Quick Answer: Hetzner's Singapore location carries a per-instance surcharge and far less included traffic than its EU sites, so APAC savings come mainly from egress — tens of euros per terabyte versus roughly $0.09/GB on AWS Singapore. Below ~2TB monthly egress, operational overhead usually erases the compute saving.

Hetzner Cloud Pricing APAC Infrastructure Cost: What the 2026 Numbers Actually Mean

A Hong Kong-based DTC brand came to us last year with a familiar spreadsheet: roughly US$4,000 a month on AWS for what amounted to three application servers, a managed Postgres instance, a Redis cache, and about 6TB of monthly egress serving product imagery to customers in Hong Kong, Singapore, and Australia. The engineering lead had priced the same workload on Hetzner and come out around US$300. That 13x gap is why Hetzner cloud pricing and APAC infrastructure cost keep showing up in the same conversation — and why the June 2026 price adjustment caused so much noise on Reddit and Hacker News.

Related reading: Local LLM Models vs Cloud APIs for Coding: APAC Verdict

Related reading: LinkedIn Security Backdoor Hiring Risk: APAC Vetting Guide

Related reading: The Agentic E-Commerce Checkout Protocol: An APAC Operator's View

The honest answer is that the gap is real, the arbitrage is narrower than the spreadsheet suggests, and the deciding variable for most Asia-Pacific teams is not the hourly instance rate at all. It is egress, latency geography, and how many hours a month your team can afford to spend being its own managed-service provider. Any serious Hetzner cloud pricing APAC infrastructure cost comparison has to start there.

The Singapore Location Changed the Calculation for APAC Teams

Until Hetzner opened a Singapore cloud location, the Hetzner-versus-hyperscaler debate was mostly academic for anyone serving Asian end users. Falkenstein and Helsinki are excellent value and roughly 180–200ms away from Hong Kong — fine for CI runners, batch jobs, and internal tooling, unusable for a checkout flow.

Singapore puts Hetzner inside the region. From Singapore you are typically looking at 30–40ms to Hong Kong, 60–80ms to Tokyo, 70–90ms to Sydney, and 20–30ms to Jakarta and Kuala Lumpur, depending on peering. That is a credible primary region for Southeast Asia, acceptable for Greater China via Hong Kong, and marginal for Australia and Japan if you have a latency-sensitive interactive product.

Two things APAC teams get wrong here. First, Hetzner's non-EU locations are priced differently from the German and Finnish ones — Hetzner publishes a location surcharge per instance per month for Singapore and its US sites, so the headline €3.79 or €5.49 figures circulating in pricing calculators are EU numbers, not Singapore numbers. Check the live figure on Hetzner's cloud pricing page before you model anything; the surcharge is where the APAC premium lives. Second, included traffic differs sharply by location: Hetzner's EU instances ship with 20TB of included egress, while the Singapore and US locations include substantially less (1TB at time of writing), with overage billed per terabyte. For a media-heavy commerce site, that single line changes the whole model.

Why the June 2026 Price Adjustment Hit Harder Than the Percentage Suggests

Hetzner's 15 June 2026 adjustment applied to new orders and to rescales, not retroactively to existing instances — which is exactly why the reaction split into two camps. Teams with stable, long-lived servers barely noticed. Teams running autoscaled or frequently-rebuilt fleets re-provisioned into the new rates almost immediately.

One widely-shared thread had an operator describing a 15-server fleet moving from roughly €700/month toward €2,000/month under the new pricing. Northflank's write-up on the 2026 increases framed the same shift from a per-second-billing perspective, noting that pay-as-you-go compute at CPU $0.01667/vCPU/hour and memory $0.00833/GB/hour becomes competitive once Hetzner's flat monthly rate rises. Both are worth reading, but neither is the full picture for an APAC team — because neither accounts for the location surcharge or the reduced included traffic outside the EU.

The structural driver is not Hetzner being greedy. According to Gartner's public cloud forecasts, end-user spending has been growing above 20% annually, and Synergy Research Group's quarterly data consistently shows AWS, Microsoft, and Google holding roughly two-thirds of cloud infrastructure services revenue. Those three are collectively committing hundreds of billions in annual capex to AI capacity, which bids up GPUs, DRAM, power contracts, and data-centre shells. A provider operating on thin hardware margins with no AI revenue to cross-subsidise has two options: raise prices or degrade the fleet. Hetzner raised prices. That is the less dishonest option.

