AWS Instance Purchasing Options
On-Demand is the default and the most expensive way to run an instance. Everything below is a way of paying less by giving something up: flexibility, capacity guarantees, or the right not to be interrupted.
Savings Plans are usually the first thing to consider now — they are simpler than Reserved Instances and cover Lambda and Fargate as well as EC2. Reserved Instances remain relevant mainly for RDS and for capacity reservations.
Reserved Instances
Section titled “Reserved Instances”A Reserved Instance is a commitment to a level of usage over one or three years, in exchange for a lower rate. The discount is applied automatically to matching usage; there is no separate instance to launch.
- Available for EC2 and RDS (and several other services with their own reservation models).
- Shared across accounts in an organization with consolidated billing, by default.
- Payment options: all upfront, partial upfront, no upfront. More upfront means a bigger discount.
| Standard RI | Convertible RI | |
|---|---|---|
| Maximum discount off On-Demand | up to 72% | up to 66% |
| Terms | 1 or 3 years | 1 or 3 years |
| Change Availability Zone, instance size, networking type | Yes | Yes |
| Change instance family, OS, tenancy, payment option | No | Yes |
| Benefit from later price reductions | No | Yes |
| Sellable on the Reserved Instance Marketplace | Yes | No |
Attributes
Section titled “Attributes”A reservation is defined by instance type, platform (operating system), tenancy, and optionally an Availability Zone.
- A zonal RI reserves capacity in a specific Availability Zone. This is the only purchasing option that guarantees capacity will be there.
- A regional RI applies to any Availability Zone in the Region and carries no capacity guarantee, but does give instance size flexibility within the family for Linux instances with default tenancy.
You can switch between zonal and regional through the console or API.
Spot Instances
Section titled “Spot Instances”Spot Instances use spare EC2 capacity at a large discount — AWS advertises up to 90% off On-Demand — in exchange for the right to take the capacity back.
There is no bidding. This is the single most common misconception, and it dates from a pricing model AWS abandoned in 2017. The Spot price for each instance type in each Availability Zone is set by EC2 and adjusted gradually based on long-term supply and demand. You pay that price. You may optionally set a maximum price you are willing to pay, but doing so only limits when your instance runs; it does not participate in any auction.
Request types are one-time or persistent. A persistent request is automatically resubmitted after its instance is interrupted; a one-time request is not. Defined-duration Spot — sometimes called “Spot blocks” — no longer exists.
Interruption handling is the part that determines whether Spot works for a workload:
- A Spot Instance interruption notice gives two minutes’ warning before EC2 terminates, stops or hibernates the instance.
- An EC2 instance rebalance recommendation is emitted earlier, when an instance is at elevated risk, giving a chance to drain work or launch a replacement before the two-minute clock starts.
Design for interruption rather than trying to avoid it. Diversify across instance types and Availability Zones and use the capacity-optimized allocation strategy in an Auto Scaling group or EC2 Fleet, so that requests are placed in the pools with the most spare capacity. Watching a price chart is not a strategy.
Spot suits batch processing, data analysis, CI runners, rendering, and anything else that can be restarted. Spot usage is not covered by Savings Plans, and Savings Plans commitments are not consumed by Spot spend.
Dedicated Instances and Dedicated Hosts
Section titled “Dedicated Instances and Dedicated Hosts”Dedicated Instances run on hardware dedicated to a single AWS account. They can be purchased On-Demand, Reserved or Spot. There is an additional charge of $2 per hour in each Region where you have at least one Dedicated Instance running, and that fee is not discounted by Savings Plans. Dedicated Instances may share hardware with other, non-dedicated instances belonging to the same account.
Dedicated Hosts go further: you get a whole physical server, and you can see and control which socket and core an instance lands on. That visibility is the reason they exist — per-socket, per-core and per-VM software licences usually require it, and a Dedicated Host is what makes a BYOL arrangement defensible in an audit. Available On-Demand or as a reservation. A host runs one instance type (or, for supported families, a limited mix), and its capacity depends on the instance type.
Use AWS License Manager alongside Dedicated Hosts to track core and socket consumption against the licences you actually hold.
Choosing
Section titled “Choosing”- Steady, always-on baseline → Savings Plans, or Reserved Instances where Savings Plans do not apply.
- Interruptible or bursty work → Spot, with diversification and a rebalance-aware design.
- Capacity that must be there → zonal Reserved Instances or On-Demand Capacity Reservations.
- Licence-bound software → Dedicated Hosts.
- Unpredictable or short-lived → On-Demand, and revisit once there is a usage history to model against.
Most estates end up with a mix: a Savings Plan covering the baseline, Spot for the elastic layer, and On-Demand absorbing the rest.