Five myths about Fly.io’s 2026 global container hosting pricing

Fargate is not free.

Fargate compute is not the complete bill. While AWS Fargate compute costs seem low, a single task running in a public subnet with a public IP incurs a $0.005 per hour charge for the dedicated Elastic Network Interface. You also pay $0.045 per hour for NAT Gateway uptime. Fargate bills for provisioned resources, not utilized resources. An idle task at 0% CPU costs the same as one running full throttle. AWS Lambda also adds a mandatory API Gateway tax of $1.00 to $3.50 per million requests.

Fargate also charges for ephemeral storage above the 20 GB free tier. Additionally, the billing clock starts when Fargate begins pulling your container image from ECR. For a 2 GB container image, a cold pull adds 90 to 120 seconds of billable time. You must also account for CloudWatch Logs ingestion at $0.50 per GB. Fargate task sizes range from 0.25 vCPU up to 32 vCPU, and Windows containers require an additional OS license fee and a 5-minute billing minimum. AWS Fargate provides specific memory ranges for each CPU, such as 16 vCPU tasks supporting up to 120 GB of RAM.

AWS Lambda uses a unique GB-second model where memory and compute are bundled together. Users choose a memory allocation from 128 MB to 10 GB, and CPU power scales proportionally. This makes direct vCPU-to-vCPU comparison with other platforms difficult.

Cloud Run hides costs in egress and minimum instances. If you enable minimum instances to eliminate cold starts, those instances bill at the full compute rate even when idle. Cloud Run also charges $0.12 per GB for egress in most regions.

Egress costs add up.

Scaling through request volume

At 5 million requests per month, Fly.io costs $2.02. For the same workload using 0.5 vCPU and 512 MB of memory, Cloud Run costs $9.23. This Cloud Run total includes 320,000 billable vCPU-seconds at $7.68, 140,000 billable GiB-seconds at $0.35, and 3 million billable requests at $1.20. AWS Lambda reaches nearly $10 because the mandatory API Gateway surcharge adds $1.00 to $3.50 per million requests.

Fly.io wins.

Railway and Render become expensive at scale because their always-on compute models cannot optimize for bursty traffic. At 10 million requests per month, the AWS Lambda API Gateway surcharge reaches $10 to $35. Cloud Run request-based pricing remains efficient, and Cloud Run can handle up to 1,000 concurrent requests per instance. However, the gap between Fly.io and the major providers becomes material as request volume increases.

Regional egress and free tier realities

Cloud Run offers a monthly free tier of 180,000 vCPU-seconds and 360,000 GiB-seconds, but you must use specific regions like us-central1 to receive these credits. If you deploy to Europe or Asia, you do not get the free tier. Cloud Run also charges $0.40 per million requests beyond the free allotment and provides 1 GB of egress free per month in North America.

Fly.io’s egress is metered by region group. In North America and Europe, Fly.io charges $0.02 per GB. In Africa and India, that cost rises to $0.12 per GB. For private network transfer, Fly.io charges $0.006 per GB in North America and Europe, but $0.050 per GB in Africa and India. Fly.io also charges $0.15 per 1GB of rootfs for a Machine stopped for 30 days and $2 per month for a dedicated IPv4 address. You can also purchase wildcard certificates for $1 per month.

Fly.io offers reservation blocks for shared machines, such as $36 per year for $5 per month of usage. This provides a 40% discount for those who pay upfront.

You should verify your regional egress rates before you scale. Can scale-to-zero platforms always remain competitive as traffic grows?

Fly.io remains the most cost-effective choice for high-request workloads.