Cloud Bill Doubling Every Quarter: What a DevOps Engineer Finds and Fixes in 30 Days

As a DevOps Engineer at Acquaint Softtech, a software development partner, cloud cost audits are one of the most common requests I handle. The pattern is consistent: a SaaS company or startup has been growing, deployment frequency has increased, and the AWS or GCP bill has been climbing for six months without a clear reason. Nobody changed anything deliberately. The bill simply doubled. This article covers the seven places where cloud spend accumulates without producing business value, what a DevOps engineer does to find each one, and what the realistic savings look like within 30 days.

---

A cloud bill that doubles every quarter without a corresponding doubling of users or revenue is one of the clearest indicators that infrastructure is accumulating waste. Cloud providers make it easy to provision resources. They do not make it easy to find unused resources. A DevOps engineer who know where to look can typically identify 20 to 40% of cloud spend as either waste (resources consuming cost with no business value) or inefficiency (resources consuming more than necessary for the value they deliver).

The scalability infrastructure that prevents crashes during traffic spikes is covered in the traffic spike infrastructure guide. This article focuses specifically on the cost side of the cloud equation: what is consuming budget that should not be, and what the fix looks like.

The 7 Places Where Cloud Waste Accumulates: What I Find First

Every cloud cost audit I have run at Acquaint Softtech finds the same categories of waste in different proportions. Here is what to look for and where.

Unused or idle EC2 instances

Development and staging instances that were launched for a project and never terminated. Instances that run 24/7 when they only serve traffic during business hours. Oversized instances running at 5 to 15% CPU utilisation. On a 20-instance fleet, it is common to find 3 to 5 instances that are either completely unused or consistently underutilised. Terminating unused instances and right-sizing underutilised ones is the fastest cost reduction with zero risk to production.

Unattached EBS volumes and snapshots

When an EC2 instance is terminated, the root EBS volume is often retained rather than deleted. After 12 months, a typical AWS account has dozens of unattached EBS volumes accumulating cost at $0.10 per GB per month. Old snapshots compound this: automated snapshot policies create daily snapshots but snapshot deletion policies are not always configured. An account with 200 snapshots at 50GB average stores 10TB of snapshots for $250/month with no recovery plan that uses them.

Oversized RDS instances

A database instance selected during initial setup based on expected load rather than actual load. Six months later, the database runs at 20% average CPU and 30% memory utilisation. Downsizing an RDS db.r5.xlarge to a db.r5.large, for example, cuts RDS cost by 50% for that instance with no performance impact at the actual utilisation level.

Data transfer costs between services

AWS charges for data transfer between availability zones and between regions. An application that routes data from an EC2 instance in us-east-1a to a database in us-east-1b pays cross-AZ transfer fees on every database call. Placing the application and database in the same AZ, or using VPC endpoints for S3 and DynamoDB access, eliminates a category of cost that appears gradually and is difficult to attribute without explicit monitoring.

On-demand pricing for predictable workloads

Any workload that runs continuously or on a predictable schedule should be on Reserved Instances or Savings Plans, not on-demand pricing. On-demand EC2 is 40 to 60% more expensive than a 1-year Reserved Instance or Compute Savings Plan. A fleet of 10 on-demand m5.large instances at $0.096/hour costs $8,410/month. The same fleet on a 1-year Savings Plan costs approximately $5,040/month, a saving of $3,370/month.

Oversized NAT Gateway usage

Applications that route all internet traffic through a NAT Gateway for private subnet instances pay per-GB data processing fees. A high-traffic application routing 10TB/month through a NAT Gateway pays approximately $450/month in NAT processing fees. For traffic destined for AWS services (S3, DynamoDB, CloudWatch), VPC Gateway Endpoints eliminate the NAT Gateway fee entirely.

CloudWatch log retention without lifecycle policies

Application logs written to CloudWatch Logs accumulate indefinitely if log group retention policies are not configured. CloudWatch Logs storage costs $0.03/GB/month. An application writing 50GB of logs per day stores 1.5TB per month. Without a 30-day retention policy, after 6 months the account stores 9TB of logs for $270/month, most of which is data nobody looks at.

The 30-Day Fix Sequence: What Gets Done in What Order

Not all cloud cost fixes are equal in effort and impact. A DevOps engineer executing a cost optimisation engagement prioritise the changes that produce the largest savings for the lowest implementation risk and effort. Here is the sequence I follow.

