How many times have you seen someone build an MVP with all the cloud bells and whistles, only to watch it go nowhere?
The product had Lambda functions. It had API Gateway. It had Cognito. It had S3, CloudFront, DynamoDB, CloudWatch, IAM policies, and more. The architecture diagram looked like a subway map. And yet... nobody used it.
The truth is simple: you don't need AWS to build something users love.
The Overkill Problem
There is a common trap that builders fall into. You read a few blog posts or see a diagram on Twitter and suddenly you think your tiny project needs the same architecture as Netflix.
You don’t.
Most early-stage projects die not because they lacked scalability, but because they lacked users. Or because the product was confusing, buggy, or didn’t solve a real problem.
Overengineering your infrastructure is a great way to waste your time, burn out, or never launch at all.
What You Actually Need
If you are a solo developer or small team trying to build and ship something useful, this is probably all you need:
- A €5 to €20 per month VPS from Hetzner, DigitalOcean, or similar
- Docker Compose to run your app and a database
- A managed container platform like Sliplane if you want to avoid server management entirely (yes I am biased as the founder)
You do not need Kubernetes. You do not need to worry about auto-scaling. You do not need to connect six AWS services to display a single page.
Most indie products run perfectly fine on a single server for a long time.
When AWS Does Make Sense
Let's be fair. There are situations where AWS is the right choice:
- You want to learn AWS because you're job hunting or building cloud career skills
- You have very specific requirements, like needing data stored in a government cloud or being close to customer infrastructure that is also on AWS
- You are solving a problem that truly needs global scale from day one
- You're already deep in the AWS ecosystem and have a lot of expertise in it
These are good reasons. But be honest with yourself. Most projects don't start here.
And even if you do need AWS later, that's fine. You can always migrate when the time comes. At that point, you will hopefully have revenue, users, and a better understanding of your needs.
Remember this:
Your product is far more likely to fail because of what it does, not where it runs.
How to Get Started Without AWS
Want to launch something fast without spending weeks learning cloud architecture? Here’s a solid starting point:
- Use Docker Compose to define your app, database, and any background workers
- Deploy it to a VPS with ssh and docker compose up
- Or use a platform that abstracts the ops away and lets you focus on code
- Pick open-source tools for things like monitoring, auth, or task queues
That’s it. You can go from zero to deployed in an afternoon. No certification required.
Final Thoughts
You don’t need AWS to build something great. What you need is focus, a working product, and the ability to ship quickly. Big infrastructure will not save a bad product. Simple infrastructure will not kill a good one.
Start small. Launch early. Learn fast. You can always scale later.
Cheers,
Jonas, Co-Founder of Sliplane
PS: Just for clarification, I love AWS and I often even recommend it for specific use cases
Top comments (126)
Love the no-nonsense vibe here, honestly I'm always wasting way too much time setting up cloud stuff instead of actually building. you ever find yourself stuck on tech choices instead of just shipping?
Hahahaha all the dam time and then finding yourself back to your first choice.
Yep, ReactJS has morphed into an over-engineered framework for syncing what was once just a simple redundant text file. Back in the day, a lean PHP, Bash, and Python stack was more than enough for a solo developer—and it’s ironic considering many of today’s tech giants began as one-person shows.
What started as straightforward file loops now demands JSX, npm packages, and complex build pipelines, mirroring the very complexity it was meant to eliminate.
Meanwhile, AWS flaunts over 200 flagship services (117 with built-in encryption) and more than 1.3 million active certifications—each peddled as “mission-critical” for secure, scalable infrastructure.
That relentless marketing blitz sidelines nimble startups before they ever get a shot at challenging Facebook or Amazon; small teams simply can’t chase every rebrand, documentation overhaul, and new certification.
And if you doubt big-tech ethics, just watch how they leverage massive media muscle to shape narratives, pad profits, and lock in users—pure corporate prestidigitation.
