Northflank Containers

What Northflank actually is

Northflank is a developer platform that treats deployment as a product: you connect GitHub, GitLab, or Bitbucket, it builds Docker images from your commits, and it runs your services, databases, cron jobs, and background workers on Kubernetes it manages for you either on Northflank's own cloud or inside your own AWS, GCP, Azure, or Oracle account. The company is Northflank Ltd, based in London, founded in 2019 by Will Stewart and Frederik Brix, who met as teenagers through multiplayer gaming and got into infrastructure the honest way: running game servers on Rancher and Mesos before Kubernetes existed. In November 2024 it announced $22.3 million in new financing: a $16 million Series A led by Bain Capital Ventures plus a $6.3 million seed led by Vertex Ventures US, bringing total disclosed funding to roughly $24 million.

The product surface is broad and unusually coherent. The CI/CD side gives you build pipelines from any commit, stored builds with one-click rollback, release pipelines across preview, staging, and production, and ephemeral preview environments spun from pull requests. The runtime side covers services, one-off and scheduled jobs, managed databases and addons, secret groups, persistent NVMe volumes, autoscaling, RBAC, and a choice of UI, CLI, API, and GitOps control. The part that separates it from the Heroku-likes is the AI workload support: GPU containers (L4, A100, H100), time-slicing and MIG multi-tenancy on GPU nodes, Jupyter notebooks, secure execution for untrusted code-generation workloads, spot instances, and read-write-many volumes: features aimed at inference, agents, and training jobs rather than just web apps.

The second differentiator is bring-your-own-cloud. On BYOC, Northflank runs its platform against a cluster in your VPC (GKE on Google Cloud, EKS on AWS, and so on), you pay your cloud provider directly for the infrastructure so existing commitments and credits keep working, and Northflank bills only for the node vCPU it manages. The pitch on the homepage is accurate as a summary: deploy any project in seconds, in our cloud or yours. The company cites 2,000+ teams including Sentry, Writer, and Chai Discovery, and claims 130 billion-plus requests processed and millions of containers run since 2019.

Pricing: what you really pay

Northflank publishes per-second rates and a compute ladder, which is more transparent than most of the platform market. Northflank Cloud: CPU at $0.01667 per vCPU-hour, memory at $0.00833 per GB-hour, ephemeral storage free, NVMe persistent disk at $0.11657 per GiB-month (first 3,000 IOPS and 140 MiB/s per disk included; extra IOPS $7.37 per 1,000 IOPS-month, extra throughput $5.85 per 100 MiB/s-month), network egress at $0.06 per GB with ingress free, and a dedicated-IP egress gateway at $15 per month. BYOC bills $0.02083 per managed node vCPU-hour and nothing else: no memory, disk, or egress line items from Northflank.

The compute plans translate that into familiar shapes: nf-compute-50 (0.5 vCPU, 1 GB) at $12.00 per month, nf-compute-100-2 (1 vCPU, 2 GB) at $24.00, nf-compute-200 (2 vCPU, 4 GB) at $48.00, nf-compute-400 (4 vCPU, 8 GB) at $96.00, nf-compute-800-16 (8 vCPU, 16 GB) at $192.00, and 21 plans in total up the ladder with custom combinations available. GPUs are hourly: NVIDIA L4 at $0.80/hour, A100 40 GB at $1.42, A100 80 GB at $1.76, H100 at $2.74. There are no seat fees: unlimited users on every plan, and Northflank's own comparison table claims 36-45% below Render at the 2-8 vCPU tiers and 53% cheaper persistent disk.

The free tier is real: the Sandbox plan includes 2 services, 2 jobs, 1 database, up to $20 per month of resources on Northflank Cloud, one BYOC cluster up to 12 vCPU and 48 GB, and unlike most free PaaS tiers, always-on compute with no sleeping. The pricing page advertises no card needed, though the billing docs state that all users must add a payment method to create resources; we would verify which behavior applies at signup before planning around it. Enterprise is custom: a platform fee plus usage, 24/7 support and SLA, white-labeling (including CI/CD under your own brand), and volume discounts at spend thresholds.

How does that compare to raw infrastructure? A 1 vCPU/2 GB container at $24 per month is roughly 3-5x a small VM from the value providers on our benchmarks: you are paying for the build system, the panel, the rollback story, and the ops burden being someone else's problem. In the PaaS bracket it is well priced; the decision is which bracket you are buying.

Performance & infrastructure

Every compute plan runs on latest-generation Intel and AMD compute-optimized cores with NVMe storage, and the platform's Kubernetes foundation means horizontal and vertical scaling, custom autoscaling policies, and multi-region deployments are configuration rather than projects. Region coverage comes in two flavors: 6+ Northflank Cloud regions, and 600+ BYOC regions across AWS, GCP, Azure, and Oracle, which the marketing pages aggregate as 330+ availability zones. In practice, EU, US, and Asia-Pacific placement is available either way, and BYOC means the physical location is whatever your cloud account can reach.