Week 1: Audit and quick wins

Run AWS Cost Explorer analysis, EC2 utilisation metrics, and CloudWatch data. Identify unused instances, unattached volumes, and orphaned snapshots. Terminate unused resources and set EBS volume deletion on instance termination for new instances. Configure CloudWatch log retention policies. These changes cost zero in terms of risk and typically produce $300 to $3,000 in immediate monthly savings depending on account size.

Week 2: Right-sizing and Reserved Instance planning

Review CPU and memory utilisation across the instance fleet. Downsize overprovisioned instances. Generate Reserved Instance and Savings Plan recommendations from AWS Cost Explorer. Purchase Savings Plans for the confirmed baseline of on-demand usage. This is the change that produces the largest sustained monthly saving in most accounts: 30 to 50% reduction in EC2 and Fargate costs for predictable workloads.

Week 3: Data transfer and networking costs

Audit NAT Gateway and data transfer charges in the AWS bill. Configure VPC Gateway Endpoints for S3 and DynamoDB. Review cross-AZ traffic patterns and consolidate where possible. These changes require careful architecture review to avoid unintended changes to network security, but produce significant ongoing savings for high-traffic applications.

Week 4: Monitoring, alerting, and governance

Configure AWS Budgets alerts at 80% and 100% of the target monthly spend. Set up Cost Anomaly Detection to flag unexpected spending increases. Tag all resources with environment, team, and project labels to enable cost attribution. Create a monthly cost review dashboard in CloudWatch or Cost Explorer. From this point forward, cost growth is visible and attributed before it becomes a surprise.

For teams also evaluating Kubernetes cost optimisation with spot instances, the Kubernetes spot instances guide covers the specific configuration for running Kubernetes workloads on spot instances, which can reduce compute costs by an additional 60 to 70% beyond right-sizing. For SaaS teams wanting a full DevOps budget breakdown before engaging, our DevOps for SaaS startups guide covers what to budget for DevOps infrastructure across different product stages.

What the Savings Look Like: Real Numbers by Account Size

The savings from a cloud cost audit depend on the account's size, age, and how much governance has been applied previously. Here are the realistic ranges by monthly spend bracket.

Monthly cloud spend	Typical waste found	Realistic 30-day savings
$1,000 to $5,000/month	15 to 25% waste	$150 to $1,250/month saved
$5,000 to $15,000/month	20 to 35% waste	$1,000 to $5,250/month saved
$15,000 to $50,000/month	25 to 40% waste	$3,750 to $20,000/month saved
$50,000+ per month	30 to 45% waste	$15,000 to $22,500+/month saved
DevOps engineer cost (monthly retainer)	$5,000 to $7,000/month (senior, Acquaint Softtech)
Break-even point	A $5K/month platform saves its DevOps cost in month 1 if it has $15K spend

The full DevOps engineer rate comparison by region and seniority level is in the DevOps engineer cost guide.

Acquaint Softtech's hire DevOps engineers service provides vetted engineers who conduct cloud cost audits as part of every engagement. Profiles delivered within 24 hours. Engineer in your first standup in 48 hours.

For individual DevOps capacity on a monthly retainer, our staff augmentation model fits. For a vendor-managed DevOps function with broader team structure, dedicated development teams covers the full engagement.

How to Keep Cloud Costs Under Control After the Initial Audit

The initial audit produces the biggest savings. Sustaining them requires ongoing governance. Here is what a DevOps engineer put in place to prevent cloud spend from drifting back up.

AWS Budgets and Cost Anomaly Detection

AWS Budgets sends an alert when monthly spend crosses 80% of the target, giving the team two weeks to investigate before the budget is exceeded. Cost Anomaly Detection uses machine learning to identify spending patterns that deviate from the historical norm and alerts within 24 hours of an anomalous increase. Together, these eliminate the surprise cloud bill.

Resource tagging and cost allocation

Every resource is tagged with environment (production/staging/development), team, and project. Cost allocation tags enable the team to see which part of the infrastructure costs what. This makes budget conversations specific and accountability clear. A team that knows their staging environment costs $2,400/month will make different decisions about leaving test instances running.

Monthly right-sizing review

AWS Compute Optimizer provides ongoing right-sizing recommendations as utilisation patterns change. A DevOps engineer reviews these recommendations monthly and applies any instance changes that produce meaningful savings without performance risk. Over 12 months, this prevents the gradual return of overprovisioning that caused the original cost increase.

Automated cleanup for non-production resources

A Lambda function or AWS Systems Manager automation runs on a schedule to stop non-production instances outside of business hours and terminate instances older than a defined age in the development environment. This eliminates the most common source of accumulated waste: test instances that someone meant to terminate but forgot about.

Frequently Asked Questions