Hetzner Raises Prices: The Cheap-Cloud Benchmark Just Moved

What happened

On June 15, 2026, Hetzner published a price adjustment notice covering its cloud server lineup. The notice (docs.hetzner.com/general/infrastructure-and-availability/price-adjustment) lays out new monthly and hourly pricing for the CX, CPX, CCX, and CAX families, applied to both new orders and — after a grace window — existing instances. The thread hit 398 points on Hacker News within hours, which is the kind of velocity Hetzner news only gets when the number on the invoice actually moves.

The context matters. Hetzner has spent the better part of a decade being the price the rest of the industry is measured against — the benchmark that lets a founder say, with a straight face, "AWS is 8× more expensive than it needs to be." CX-series VMs at a few euros a month, dedicated AX-line machines with 64GB of ECC RAM for the cost of an AWS dinner reservation, and an unmetered 20TB egress allowance that makes Cloudflare R2's existence feel like a polite hint. When that floor moves, every cost model anchored to it moves with it.

The adjustment itself is not catastrophic in isolation — Hetzner remains, by a wide margin, the cheapest production-grade option in Europe. But "Hetzner raised prices" is a sentence the industry has rarely had to say, and the second-order effects matter more than the first-order euros.

Why it matters

First, the macro signal. Hetzner doesn't raise prices for fun. Their entire brand is built on grinding hardware, power, and bandwidth costs down a long curve. When a price-leader raises prices, it's almost always because input costs moved underneath them and they ran out of slack to absorb it. The candidates are well-known: EU industrial electricity has stayed structurally elevated since the 2022 shock and never fully reverted; AMD EPYC and Ampere ARM supply is being eaten alive by AI-buildout allocations; and NVMe pricing finally bottomed and reversed in late 2025 as hyperscalers pre-bought the next two years of capacity. Hetzner's announcement is, in effect, a public readout of the commodity-compute supply curve. Anyone running their own colo math should treat it as a calibration data point.

Second, the arbitrage compression. A non-trivial fraction of the "escape AWS" content of the last three years has rested on a specific ratio: Hetzner is roughly an order of magnitude cheaper than AWS for the same compute. Even a 10–15% adjustment at Hetzner doesn't kill that arbitrage, but it does narrow the gap precisely as AWS pushes harder on Graviton4 discounts and Savings Plans extensions. The break-even point — the workload size above which it's worth running your own infra on Hetzner rather than paying the AWS premium for managed everything — just shifted. Not by a lot. But "not by a lot" is the difference between yes and no on plenty of Series A architecture decisions.

Third, the community reaction tells you what's actually at stake. The HN thread is unusually free of the standard "vote with your feet" energy. The top comments aren't "I'm moving to OVH" — they're variations of "who else is even in this tier?" The honest answer is: very few. OVH has its own pricing pressure and a slightly more chaotic operational reputation. Scaleway's ARM offerings are interesting but limited. Contabo is cheaper on paper and dramatically worse in practice. Oracle's free tier was the dark-horse alternative until they quietly tightened the eligibility screws. The Hetzner-shaped hole in the market is real, and the lack of a credible second source is part of why a price adjustment from a single German hoster registers as industry news.

Fourth, the self-hoster knock-on. The bare-metal renaissance — the k3s-on-Hetzner-AX, Tailscale-mesh, Caddy-and-Postgres stack that's become the default for technical founders who don't want a cloud bill — was always a one-vendor bet. The community didn't really notice because the vendor never gave them a reason to. This is the first time in years that bet has wobbled, even slightly. Expect a renewed round of "compare three providers for the same workload" blog posts in the next six weeks.

What this means for your stack

If you're running production on Hetzner, the right immediate action is unglamorous: pull your monthly bill, apply the new rates by SKU, and decide whether the delta crosses any of your internal thresholds. For most teams, it won't. For teams running dozens of CX-series VMs as a sharded data plane, it might. The bigger move is to use this as a forcing function to do the cost-modeling exercise you've been putting off — annual reserved commitments, larger fewer instances vs. smaller more instances, and whether the AX-line dedicated boxes are now meaningfully cheaper per core than the cloud SKUs you defaulted to.

If you're on AWS/GCP/Azure and have been treating Hetzner as the always-available escape hatch, this is the moment to actually price the move rather than assume the spread. The spread is still there. It's just smaller than it was last quarter, and the gap between "theoretically cheaper" and "cheaper after engineering time" has always been the thing that kept teams on hyperscalers in the first place. A narrower headline gap means the engineering-time math gets stricter, not looser.

If you're a vendor selling to developers on a per-VM or per-CPU basis — observability, CI runners, edge functions, anything where your COGS is someone else's compute — your gross margin just changed and you probably haven't modeled it yet. The deployments running on Hetzner-style infrastructure (and there are a lot of them, often invisibly) just got marginally more expensive to operate. If you sell flat-rate or freemium, this hits the freemium tier first.

Looking ahead

The interesting question isn't whether Hetzner's adjustment sticks — it will, and within a quarter nobody will remember the old prices. The interesting question is whether this is a one-time recalibration or the first of several. If EU power and AI-hardware allocation keep compressing the supply of cheap general-purpose compute, the cloud-cost arbitrage that's powered a generation of bootstrapped infrastructure will quietly close, not with a hyperscaler price cut but with the floor rising to meet them. Watch OVH and Scaleway in the next 90 days. If they move too, the era of "compute is basically free if you know where to look" has a defensible end date.