Related reading: AI Inference Semiconductor Demand in APAC: What the Data Shows

What the Real Cost Delta Looks Like at Working Scale

Strip out the marketing numbers and model a concrete workload. Take a mid-sized APAC commerce stack: three application nodes at roughly 4 vCPU / 8GB, one database node at 8 vCPU / 32GB with replication, a load balancer, 500GB of block storage, and 6TB of monthly egress from Singapore.

Compute

On Hetzner Singapore, that compute footprint — CX33/CPX41-class instances plus a CX23-class utility box — lands in the low hundreds of US dollars per month including the location surcharge. On AWS Singapore (ap-southeast-1) with on-demand m7g or m6i instances plus RDS Multi-AZ, the equivalent is comfortably four to six times that before you reserve anything. Committing to a three-year Savings Plan closes maybe 40–50% of the compute gap, per AWS's own published discount ranges — but commits you to capacity you may not want in year three.

Related reading: Shopify Plus Noibu Ecommerce Analytics in APAC: What It Fixes

Egress — the line that actually decides it

According to AWS's published pricing, data transfer out to the internet from Singapore runs around $0.09/GB for the first tier. Six terabytes is therefore roughly $540/month in egress alone. Hetzner bills overage per terabyte in the low single-digit euros. You are comparing hundreds of dollars against tens. For any business serving images, video, catalogue assets, or app bundles across Asia, egress is frequently the single largest controllable infrastructure line — and it is where the Hetzner cloud pricing APAC infrastructure cost argument is strongest by an order of magnitude.

The invisible line: who runs the database

The AWS vs Hetzner comparisons circulating this year make one fair point repeatedly: self-managed Postgres costs 10–20 engineering hours a month in patching, backup verification, failover testing, and monitoring. According to Northflank's 2026 report on infrastructure cost shifts, this operational overhead is routinely underweighted in migration decisions. At Hong Kong or Singapore senior-engineer rates, 15 hours is not a rounding error — it can exceed the entire monthly saving on a small fleet. The arbitrage only works if you either already have that competence in-house or you consolidate it across enough workloads to amortise it.

Build the Comparison Properly Before You Migrate

Most "Hetzner cloud pricing APAC infrastructure cost calculator" searches end at instance rates. A usable model needs five inputs: instance cost including location surcharge, egress above the included allowance, block and object storage, snapshot/backup storage, and the loaded cost of operational hours.

If you want a fast reality check, provision one instance per candidate region and measure before you model:

1# Install and authenticate the Hetzner CLI
2brew install hcloud
3hcloud context create apac-benchmark
4
5# List what's actually available in Singapore with live pricing
6hcloud server-type list -o columns=name,cores,memory,disk
7hcloud location list
8
9# Spin up a Singapore node for latency testing
10hcloud server create \
11  --name sg-latency-probe \
12  --type cpx31 \
13  --image ubuntu-24.04 \
14  --location sin \
15  --ssh-key my-key

Then measure from where your customers actually are, not from your office:

1# From a probe in Hong Kong, Sydney, and Tokyo
2mtr -r -c 100 <singapore-ip>
3curl -o /dev/null -s -w "ttfb: %{time_starttransfer}s total: %{time_total}s\n" \
4  https://your-sg-endpoint/health

For anything you intend to keep, define it as code from day one so the migration back out stays cheap:

1terraform {
2  required_providers {
3    hcloud = {
4      source  = "hetznercloud/hcloud"
5      version = "~> 1.49"
6    }
7  }
8}
9
10resource "hcloud_server" "app" {
11  count       = 3
12  name        = "app-sg-${count.index}"
13  server_type = "cpx41"
14  image       = "ubuntu-24.04"
15  location    = "sin"
16  public_net {
17    ipv4_enabled = true
18    ipv6_enabled = true
19  }
20  labels = { env = "prod", region = "apac" }
21}

The discipline matters more than the provider. We have run managed infrastructure engagements where the team's genuine blocker to leaving a hyperscaler was not cost or latency — it was that nothing was reproducible. Click-configured VPCs, hand-tuned security groups, and an undocumented IAM graph are a far stronger lock-in than any pricing page.

Why Are Hetzner Servers So Cheap in the First Place?

Three reasons, none of them magic. Hetzner builds and operates its own data centres in Germany and Finland, where it has direct access to relatively cheap grid power and its own hardware assembly. It sells a deliberately narrow product catalogue — compute, block storage, object storage, load balancers, firewalls, private networks — rather than the 200-plus managed services a hyperscaler maintains. And it does not fund a global enterprise sales organisation, solutions-architect bench, or compliance programme covering every regulated vertical on earth.

