I'm leaving the cloud! (...and why you probably should too)

By Simon Høiberg

Share:

Key Concepts

  • Cloud Computing: Utilizing remote servers hosted on the internet to store, manage, and process data, rather than a local server or personal computer.
  • Bare Metal Servers: Physical servers dedicated to a single tenant, offering direct access to hardware resources.
  • Self-Hosting: Running applications and services on your own infrastructure, whether physical servers or virtual machines.
  • Vendor Lock-in: A situation where a customer is dependent on a vendor for products and services, and cannot easily move to a competitor.
  • Serverless Infrastructure: A cloud computing execution model where the cloud provider dynamically manages the allocation and provisioning of servers.
  • Docker: A platform for developing, shipping, and running applications in containers.
  • Kubernetes: An open-source system for automating deployment, scaling, and management of containerized applications.
  • Open-Source Software: Software with source code that anyone can inspect, modify, and enhance.
  • ARR (Annual Recurring Revenue): The predictable revenue a company expects to receive from its customers over a year.
  • DDoS (Distributed Denial of Service) Attack: A malicious attempt to disrupt the normal traffic of a targeted server, service or network by overwhelming it with a flood of internet traffic.

Transition from Cloud to Dedicated Servers

The speaker, Simon Hoyberg, who runs a portfolio of five SaaS tools used by over 50,000 users and approaching $2 million in ARR, has transitioned his business from running entirely on AWS to using dedicated bare metal servers hosted by a German company called Hetzner. Currently, 80% of his systems run on dedicated servers with open-source software, while 20% remain on AWS. This move has resulted in significant cost reductions and performance improvements, opening up new business opportunities.

Previous AWS Setup:

  • Compute: Amazon Elastic Container Service (ECS), AWS Fargate, AWS Lambda functions.
  • App Hosting: AWS Amplify.
  • Data Storage: Amazon DynamoDB.
  • Message Queues: Amazon Simple Queue Service (SQS).
  • Monitoring & Logging: Amazon CloudWatch.

This setup was entirely managed by AWS, with billing based on exact usage (milliseconds of code execution, database reads/writes, log storage duration).

New Hetzner Dedicated Server Setup:

  • Infrastructure: Three dedicated servers from Hetzner.
  • Compute & Orchestration: Docker containers managed by Kubernetes.
  • Data Storage: PostgreSQL.
  • Caching: Redis.
  • Message Queues: BullMQ.
  • Monitoring & Logging: Grafana and Prometheus.

This new setup utilizes open-source and free software (Kubernetes, PostgreSQL, Redis, BullMQ, Prometheus, Grafana Community Edition). The primary cost is the fixed monthly fee for the Hetzner dedicated servers (around $200 per month), plus the cost of the remaining 20% of infrastructure still on AWS. The speaker clarifies that "self-hosting" in this context means managing dedicated servers rather than truly owning and housing the hardware.

Top Three Reasons for Moving Off the Cloud

1. Dramatic Cost Reduction

  • AWS Costs: At the beginning of 2025, the projected AWS cost was $7,800 per month, increasing monthly with user growth due to the pay-per-usage model.
  • Hetzner Costs: After moving 80% of infrastructure, the total cost is now below $2,000 per month, with an aim to go below $1,000 per month. The Hetzner servers themselves cost approximately $200 per month.
  • ROI Analysis: While the initial migration involved significant time investment in learning Kubernetes, Docker, and system administration, the speaker argues that the upfront time cost is fixed. In contrast, the cost savings are potentially unlimited and increase over time. Time spent on infrastructure management on AWS was not zero and could be complex. After an adjustment period, infrastructure management time is now comparable to or less than before, with increasing financial savings.

2. Eliminating Vendor Lock-in

  • AWS Dependency: Services like DynamoDB, Lambda, and Fargate are AWS-specific. If AWS were to significantly raise prices or terminate services, migrating would be a complex and time-consuming process, potentially jeopardizing the business.
  • Freedom of Choice: The Hetzner setup allows the products to run on any hosting company offering VPS or dedicated servers, providing flexibility and reducing dependency.
  • Demonstration of Portability: As a proof of concept, the speaker ran one of his products for two days on three Raspberry Pis in his office, highlighting the portability of his setup, though not recommending it for production due to performance limitations.
  • Mitigating Dependencies: While Stripe and OpenAI are still dependencies, the speaker is exploring alternatives like Coinbase for payments and independent GPU vendors for AI inference.

3. Enabling Lifetime Deals and Financial Planning

  • Business Model Challenge: The speaker sells SaaS products with both recurring subscriptions and lifetime deals (one-time purchase for perpetual access). Selling a one-time purchase with ongoing, rising operational costs is problematic for financial planning.
  • Fixed Costs Advantage: By moving to dedicated servers with fixed monthly costs, the speaker can more accurately forecast expenses and offer lifetime deals with less financial risk.
  • Guaranteeing Lifetime Access: To address user concerns about the longevity of lifetime deals, the speaker is working on making all products fully self-hostable. This allows customers to install and run the software on their own servers if the business is unable to host it, truly guaranteeing lifetime access.

Services Still on AWS and Challenges

Despite the significant migration, approximately 20% of the infrastructure remains on AWS due to specific challenges:

  • Storage (S3):

    • Challenge: Vendor lock-in. While Hetzner and Cloudflare R2 offer S3 compatibility and are cheaper, they still represent a form of lock-in.
    • Attempted Solution: Using Min.io, an open-source S3-compatible object store, with a 15 TB hard drive from Hetzner.
    • Outcome: The setup became complicated, especially regarding backups, and was not very stable, likely due to Min.io's design for distributed systems requiring multiple drives per node. The speaker reverted to S3.
  • User Authentication (Cognito):

    • Challenge: Security is paramount, and finding a robust, easy-to-manage open-source alternative is difficult.
    • Attempted Solution: Keycloak, an open-source option.
    • Outcome: The speaker found Keycloak's user and developer experience to be poor, deeming it too risky to implement for such a critical function and returning to AWS Cognito.
  • Public Entry Point (CloudFront):

    • Reason for Keeping: AWS CloudFront provides essential security features like firewalls and DDoS prevention, which the speaker would find difficult to implement and manage on his own.
  • AI Services (OpenAI, Replicate, Vast AI):

    • Challenge: Current standalone servers lack the GPU power for reliable, self-hostable AI inference.
    • Hope: The speaker anticipates that future standalone servers will offer sufficient GPU capabilities.

Conclusion and Recommendation

The decision to move off the cloud and self-host depends on the business's stage and goals:

  • Hobby Projects: For small, non-scaling hobby projects, staying on managed cloud services like Vercel, Netlify, or Blobable is recommended.
  • Large-Scale or Mission-Critical Businesses: For businesses requiring significant scaling capabilities or those that are mission-critical, managed hyperscaler platforms like AWS are the preferred choice.
  • Mid-Range Businesses: For businesses in the middle, considering dedicated servers is a sensible option, offering significant cost savings and the ability to manage infrastructure.

The speaker also highlights the benefit of leveraging open-source tools for internal business operations, citing a separate video where he saved $10,000 annually by replacing expensive proprietary tools with self-hostable alternatives. The core takeaway is that while moving off the cloud requires an upfront investment in learning and setup, the potential for cost savings and increased control is substantial.

Chat with this Video

AI-Powered

Load the transcript when you're ready to chat so the initial page stays lighter.

Ready to summarize another video?

Summarize YouTube Video