Could any tech giant really go that far to protect its bottom line? No, of course not…
Hahahahahahaha you hit the nail on the head, I can read the pain in your words. Dam the suffering we endured. I can't call myself a full developer but recently while building projects. PROJECT 1 - LE ME 0 🤣😂
Hhuh
💯 When working on net-new product or capability, greatest risk is often market- or user-fit. Testing and rapid iteration is most important. Often we see the value most in the higher-order services. Need to transcribe real-time audio from websockets? Cloud service is often easier and more reliable to experiment with.
But trying to build an event driven architecture practice as a pre-req for your rapid experimentation? You'll be arguing over fat or thin schemas-- and even the philosophical meaning of what events are and will never get to actual innovation (the knowledge of what combination of features, UI, UX and target persona needs the improvement).
I remember setting SSO in AWS, where Cognito was involved. It was a mess. It was frustrating. Many bugs. Many parts didn’t even work. It was about 3 years ago. In general all aws works with very slow web admin, everything is complicated with their iam/ roles/permissions and they actually let you program their service instead of providing a clear, intuitive and simple admin and SDK.
Instead, today I look for new services from young and brilliant startups. These services are done by gurus so their “beta” is already stable by far from any AWS years of production. I’m not a security expert but comparing e buggy AWS to an innovative startup of gurus, I’m sure it’s the same about the service security.
Thousand armature AWS programmers can’t beat a single innovative startup guru.
What many of these so called "gurus" don't tell you is that they are using AWS under the hood and adding a premium on top of their shell, that you pay for.
Yes, since Ai joined the game.. one has to be really careful with how derailing all those extra npm, pips, aptgets, docker extravaganza can lead to.. it’s easy to go with a ”-y” knowing our ”vibe” tools can fix the bugs; and woha.. can be hard to cut the looses of a tech stack realizing down the line that such introduced complexities always bares future costs too; focus & ”archetypes” still rules.. easy to forget how big the incentives are for a new shiny tool echoing on social media, even via our dev friends; it’s what fuels the whole circus; affiliate funnels & more viewers for the short term; the more lost - the more valuable in a milking ”influencer” economy.
A few API calls put in the right places in a chmoded textfile can replace gigabytes of introduced risks from a day of Ai assisted repo explorations.
Sliplane offers no geographically appropriate options and I love it when you say "deploy anything that fits in a container" and then list a stack of JavaScript frameworks which could all easily be hosted on Vercel. Possible poorly chosen stacks the marketing bits.
I struggle to read posts from founders about how not to use X but do Y when there is a clear vested interest in bashing X.
I had the same opinion before I ever had vested interest :D
There are 2 sides to that. Yes, if I need a quick, simple instances, there are always cheaper provides than the big cloud services. Even secure enough. Like OVH, Hetzner, .. (even from Hetzner I had bad experience and the impression stays).
Big providers (AWS, Azure, GPC, IBM, ..) have different advantage (aside scaling). There are all the services matched together , from network, storage, backup, ... and most important - security is managed from a single source with easy delegation and fine-grained control. When you are a single dev, you may not care. Where there's a team/enterprise and when there are security concerns, that's a price for assurances.
Vendor-lock-in really is a rabbit-in-a-hat trick: non-technical execs are dazzled by “security” promises, while the real magic lies in simply moving data around within the same S3 bucket. Meanwhile, AWS keeps pulling new services out of its top hat—each rebrand, bells-and-whistles feature, and AI-powered dashboard demo the latest flourish in its buzzworthy show.
It’s fun to watch, but for freelancers and lean startups on a shoestring budget, the endless renames and tutorial tours torch runway faster than you can say “magician.” Suddenly you’re chasing hype over a text file lost in a labyrinth of permissions, all while juggling family life, mortgage stress, and looming deadlines.
Innovation should be a spotlight, not a smoke screen. Let’s keep applauding big-tech wizardry—but not at the expense of the little guys still learning how to pull their own rabbit out of the hat.
Solid post. The point about overengineering early is spot on. It’s common to see MVPs built with a full suite of AWS tools that never make it in front of real users. There’s definitely a temptation to mimic big-company stacks because it feels like “professional” work, but for most small teams or solo devs, it’s more of a distraction than a benefit.
