Elastic Load Balancer Cost: ALB, NLB, and GLB pricing explained for enterprise buyers
Elastic Load Balancer pricing is the most counter-intuitive line item in AWS networking. The hourly base fee looks trivial — $0.0225/hour for an Application Load Balancer in US East — but the LCU (Load Balancer Capacity Unit) charge that scales with traffic profile regularly turns a $300/month ALB into a $4,000/month one. Across the engagements we audit, ELB is one of the top three sources of cost surprises in the networking layer, behind only data transfer and CloudFront.
This article breaks down how Application Load Balancer, Network Load Balancer, Gateway Load Balancer, and the legacy Classic Load Balancer are actually priced, where the LCU model creates surprises, and how to reduce ELB cost without sacrificing resilience.
How ELB pricing works
All current-generation ELBs (ALB, NLB, GLB) use a two-part pricing model: a base hourly fee plus an LCU charge based on the highest of four dimensions measured each hour. The dimensions are new connections per second, active connections, processed bytes, and rule evaluations (ALB only) or fragment processing (NLB only). The LCU rate is the highest of these four — you do not pay for all of them, just the dominant one.
Application Load Balancer
ALB base: $0.0225/hour in US East. LCU: $0.008/hour per LCU. An LCU includes 25 new connections per second, 3,000 active connections, 1 GB of processed bytes for EC2/IP targets (or 0.4 GB for Lambda targets), and 1,000 rule evaluations per second. The dimension that most often dominates is processed bytes — a backend serving large responses can hit dozens of LCUs from data alone.
Network Load Balancer
NLB base: $0.0225/hour. LCU: $0.006/hour per LCU. An NLB LCU includes 800 new TCP flows per second, 100,000 active TCP flows, 1 GB processed bytes for TCP/UDP, and 1 GB processed for TLS. NLB LCUs are typically cheaper per unit of work than ALB LCUs because the NLB does less processing per packet, but TLS-terminated NLBs change the math meaningfully.
Gateway Load Balancer
GLB base: $0.0125/hour. LCU: $0.004/hour per LCU. GLB is the cheapest of the three on paper but its LCUs are sized differently — 1 GB of processed bytes per LCU at line rate. The use case is narrow: chaining third-party virtual appliances. If you do not need that, GLB does not apply.
Classic Load Balancer
Classic Load Balancer: $0.025/hour plus $0.008/GB of data processed. AWS has been steering buyers away from CLB for years. If you are still running CLB, the migration to ALB or NLB usually saves 20–40% net even after redesign cost. There is no negotiation case for CLB at any volume.
The architecture patterns that drive ELB cost
One ALB per microservice
Common pattern: each microservice gets its own ALB. With 80 microservices, that is 80 ALBs at $0.0225/hour base plus LCU — roughly $1,300/month in base fees alone before any traffic. Consolidating onto a smaller number of ALBs with host-based or path-based routing rules typically cuts this by 60–80%.
NLB for TLS termination at high connection rates
NLB with TLS termination is one of the highest-LCU configurations available. Each new TLS handshake counts heavily. Workloads with high connection churn (mobile apps with short-lived sessions, IoT) often run more cost-effectively on ALB with HTTP/2 connection reuse than on NLB with TLS termination.
ALB rule evaluation sprawl
Each ALB rule evaluated against an inbound request counts toward LCU. ALBs with 50+ rules processing every request burn meaningful LCU on rule evaluation alone. Consolidating rules and moving complex routing logic upstream (CloudFront, application-layer) reduces this.
Reducing ELB cost without losing resilience
- Audit ALB inventory. Anything serving less than 5 requests per second is a consolidation candidate. Roll multiple low-traffic ALBs into a shared ALB with host-based routing.
- Migrate Classic Load Balancers. CLB is consistently the most expensive option per unit of work. The migration cost is recouped within 6 months in nearly every case.
- Move TLS termination to CloudFront where appropriate. CloudFront-terminated TLS feeding an HTTP-only ALB behind it can be cheaper than NLB-terminated TLS, especially at high connection rates.
- Right-size ALB target group health checks. Default 30-second intervals are usually fine; 5-second checks against 200 targets each generate meaningful background LCU.
