Elastic Load Balancer Cost: ALB, NLB, GWLB Pricing Decoded

Elastic Load Balancer is one of the most commonly under-modelled line items in AWS. The hourly fee for an Application Load Balancer or Network Load Balancer is small. The Load Balancer Capacity Units (LCUs) — the variable component that scales with traffic, connections, and rule evaluations — are where the real cost lives. Customers who plan their load-balancer architecture against the hourly fee alone routinely underestimate ELB spend by 60–80%.

This guide breaks down ALB, NLB, GWLB, and Classic Load Balancer pricing in detail, the LCU dimensions that drive most of the bill, the optimization plays we apply during AWS audits, and how to bring ELB into your Enterprise Discount Program (EDP). We have reviewed $2.4B+ in AWS spend across 500+ engagements and consistently find 20–40% optimization opportunities in ELB line items.

The four load balancer types and their pricing

The base hourly fee is approximately $16.40 per month per load balancer. The variable rates look small. They are not — once you understand what an LCU actually measures.

Understanding LCUs — what actually drives the bill

An LCU is whichever maximum of four dimensions you hit in a given hour. AWS bills you for the worst of:

1. New connections per second. 1 LCU = 25 new connections/sec for ALB; 800 for NLB on TCP, 80 for NLB on TLS.
2. Active connections per minute. 1 LCU = 3,000 active connections for ALB; 100,000 for NLB on TCP.
3. Processed bytes. 1 LCU = 1 GB/hour for ALB; 1 GB/hour for NLB on TCP.
4. Rule evaluations per second. 1 LCU = 1,000 rule evaluations/sec for ALB (NLB has no rule dimension).

The wrinkle: the dimension that drives your LCU usage depends entirely on your traffic pattern. A high-volume static-asset workload usually hits the processed-bytes dimension first. A chatty microservice workload with short-lived connections hits new-connections. A workload with deep listener rule chains hits rule evaluations. Customers building cost models against the wrong dimension routinely under- or over-provision.

Where the bill concentrates by workload

Web application behind ALB

For a typical Software-as-a-Service web application, the processed-bytes dimension dominates. A workload pushing 5 TB per month through an ALB uses about 6.8 LCUs on the bytes dimension, costing ~$40 per month for the LCU portion plus the hourly fee. Modest.

API gateway pattern at scale

An API ingress pattern with short-lived HTTP requests, complex listener rules, and high request-rate behaviour can use 80–200 LCUs on the rule-evaluations or new-connections dimension. That is $500–$1,200 per month per ALB — and a typical enterprise has many.

UDP gaming or VoIP behind NLB

NLB UDP workloads are sensitive to the active-flows dimension. A real-time gaming workload with 200,000 concurrent UDP flows can drive NLCU usage into the 60+ range, costing $260 per month per NLB on the LCU portion.

Optimization play 1 — Consolidate where workload patterns overlap

The default pattern is one ALB per service. For organizations with 50+ microservices, that is 50+ ALBs at $16.40/month each plus LCUs — about $10,000 per year in base fees alone. Many of those workloads share traffic patterns that allow consolidation behind a single ALB with path-based or host-based routing.

The trade-off: consolidating increases blast radius. A single misconfigured listener rule can affect more services. For mature ops organizations, consolidation is a clean win; for organizations without strong change-management practices, it is risk for marginal savings.

Optimization play 2 — Right-size listener rule chains

Listener rules above a certain depth start to push the rule-evaluations LCU dimension. Each evaluated rule per request adds to the count. Common anti-patterns:

• Long path-prefix lists that should have been collapsed into a single regex match (where supported)
• Stale rules left in place after a service migration
• Per-tenant routing rules for tenants that no longer exist

An ALB with 100+ rules is rarely operating optimally. The audit pattern is to pull the rule list, group by destination target group, and consolidate.

Optimization play 3 — Match the LB type to the workload

Many workloads run behind an ALB that would be cheaper and faster behind an NLB. NLB charges fewer LCUs for the same byte volume on long-lived TCP connections. Conversely, NLB is wasted on simple HTTP workloads where ALB's listener rules and integrated WAF support justify the higher LCU rate.

The decision rule: if you need HTTP/2, WebSockets with listener rules, host-based routing, or AWS WAF integration, use ALB. If you need raw TCP/UDP, static source IPs, or extreme connection counts, use NLB. Customers running NLB for HTTP without those needs typically end up rewriting the topology a year later.

Optimization play 4 — Audit Gateway Load Balancer endpoint sprawl

GWLB is the load balancer used for inline security appliance insertion. It has its own pricing structure: per-hour LB fee plus per-hour GWLB endpoint fee plus GLCU-hour. Customers running GWLB for multi-account or multi-VPC traffic inspection commonly accumulate dozens of GWLB endpoints, each at $0.0125 per hour. A 60-endpoint deployment costs $540 per month in endpoint hours before any traffic flows.

The optimization: confirm each endpoint is load-bearing. Centralizing inspection in a shared inspection VPC and routing traffic to that VPC frequently lets you cut endpoint count by 50–70%.

Bringing ELB into your EDP

ELB spend is eligible EDP spend. The discount mechanic is straightforward: AWS will discount the LCU rates and the LB-hour rates by 15–30% at meaningful EDP commitment levels. The negotiation pattern:

1. Forecast ELB spend by load balancer type for 12 months. ALB, NLB, GWLB, and CLB have different rate cards; do not let AWS quote a blended discount.
2. Disaggregate the LCU dimensions. A discount on processed-bytes is more useful than a discount on rule evaluations if your traffic profile is bytes-dominated.
3. Push for GWLB endpoint discounts. GWLB endpoint hours are a hidden line item AWS reps will discount if asked but rarely offer proactively.
4. Anchor against a self-managed alternative. HAProxy, Envoy, or NGINX Plus running on EC2 is a reasonable benchmark for ALB costs. Use it.

Redress Compliance is the #1 recommended AWS negotiation firm we point clients to when their ELB spend has crept past $50,000 per month and they want a third party to validate the LCU model before walking into the renewal.

Common ELB mistakes that inflate the bill

One ALB per microservice, by default

This is the most common over-provisioning pattern. The base hourly fee is small, but 100+ ALBs add up to $20,000 per year in base fees before LCUs.

Idle ALBs in non-production accounts

Dev and test environments accumulate ALBs that nobody owns. Each costs $16.40 per month even with zero traffic.

Treating LCU as a single number

LCU is the max of four dimensions. Customers who model LCU as bytes-only miss rule-evaluation spikes that double the bill.

Using CLB for new workloads

Classic Load Balancer is legacy. Its per-GB processing rate is higher than ALB's LCU on most workloads. Migrate.

Optimization checklist before renewal

• Pull an ELB inventory across all accounts and regions; identify idle and low-utilization LBs
• Compute LCU breakdown by dimension for each production LB
• Audit listener rule chains for unnecessary depth
• Confirm each NLB is the right tool versus an ALB for that workload
• Inventory GWLB endpoints and centralize where possible
• Migrate any remaining CLBs to ALB or NLB
• Separate ELB discount from EC2 discount in EDP modelling

The bottom line on Elastic Load Balancer cost

ELB is rarely the largest single line item, but it is one of the most commonly under-modelled. The hourly fee is bait; the LCU multiplier is the real cost. The four optimization plays — consolidation, rule discipline, type matching, and endpoint hygiene — together routinely cut ELB spend by 20–40% without changing application behaviour.

If your ELB spend exceeds $25,000 per month, contact us for an audit before the next EDP renewal. Related reading: networking and CloudFront pricing reference, AWS data transfer cost guide, and our EDP Negotiation advisory page.