Reliability signals are moderate but credible for a company of this age: 130 billion-plus requests processed, claims of millions of containers, marquee production customers like Sentry, and enterprise features (audit logs, RBAC, private networking, dedicated support) that suggest real production use rather than demo-ware. On the flip side, this is a Series A company of a few dozen people; the platform is young, the brand is unproven at decade scale, and there is no public status history or third-party uptime data we could point to. We have not load-tested Northflank and will not claim numbers we do not have.

Support splits by tier in the standard way: self-serve documentation and ticket support on Sandbox and pay-as-you-go, with 24/7 support and a contractual SLA reserved for Enterprise. The SLA is the honest dividing line: if availability is a contractual requirement for you, it is an Enterprise conversation, not a checkout option.

Where it falls short

The cons on our listing said less mature than Heroku and documentation could be improved, and both hold up, with nuance. Maturity: Heroku has fifteen years of operational scar tissue and an enormous add-on ecosystem; Northflank has strong engineering and a young track record. The gap is not features (Northflank's feature table is richer than Heroku's in 2026) but the boring institutional knowledge of failure modes at scale. Documentation: the docs cover the platform thoroughly and the pricing page is a model of clarity, but the middle layer (migration patterns, cost modeling for multi-service architectures, BYOC networking edge cases) is thinner than Kubernetes-native competitors, and some secondary surfaces disagree with each other: the pricing page says no card is needed for the free tier while the billing docs require a payment method, and the northflank.ai mirror publishes different unit rates than northflank.com.

Second, cost predictability on Northflank Cloud requires attention: per-second billing is fair, but disk, IOPS, throughput, egress, and gateway line items stack, and a database-heavy architecture will find that storage and network dominate the bill before compute does. The $20/month free allowance is generous but caps quickly against real workloads. Third, the platform abstraction is deep, which is the point but also the lock-in: pipelines, secret groups, and addon definitions are Northflank objects. BYOC mitigates infrastructure lock-in; application-level portability is still on you. Fourth, the AI feature surface (GPU time-slicing, secure code execution, agent runtimes) is advancing fast enough that documentation trails the product: budget engineering time to learn it properly.

Who should pick this

Our pick for engineering teams that want Heroku-grade developer experience with Kubernetes-grade control: startups shipping from GitHub with preview environments per pull request, platform teams standardizing deployment across AWS/GCP/Azure without rebuilding CI/CD per cloud, and AI teams needing GPU inference, notebooks, and untrusted-code execution behind one platform. The BYOC model makes it especially attractive for companies with existing cloud commitments: you keep your enterprise discounts and let Northflank drive the cluster.

We would pick a plain managed Kubernetes service instead when the team is already Kubernetes-fluent and wants maximum control per euro, and we would look at the simpler app platforms in our managed cloud comparison if the workload is a single web app and budget dominates. But for a growing product with services, jobs, databases, previews, and the occasional GPU workload, Northflank is one of the most complete self-serve platforms we have priced out, and its published rates let you check that claim yourself.

Key takeaways

• London-based Northflank Ltd (founded 2019 by Will Stewart and Frederik Brix); $22.3M raised in Nov 2024 ($16M Series A led by Bain Capital Ventures), roughly $24M total
• CI/CD from GitHub/GitLab/Bitbucket with per-PR preview environments, stored builds and rollback, cron jobs, managed databases, secrets, RBAC, and a UI/CLI/API/GitOps surface
• Transparent per-second pricing: $0.01667/vCPU-hour and $0.00833/GB-hour on Northflank Cloud; compute plans from $12/mo (0.5 vCPU/1 GB) to $192/mo (8 vCPU/16 GB); GPUs from $0.80/hour (L4) to $2.74/hour (H100)
• Free Sandbox tier: 2 services, 2 jobs, 1 database, up to $20/mo of resources, 1 BYOC cluster (12 vCPU/48 GB), always-on with no sleeping, but card requirements conflict between the pricing page and docs
• BYOC runs the platform in your AWS/GCP/Azure/Oracle VPC at $0.02083 per managed node vCPU-hour; you pay your cloud provider directly and keep existing commitments
• Enterprise adds the contractual SLA and 24/7 support; the platform is strong but the company is young, and track-record depth is the trade-off versus incumbents

Sources

• https://northflank.com/pricing
• https://northflank.com/docs/v1/application/billing/pricing-on-northflank.md
• https://www.northflank.ai/pricing
• http://www.northflank.io/
• https://www.businesswire.com/news/home/20241111904885/en/Northflank-Raises-22M-to-Simplify-App-and-Database-Deployment-in-Your-Cloud
• https://techcrunch.com/2020/07/07/northflank-announces-2-6m-seed-to-create-end-to-end-devops-workflow-in-cloud/
• https://tracxn.com/d/companies/northflank/__evAHwlxtCFdfVmkZVe470E8SpGCpgyNhN94GPbcB5Hw
• https://platform.tracxn.com/a/d/company/5ca219ec252ec22311b90c79/northflank