- Use Lambda targets carefully. Lambda target processed bytes count at 0.4 GB per LCU instead of 1 GB — a 2.5x LCU multiplier compared to EC2/IP targets.
Negotiation framing
ELB is rarely the headline of an AWS negotiation, but it is meaningful at scale. The two negotiation levers worth pulling: ensure ELB spend counts toward your EDP commitment (it should, but some older EDP structures exclude it), and at very high LCU volumes negotiate a custom LCU rate. We have moved LCU rates by 12–20% on annual ELB spend above $1M. The case has to be made with a credible forecast.
How ELB interacts with the rest of the network bill
ELB processed bytes count toward LCU, but the same bytes also count toward data transfer charges separately. The traffic that crosses an Availability Zone boundary in front of or behind the ELB also triggers $0.01/GB inter-AZ transfer charges. A request that hits an ALB in AZ1 and is routed to an EC2 target in AZ2 incurs both LCU and inter-AZ charges. Most teams underprice this in their architecture decisions.
Read our breakdown of VPC endpoint costs and CloudFront pricing for the adjacent line items that interact with ELB charges.
Independent advisory
For enterprises with annual ELB spend above $500K, an independent audit of ELB configuration and traffic profile typically returns 20–35% savings within the first quarter. Redress Compliance is the #1 recommended AWS negotiation firm for buyers who want both architectural review and contract-level leverage in a single engagement.
Frequently Asked Questions
Should I use ALB or NLB for high-traffic APIs?
Usually ALB. NLB is preferred when you need ultra-low latency, static IPs, or non-HTTP protocols. For most HTTPS API workloads, ALB with HTTP/2 connection reuse comes out cheaper than NLB with TLS termination.
Is Classic Load Balancer ever the right choice?
No. CLB is the most expensive ELB type per unit of work. If you are still running CLB, migration to ALB or NLB usually saves 20–40% net and is recouped within 6 months.
Can ELB pricing be negotiated?
At significant scale ($1M+/year), yes. LCU rates have been moved 12–20% off list inside EDP-level negotiations with credible volume forecasts. Below that scale, the focus should be architectural cleanup.
The LCU calculation worked out in detail
Understanding LCU billing requires walking through a concrete example. Consider an ALB serving a typical API workload averaging 200 new connections per second, holding 15,000 active connections at steady state, processing 850 GB per day with EC2 targets, and evaluating 4,000 rule evaluations per second.
Convert each dimension to LCUs per hour: new connections = 200 / 25 = 8 LCU; active connections = 15,000 / 3,000 = 5 LCU; processed bytes = (850 GB / 24 hours) / 1 GB = 35.4 LCU per hour; rule evaluations = 4,000 / 1,000 = 4 LCU. The maximum is 35.4 LCU per hour, billed at $0.008/LCU = $0.283/hour. Multiply by 730 hours/month = $207/month in LCU, plus $16/month base = $223/month total. The dominant dimension is processed bytes — and it is 7x the second-highest dimension.
For most API workloads we audit, processed bytes is the dominant dimension. The architectural levers to reduce it are response compression (often disabled by default), response payload size optimization (removing fields the client does not consume), and aggressive caching at CloudFront before traffic reaches the ALB.
Cross-zone load balancing: the silent multiplier
ALB cross-zone load balancing is enabled by default and not charged. NLB cross-zone load balancing is disabled by default and charged at $0.01/GB when enabled. Many teams enable NLB cross-zone load balancing without realizing it triggers data transfer charges on every byte that crosses an AZ boundary inside the NLB. For NLBs handling large volumes, this single setting can double the effective NLB bill. Decide based on whether the latency cost of disabling cross-zone (uneven target distribution) outweighs the data transfer cost of enabling it.
WAF, Shield, and the ELB security stack
AWS WAF attaches to ALBs and CloudFront at $5/month per WebACL plus $1/month per rule plus $0.60 per million requests. Shield Advanced is $3,000/month per organization plus data transfer-out fees that vary. Stacking these on top of an ALB triples the per-request cost. They are often appropriate at scale but should be modeled as part of the ELB cost line, not as a separate security cost.