AWS is powerful, no question - but complexity adds overhead. Most apps don’t need global scale or auto-everything on day one. They need clarity, feedback, and iteration.
If you’re just trying to build and learn, keeping it simple makes a lot more sense. A VPS and Docker gets you 90% of the way without the cognitive load. And if you grow past that, great - migrate later with more confidence.
Start with the simplest infrastructure that supports your current needs, not hypothetical ones., and revisit complexity only when there’s a real bottleneck or operational need.
thank you for breaking it down! i needed to read this
Cargo cult is a phenomenon when external actions or attributes are copied without understanding their internal essence and cause-and-effect relationships, in the hope of obtaining the same result as the object of imitation
Wow.. That's great
Thank you for one of the most logical posts I've read, I've been trying to explain this to people and friends for months or even years. The unnecessary complexity of AWS already makes things more difficult and its pricing is a bit of salt and pepper, on the other hand, as you mentioned, alternatives such as digitalocean and even koyeb not only make things easier, but also relieve the project and us with their fair pricing.
I always tend to go to towards Firebase first. It's dead simple and deploys what I need in seconds to build my MVP. It's also tied into GCP so should I really need something more complex than a database and some cloud functions it's there for me to use.
I totally agree with the author. In many cases AWS is overkill, especially for smaller projects. Their billing is tricky too. If you are not careful, you can get a large bill, even if you use their "free" services.
To be fair, never had that issue with AWS. The billing is tricky, but luckily never got a surprise bill!
Realistically, you don't even need docker. If you already have a template project with it cool; otherwise deploying an app directly on a server with an SQLite database somewhere in /var/www is good enough for your first several thousand users.
There's a reason why so many successful apps run on rails, and it's 0% about how good rails is for large apps and 100% abobut how quickly you can build and deploy something that works.
I’ve spent over 25 years developing all kinds of projects and products, and I completely agree with this article. Personally, I’ve been living and breathing VPS environments since 2008—and I wouldn’t trade it for anything.
Over the years, with small tweaks and smart optimizations, I’ve made solid projects run like Swiss watches—without ever needing to migrate to expensive corporate services or overengineer the infrastructure.
More than once, I’ve watched competitors collapse under the financial weight of platforms like AWS or GCP, all while their projects hadn’t even validated a real business model. All that just to look bigger than they really were.
The key is to start simple, scale wisely… not with ego.
Do you recommend to your customers a VPS hosted solution and then provide general admin support services ? I would have thought most companies are hesitant to go with a simple VPS based solution? It too basic or limitations.
Do you think the VPS is a strong enterprise solution? Would love to hear more specifics about your use cases. Would you recommend to host custom LLMs via a VPS for small companies to integrate with their in-house apps and databases (for example)?
Absolutely. I work exclusively with long-term clients — some have been with me for over 14 years, and the “newest” has been with me for 5.
My standard approach is to develop the client’s entire project on my own VPS, managed with WHM/cPanel, and assign a dedicated account under the domain we’ll be working with. Once the client approves the project and is happy with the results, I offer them the option to either stay hosted on my VPS while they gain traction, or to migrate everything to their own private VPS.
In my experience, a properly configured VPS is not only more than enough — it’s a strategic choice. It gives me full control, security, and performance without unnecessary overengineering. It also allows for lean infrastructure and full transparency with the client.
Would I recommend using VPS to host custom LLMs for small companies integrating with their internal tools and databases?
Definitely — with proper resource allocation, smart architecture and strong security practices, it’s a flexible and scalable solution.
When someone trusts you with the performance and stability of their business, that’s a responsibility I don’t take lightly. That’s why my clients don’t let me go, and I never take my eyes off their systems.
That’s my take on that 🙂
Thank you @auvitijuana, very helpful and reassuring to know the VPS can be a solid enterprise ready solution 🏗️
Some comments may only be visible to logged-in visitors. Sign in to view all comments.