AWS free tier vs paid services: Which tier actually makes sense for SMBs

AWS offers a free tier to let developers and organisations explore services without immediate cost. The free tier is genuinely useful. It is also routinely misunderstood, leading to two common mistakes: trying to build production workloads on free tier limits and receiving surprise invoices when usage exceeds those limits, or dismissing the free tier entirely and paying for services that a scaling startup could run free for months.

This guide covers what the free tier actually covers for services SMBs use most, where it runs out, and how to make the transition to paid services without bill shock.

The three types of free tier offers

Not all free tier offers are the same. AWS uses three distinct categories:

Always free: Services that remain free forever up to a specified limit, regardless of how long you have been an AWS customer. Examples: 1 million Lambda function requests per month, 25 GB DynamoDB storage, 1 million CloudWatch custom metrics per month. These are permanent, not expiring.

12 months free: Services that are free for the first 12 months after your AWS account creation date. The clock starts when you create the account, not when you first use the service. 750 hours per month of t2.micro or t3.micro EC2 (enough for one instance running continuously), 5 GB S3 standard storage, 750 hours RDS (t2/t3.micro, single-AZ), and 15 GB of data transfer out per month fall into this category.

Free trials: Short-duration trials for specific services, typically 30-90 days from first activation. Amazon Redshift Serverless (90-day trial), Amazon Detective (30 days), Amazon Macie (30 days). These do not start when you create the account but when you activate the specific service.

Critical distinction: 12-month free tier expires based on account creation date. If you opened an AWS account for personal exploration and are now using it for business, your 12-month free period may already have expired even if you only recently started using EC2 seriously.

What SMBs actually get free (and what they do not)

Lambda (always free): 1 million requests and 400,000 GB-seconds of compute time per month, always. For many SMB applications with moderate traffic, Lambda stays within the free tier indefinitely. A function invoked 500,000 times per month with a 200ms average duration and 512MB memory uses roughly 50,000 GB-seconds, well inside the free threshold.

EC2 (12 months, then paid): The 750 hours/month of t2.micro or t3.micro is enough for one instance running around the clock for the first year. For actual production workloads, t2.micro is rarely sufficient (1 vCPU, 1 GB RAM). The free tier lets you learn EC2 without cost, not run production on it.

S3 (12 months, then paid): 5 GB storage, 20,000 GET requests, 2,000 PUT requests per month. This covers experimentation. A production application generating significant S3 activity will exhaust the free tier quickly. 5 GB is also not much for media files, backups, or log storage.

RDS (12 months, then paid): 750 hours of db.t2.micro or db.t3.micro single-AZ. Enough for a development database. Not suitable for production: single-AZ means no automatic failover, and t2/t3.micro is too limited for most production database workloads.

CloudFront: 1 TB data transfer out to the internet and 10 million HTTP/HTTPS requests per month, always free. For SMBs serving static assets through CloudFront, this covers substantial traffic before charges begin.

CloudWatch: 10 custom metrics, 10 alarms, 5 GB log ingestion, and 5 GB log storage, always free. A basic monitoring setup for a small application fits within this. A production monitoring setup with dozens of custom metrics and centralised logging quickly exceeds it.

Where the free tier ends unexpectedly

Several common AWS services have no free tier at all:

  • NAT Gateway: Charged per hour and per GB processed from the moment it is deployed. A NAT Gateway running continuously costs approximately £30-40 per month even with no traffic. Many tutorials deploy a NAT Gateway without noting that it immediately starts billing.
  • Application Load Balancer: No free tier. Charged per hour and per LCU (Load Balancer Capacity Unit). A lightly loaded ALB costs approximately £15-20 per month.
  • ElastiCache: No free tier for Redis or Memcached. Even the smallest cache.t2.micro instance costs approximately £15-20 per month.
  • Data transfer out to the internet: The free 15 GB per month (12-month free tier) is consumed quickly by any application serving content to users. Above 15 GB, standard egress rates apply.

Transitioning from free tier to paid

Set up AWS Cost Explorer and Budgets before you reach the end of the free tier period. The transition from £0 to paying invoices is not the problem; the unexpected first invoice because nobody was monitoring usage is.

Create a budget at £50-100 for your first billing period after the 12-month anniversary. Alert at 50% and 80%. The first invoice after the free tier expires is educational: it shows exactly which services are generating what costs and lets you make informed decisions about Reserved Instances and right-sizing.

Reserved Instances for predictable workloads: Once you understand your usage pattern (EC2 instances that run continuously, RDS databases that are always on), Reserved Instances save 30-40% compared to on-demand pricing with a 1-year commitment, or 60-70% with a 3-year commitment. Do not commit until you have at least 2-3 months of on-demand billing data to confirm the instance types and sizes you actually need.

Savings Plans: More flexible than Reserved Instances. Compute Savings Plans apply a discount to any EC2 usage (regardless of instance type or region) in exchange for a commitment to a specific spend level per hour. Easier to manage than Reserved Instances when your instance mix is still evolving.

Where Critical Cloud comes in

Moving past the free tier is when AWS cost management becomes genuinely complex: multiple services, data transfer costs, multiple pricing dimensions per service, and Reserved Instance commitments to manage. We run AWS cost optimisation for technology-led businesses as an ongoing operational function rather than a periodic review. As the world's first Powered by Datadog accredited partner, we surface spend anomalies and rightsizing opportunities in the same operational view as application performance. See how Critical Support works.