The Cloud Tax

Happy Monday,
Short and sweet post today (and late…), as I’ve spent most of my spare time working on my new website! It lives entirely within the free tier of AWS, and yet I'm quite happy with it. Let me know what you think of the blog's new home: here.

Hosting a website on AWS has a lot of steps, especially for those of us that decide to buy a domain name from a not-AWS provider. I've worked hard keeping everything cheap, being careful and setting billing limits, but I can absolutely see a world where I do need to break out of the free tier soon. For example, I’ve been working on getting FiNet (my full stack project) in a state that can run 24/7 on AWS without costing too much. On top of this, I plan on hosting a handful of APIs and smaller projects as well very soon, just for some practice. No doubt I’ll hit a point that AWS will go from sandbox to full on construction zone and as such I’ve been thinking about possible alternatives before I really break ground.

After looking into it I’ve decided that dedicated server renting makes more logical sense for project scopes in this zone. With as little as $42 USD / Month ( Hetzner Link ) you can get a fully capable server with 32+GB Ram and modern CPU/GPU hardware. Once you start having AWS products run 24/7 the cost prediction calculator skyrockets, for the same specs an EC2 instance might cost nearly $500 / Month.

This is assuming your project isn’t ready to commit to 1+ Years of reserved compute power. Obviously these projects might not actually be up all 730 Hours but even considering these variables the cost difference is apparent.

There is a huge flexibility difference, which shouldn’t be underestimated. But there are definitely use-cases where AWS is significantly more expensive than simply hosting applications on some dedicated German servers.

I also briefly looked into Owning hardware as an alternative to this abstracted virtual ownership situation, just to consider all options. A comparable mini workstation pc can be purchased for as little as $680 on Amazon, which based on our AWS cost estimating would pay for itself in less than 2 months (Hetzner in ~16 months). Of course my network router won’t even let me host publicly accessible websites without an addition fee and I experience nearly monthly internet outages. Needless to say this absolutely isn’t an option right now.

I’d be interested in doing these kinds of cost breakdowns for a small-business scale application, I’d imagine the period before seeing returns on the investment is likely shorter than it might seem when looking at all the upfront costs, mostly due to how ridiculously pricey AWS is.

Owning hardware and renting dedicated hardware both have the same issue, upfront costs with no flexibility in usage. Cost cutting on AWS is much more straightforward, minimize compute time and compute resources as much as possible. In situations like Shopify, this makes a lot of sense and is worth the effort. However, in my case where demand varies very little (0-3 Users), it doesn’t seem likely to be worth my time or the luxury tax that comes with hardware scaling.

The question is where does that line lie and how much value is there in planning ahead for when demand crosses that line. I constantly see startup employees on Reddit tasked with designing complex microservices for products that lack market results; to me, this looks less like preparation and more like setting the company up for failure.

I’m interested in the option for small businesses to offer compute capabilities during down time, maybe a company that facilitates the process of renting and hosting (AirBnB for compute). It’d give companies another little push in the direction of hardware ownership, which seems undervalued in today’s market. There exists DePIN networks but these are typically are much more specialized than what AWS and Hetzner offer. DePIN broadly is something I definitely want to look into more, and definitely isn’t the topic of this blog post.

---