A system that can scale from 10 to 1,000 servers is described as both 'scalable' and 'elastic' — are these actually the same property, and does the distinction matter practically?

Scalability is the capability to handle increased load by adding resources — a system either can or can't grow to meet demand, regardless of how automatically or quickly that happens. Elasticity is specifically about how automatically and rapidly that scaling happens in both directions — scaling up under load and, critically, scaling back down when load decreases, without manual intervention. A system can be scalable without being elastic (it can grow to handle load, but requires manual provisioning and doesn't shrink back down automatically), and the distinction matters directly for cost, since a scalable-but-not-elastic system tends to accumulate over-provisioned capacity that never gets reclaimed.

The two terms describe genuinely different properties, and conflating them obscures a real, practically important distinction — particularly around cost, which is where the difference actually shows up most concretely.

Scalability is about capability — can the system grow to handle more load at all: a traditional on-premises data center can be scalable in this sense (you can buy and rack more servers to handle growth), but this scaling is manual, slow (procurement lead time), and — critically — essentially one-directional in practice, since nobody decommissions and sells off hardware just because load temporarily decreased.

Elasticity is about the automatic, bidirectional, rapid nature of that scaling: a genuinely elastic system scales up automatically in response to increased load (without a human manually provisioning new capacity) and, just as importantly, scales back down automatically when load decreases — this bidirectional, automatic behavior is specifically what cloud auto-scaling groups, serverless compute, and similar mechanisms provide, and it's the property that on-premises infrastructure structurally can't easily replicate.

The cost implication is the concrete, practical reason this distinction matters: a scalable-but-not-elastic system (say, cloud infrastructure that was manually scaled up for a traffic spike but never scaled back down afterward) accumulates idle, over-provisioned capacity that continues costing money indefinitely — the system was capable of scaling, but without the automatic bidirectional behavior, the actual cost efficiency benefit of "only pay for what you're using" never materializes.

A system can technically be scalable without being elastic, but true elasticity requires underlying scalability: elasticity is really "scalability plus automation plus bidirectionality" — you can't have an elastic system that isn't also scalable (since elasticity requires the capability to grow in the first place), but you can absolutely have a scalable system that isn't elastic (the capability exists, but the automatic, bidirectional mechanism to actually use it efficiently doesn't).

In practice, cloud-native architecture aims for elasticity specifically, not just scalability: the actual cost and operational value proposition of cloud infrastructure over traditional on-premises scaling is largely about elasticity — the ability to automatically match capacity to actual real-time demand in both directions — rather than just the theoretical capability to eventually grow, which traditional infrastructure could also technically achieve, just slowly and manually.

• Scalability is the capability to handle increased load by adding resources — it says nothing about how automatically or quickly that happens, or whether it happens in both directions.
• Elasticity is specifically the automatic, rapid, bidirectional nature of scaling — growing under load and shrinking back down when load decreases, without manual intervention.
• A scalable-but-not-elastic system accumulates idle, over-provisioned capacity that was manually scaled up but never scaled back down, directly costing money without corresponding benefit.
• Cloud-native architecture's real cost and operational advantage over traditional infrastructure is largely about elasticity specifically, not just the theoretical capability to eventually scale.

• AWS: What is elasticity in cloud computing?
• NIST: The NIST Definition of Cloud Computing