That last point is the trade-off, stated plainly. You are not buying AWS Artifact, a named technical account manager, GovCloud-equivalent isolation, or a 24/7 enterprise support SLA with a response-time guarantee. For an APAC startup pre-Series B, none of that matters. For a licensed financial institution in Hong Kong or Singapore under HKMA or MAS outsourcing expectations, several of those items are not optional — and the procurement conversation ends before the pricing conversation starts.

What Is Actually Cheaper Than Hetzner in Asia?

On pure per-vCPU price in the region, a handful of options genuinely undercut Hetzner's Singapore rates: Alibaba Cloud's entry tiers during promotional periods, Tencent Cloud in markets where it is pushing share, and smaller regional VPS providers in Singapore, Japan, and Indonesia. OVHcloud's Singapore presence sits in a similar price band. Oracle Cloud's Always Free Arm allocation is the only meaningful free tier of any size in the region, and it is capacity-constrained in popular availability domains.

But cheaper per vCPU is not the same as cheaper per deployed application. Two things typically erase the saving: egress pricing that reverts to hyperscaler-like per-gigabyte rates, and API/tooling maturity. Hetzner's CLI, Terraform provider, and cloud-controller-manager for Kubernetes are well-maintained and well-documented. On several lower-cost regional providers, you will write and maintain that glue yourself.

The more interesting pattern we see across Asia-Pacific teams is not single-provider migration but deliberate splitting: latency-sensitive edge and stateful regulated data on a hyperscaler in-region, and everything bandwidth-heavy or batch-oriented — CI/CD runners, image processing, data pipelines, staging environments, internal analytics — on Hetzner. A Greater China retail client of ours running a dealer-network portal landed in roughly that shape, keeping the transactional core on a hyperscaler for audit reasons while moving build infrastructure and asset processing off it entirely. The mechanism is simple: you move the workloads where egress and raw CPU dominate, and leave the ones where compliance and managed services dominate.

Using Asia as the Operations Hub, Not Just the Market

For US and European companies, the Singapore location is more interesting as a hub than as a market endpoint. Singapore has credible subsea connectivity to India, Japan, Australia, and Southeast Asia; a stable regulatory environment; and the region's deepest pool of multilingual DevOps talent. A European company serving APAC customers can now run its Asian footprint on the same provider, same Terraform modules, and same billing relationship as its Frankfurt infrastructure — with one location attribute changed.

The constraint is Singapore's power and land cost, which is precisely why the location surcharge exists and why it is unlikely to disappear. According to the Infocomm Media Development Authority, Singapore's data-centre moratorium and subsequent capacity-allocation framework mean regional compute supply is constrained by policy as well as economics. Expect APAC infrastructure to stay structurally more expensive than Falkenstein indefinitely, regardless of provider.

Where This Leaves Your Budget

My practical rule after modelling Hetzner cloud pricing against APAC infrastructure cost for commerce, catering, and dealer-network clients across the region: if egress exceeds about 2TB a month, or your compute bill exceeds roughly US$1,500 a month and your workload is stateless, the comparison is worth a hard look even after the 2026 adjustments. Below that, the operational overhead eats the delta and you should spend the engineering time on your product instead.

And this advice is not for everyone. If you are a regulated financial institution, a healthcare provider handling patient data, a team with no one who can confidently restore a Postgres cluster from a WAL archive at 3am, or a company whose enterprise buyers demand specific certifications in the security questionnaire — stay where you are and negotiate a committed-use discount instead. The cheapest infrastructure is the one your team can actually operate. Price adjustments like June 2026 will keep coming from every provider as AI capex reshapes the hardware market; the defensible position is not picking the cheapest vendor today, but keeping your stack portable enough that next year's announcement is a Terraform variable rather than a crisis.

If you are modelling an APAC infrastructure move and want a second set of eyes on the egress and operational-hours assumptions, talk to Branch8 — we build and run this kind of cross-border infrastructure across Hong Kong, Singapore, Taiwan, and Australia.

Sources

• Hetzner — Cloud Hosting Pricing
• Hetzner — Documentation Portal
• Hetzner Cloud — Terraform Provider Documentation
• AWS — EC2 On-Demand Pricing
• Microsoft Azure — Pricing Calculator
• Gartner — Newsroom and IT Spending Forecasts
• Synergy Research Group — Cloud Infrastructure Market Data
• IMDA Singapore — Digital Infrastructure and Data Centres