Hetzner Prices increase 30-40%
In this technical retrospective (2026), @williausrohr shares practical architectural lessons learned from running, scaling, or migrating production software under lean constraints.
Core Principles for Independent Builders:
Defaulting to simpler architectures (such as monolithic frameworks, SQLite, or managed VPS) prevents early engineering bankruptcy.
Evaluating true compute and egress costs reveals that high cloud bills can often be slashed by 80%+ through judicious re-platforming.
Code you write at 2 AM as a solo engineer must be easy to debug six months later without complex microservice dependencies.
Community Debate Themes:
HN engineers debated the tradeoffs between cloud convenience and raw server economics, database indexing strategies, and resisting premature architectural complexity.
CATEGORY Frugal Engineering →
HISTORICAL ERA Post-ZIRP & AI (2023–2026)
COMMUNITY SCORE 4 points
TOTAL DISCUSSIONS 0 comments
I just started the process of migrating to them yesterday. They are still very affordable. But a bit less. I'm estimating that our quite lean GCP setup cost is going to be cut to about 20-25% when I'm done. So, it doesn't affect my decision to go with them literally yesterday morning.
It's all a bit barebones and primitive but I don't mind. I spent yesterday tweaking some ansible scripts with codex to setup stuff like bastion hosts and nat networking. I expect I have most of the rest ready in a few days.
The benefits of having an uncomplicated docker compose and boring tech stack. No microservices. Just a monolith.
One issue that I don't have a solution for yet is disk encryption and encrypted bucket content. Probably solvable but not natively supported. Might trigger compliance issues with some of our customers.
Something I don't quite understand is why the price of a used dedicated server is impacted by anything beyond electricity and land prices. It's not like the RAM gets replaced by new one every now and then. That's old RAM chip that's been bought years ago (and I believe largely amortized by now).
Is there any collocation space from which you can actually buy and own a physical machine without going there, and pay only the rent and traffic? I miss calling a server "mine".
the 500% DRAM cost jump makes more sense when you factor in that Samsung and SK Hynix have been shifting capacity from commodity DRAM to HBM for AI accelerators. same fabs, different product mix. so consumer/server DRAM supply contracts right when everyone is spinning up new cloud infra to run inference workloads. supply squeeze from both sides.
From the email:
That is an utterly insane price hike. Is production being scaled up, and will that take years? Or are producers happy charging 5x the price for the same amount of effort?
How much of this is driven by speculation vs actual demand?
I have been buying older servers by the truckloads. Older being a year or so. It will be enough to host whatever outside AI that we need for the coming 15-20 years. And the all were great deals, will have them paid for within a month per server. I have my own cage full with empty racks bought from a bankrupt company in AMS.