quota, or 1?
Ten. AWS states that the quota counts instances, not purchase orders. One Reserved Instance configuration bought with an instance count of 10 counts as 10 against the monthly quota, the same as 10 separate purchases of 1.
Can I buy more Reserved Instances than I have running instances?
Not with a regional Reserved Instance. AWS states you cannot exceed your running On-Demand Instance limit by purchasing regional Reserved Instances, so request the On-Demand limit increase before the purchase, not after. A zonal Reserved Instance is the exception: it carries its own capacity reservation, so it can exceed the running On-Demand Instance limit.
Do RDS and ElastiCache have the same monthly purchase quota as EC2?
No. AWS documents no monthly purchase quota for either service. RDS caps how many reserved DB instances you can hold at once: 40 per Region, 20 in ap-south-1. ElastiCache publishes no reserved-node quota at all, and its 300-node Region quota explicitly excludes reserved nodes.
Can the EC2 or RDS quota be raised?
Yes. Both are adjustable through Service Quotas. Request the increase before a large purchase, since the request itself takes time to process.
How does this connect to renewal laddering?
A clustered expiry month is also a clustered purchase month. Sixty EC2 reservations expiring in one Region the same March need sixty regional repurchases that same March, and the default quota stops a same-day re-buy at 20. Spreading expiry dates spreads the purchases needed to replace them.
Sources, fetched and confirmed on 2026-09-11: EC2 Reserved Instance limits; RDS quotas and constraints; and ElastiCache quotas and constraints.
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.