I host infrastructure for a living, so the Hetzner price increase landed on my desk the same week it landed on everyone else’s. And my first reaction probably wasn’t what you’d expect. I wasn’t angry about the number. I get why it happened. The part that bugs me is how it was handled.
So let me walk through what actually changed, why it makes complete sense from a business angle, and the one thing I think Hetzner got wrong.
What actually changed in the Hetzner price increase
This wasn’t one increase. It was two, a few months apart.
The first came on 23 February 2026, when Hetzner announced a price adjustment effective 1 April. That round touched both existing products and new orders. Tom’s Hardware pegged it at up to 37%.
Then on 27 May, Hetzner announced a second round effective 15 June. This one bundled a price hike with a portfolio cleanup: dedicated servers now ship as fixed configurations named “-1”, “-2”, and “-3” instead of fully custom RAM and storage builds. German outlet heise reported the cloud side as up to 200%, a partial tripling.
The published numbers back that up. Hetzner’s own price list shows the dedicated-vCPU CCX line roughly tripling:
| Cloud plan | Old €/mo | New €/mo | Change |
|---|---|---|---|
| CCX13 | 15.99 | 42.99 | +169% |
| CCX23 | 31.49 | 85.99 | +173% |
| CPX22 | 7.99 | 19.49 | +144% |
| CAX11 (Arm) | 4.49 | 5.99 | +33% |
| CX23 (shared) | 3.99 | 5.49 | +38% |
Prices exclude VAT, taken from Hetzner’s official 15 June 2026 price-adjustment page. The dedicated-CPU plans took the hardest hit; Arm and shared plans rose far less.
A few things worth knowing before you panic:
- Existing dedicated servers are safe. The 15 June change applies to new orders and rescales only. Keep your box as-is and your terms don’t move.
- Not everything went up. Web hosting, managed servers, Server Auction machines, IPs, storage products, Load Balancers, Volumes, Snapshots, and Object Storage were all left out of the June round.
- Setup fees actually dropped for most dedicated servers in the same move. So it’s a higher monthly rate against a lower one-time cost, not a straight grab.
That last point matters. A provider purely chasing margin doesn’t cut your setup fee on the way through.
Why the Hetzner price increase makes business sense
Here’s the thing people skip past: the hardware market genuinely went crazy.
Server memory is the headline. TrendForce reported that conventional DRAM contract prices jumped roughly 93% to 98% in the first quarter of 2026 alone, pushing total DRAM industry revenue up 81% in a single quarter to $97 billion. And it wasn’t a one-off spike. TrendForce expected another 58% to 63% rise the following quarter.
Storage tells the same story. Tom’s Hardware reported that SSDs now cost up to 16 times more per terabyte than hard drives, as the AI supply chain swallows NAND output. The cause is the same on both fronts: AI data centers are buying high-bandwidth and high-capacity memory faster than the fabs can make it, and the major suppliers have shifted production toward those fat-margin server parts. Everyone else queues behind them.
Hetzner buys this stuff by the rack. When the input doubles, the math stops working.
What I respect is that they didn’t pass it on immediately. In the Hetzner community forum, the company’s Martin Hetzner reportedly explained that Hetzner had spent roughly nine months subsidizing new server hardware and holding off on the cost handoff to customers. That tracks with the timeline. They tried to eat it, then couldn’t anymore. Their own FAQ puts it plainly: more efficient provisioning “can only mitigate” the impact, it can’t undo a procurement market that doubled.
They’re also not alone. OVHcloud’s CEO had already signaled back in November 2025 that cloud costs would climb 5% to 10% in 2026, for the same reason. This is a market-wide shift, not a Hetzner quirk.
Where Hetzner went wrong: the communication
So if the number is fair, what’s my problem?
The communication with the public. That’s the one thing I’d push Hetzner to fix.
Start with that first statement in February. It opened with “there have been drastic price increases in various areas in the IT sector” and “the costs to operate our infrastructure and to buy new hardware have both increased dramatically.” All true. But there wasn’t a single number in it. No table, no example, no “here’s what your CCX13 becomes.” Customers were told a big change was coming and left to find the actual figures somewhere else. One well-known German tech blogger admitted he had no feel for how the prices had actually moved.
Then came the rhythm. The April increase had barely settled before the June one was announced, roughly two and a half months later. Two hikes in one spring, each its own statement, is a hard thing to plan around. Even a customer who fully accepts the reasoning starts to wonder when the next one lands.
And the messiest part was availability. Around the 15 June switch, people couldn’t reliably get new cloud instances or rescale existing ones. One customer described whole “Cost Optimized” models greyed out in Nuremberg and Falkenstein, with only Helsinki bookable. Another called it a gamble whether a rescale would go through, and said he waited days to resize his own machines. Where was that explained? Buried in a status page incident most people only found because another customer linked it in a comment thread. The clearest explanation of the whole pricing strategy came from Martin Hetzner himself, in a forum post sitting behind a login.
None of that is malicious. I think it’s a great engineering company treating a pricing event like a back-office update instead of a conversation with the people paying the bills.
Is the Hetzner price increase a reason to leave?
For most people, honestly, no.
Even after two rounds, Hetzner is still one of the strongest price-to-performance options in Europe. A CCX13 at €42.99 is a real jump from €15.99, but stack it against the US hyperscalers for the same dedicated vCPUs and RAM and it’s still a sane bill. Leaving costs money too: migration time, re-testing, and often higher prices wherever you land, because the memory squeeze hit every provider.
What I’d actually do is treat this as a prompt to look at your own setup:
- Right-size before you rescale. The increase bites hardest on rescales and new orders. Audit what each machine really needs before you bump a tier out of habit.
- Check the limited and auction tiers. Hetzner’s new “-1-Ltd” limited servers and the untouched Server Auction can land you cheaper hardware if you’re flexible on exact specs.
- Don’t get locked to one provider. This is the real lesson. If your stack is portable, a price change is an inconvenience. If it’s welded to one vendor, it’s a hostage situation.
That last one is how we build for clients. Infrastructure you control, on standards that move, so a pricing email from any provider is something you decide on, not something that decides for you.
Closing the loop
The Hetzner price increase is, in the end, a fair response to a hardware market that lost its mind. RAM and storage costs genuinely surged, Hetzner absorbed it for a while, and then did what any provider eventually has to do. I won’t fault them for the number.
I’ll fault them for the rollout. A clear email, real numbers up front, and an honest heads-up about the availability crunch would have turned a rough quarter into a non-event. The price was the easy part to defend. The silence around it wasn’t.
