Cloud Agentics Talk to an engineer

Work / Cost optimization

Cost optimization, January 2026

A 60 to 90 day forecast of AWS cost

An AWS bill can rise by 30 to 50 % in one month when nobody tracks resource growth. We forecast the bill 60 to 90 days ahead, so the client acts before the increase is invoiced.

60 to 90days of forecastdelivered each month
18 kUSD saved on RDS storagegrowth found before the extension
0deployments blocked by a quotain 12 months
23 %higher return on Reserved Instancesrightsizing before the commitment

Bills that rise without warning

Several clients saw their AWS bill rise by 30 to 50 % in one month because nobody tracked resource growth. By the time the invoice arrives, the client has paid for weeks of the increase. Three cases show where the increase comes from.

RDS storage

The RDS instance was set to extend its storage automatically. We forecast that it would reach 2 TB in six weeks. The client archived old data first and avoided an increase of 400 USD per month.

ElastiCache cluster

The cache hit ratio was falling, so the cluster would have needed more nodes. We forecast the cost and optimized the cache keys instead. The client avoided an upgrade of 1,200 USD per month.

Lambda concurrency

A deployment of a security patch reached the account's concurrent execution quota. Production was blocked for 3 hours while the quota increase was requested. We now check quotas and request increases 30 days ahead.

What we track

Each month we analyse growth and deliver a forecast for the next 60 to 90 days with the actions to take. The forecast covers three groups of measures.

RDS allocated storage approaching the next pricing tier S3 bucket growth and the effect of lifecycle policies EBS volumes approaching their provisioned IOPS ElastiCache memory utilization and eviction rate
Lambda concurrent executions against the account quota API Gateway throttling limits EC2 instance quotas by type and region RDS instance count per account
Services that scale automatically and raise the bill Data transfer cost trend NAT gateway usage surges CloudWatch Logs retention cost growth

How we prevent blocked deployments

One client could not deploy a security patch because the Lambda concurrent execution quota was fully used. The emergency quota increase took 3 hours. The process below prevents a repeat.

Weekly check Growth projection Quota request, 30 days ahead Deployment
1

Weekly check

We compare every service quota with current usage.

LambdaAPI GatewayEC2RDS
2

Growth projection

We calculate when each quota will be reached from the 30, 60 and 90 day trends.

30 day trend60 day trend90 day trend
3

Quota request

We submit the AWS quota increase 30 days before the projected date.

30 days before the limit
4

Deployment

Security patches and new code can be deployed at any time.

0 deployments blocked in 12 months

Reserved Instances: commit after the data, not before

Reserved Instances offer a discount of up to 72 %. Clients lose money when they commit too early. We have seen four causes, and we require four checks before we recommend a purchase.

Four causes of a lost discount

Wrong instance size

A reservation for r6g.xlarge when the workload needs r6g.large. 50 % of the reservation is wasted.

Change of usage pattern

A three-year reservation, then a migration to containers six months later.

Region lock-in

A reservation in us-east-1, then a multi-region requirement for compliance.

Over-provisioned baseline

A reservation sized for peak load when average utilization is 40 %.

Four checks before a purchase

90 day review

A review of utilization across all hours, not only averages, with rightsizing first.

Stability check

No migration, refactoring or architecture change planned in the next 12 months.

Savings Plans compared with Reserved Instances

Compute Savings Plans often give more flexibility at a similar discount. We calculate both.

Staged commitment

One-year terms on 60 % of the baseline, more after six months of stable usage.

The figures

18 kUSD saved on RDS storagestorage growth found before the automatic extension
0deployments blocked by a service quotain 12 months, quotas checked weekly and increased 30 days ahead
23 %higher return on Reserved Instancesrightsizing before the commitment instead of an early purchase

Act 30 days before the limit

Every forecast carries the date on which a storage tier, a service quota or a commitment will be reached. A quota increase is requested 30 days before that date. In 12 months, no deployment was blocked by a service quota.

What the client receives

  • A forecast of the AWS bill for the next 60 to 90 days, delivered each month.
  • The date on which each storage tier, service quota or commitment will be reached, with the action to take.
  • Quota increase requests submitted 30 days ahead.
  • A Reserved Instance or Savings Plan recommendation only after 90 days of utilization data and a stability check.