Ask ten developers where to rent a cheap Linux box in 2026 and you’ll get ten different answers, and most of them are outdated. DigitalOcean’s brand recognition, Hetzner’s cult following among self-hosters, and AWS Lightsail’s “it’s just Amazon” convenience all pull in different directions. Meanwhile Vultr and Linode (now folded into Akamai’s cloud division) keep quietly undercutting everyone on paper while adding fine print that changes the real math.
This comparison pulls current September 2026 pricing straight from each provider’s own pricing page, cross-checks it against three independent benchmark trackers, and works out what a 2 vCPU / 4 GB box actually costs once bandwidth overages and hidden fees are added back in. The short version: Hetzner wins on raw price-per-core, DigitalOcean wins on developer experience, and AWS Lightsail wins if you’re already inside the AWS ecosystem and don’t want a second vendor relationship. The long version explains why that answer changes depending on your workload.
Why cloud VPS pricing got confusing again in 2026
Budget VPS pricing used to be stable for years at a stretch. That changed in August 2026, when Hetzner raised prices on its shared x86 CX line by roughly 30-50%. The CX23 plan (2 vCPU, 4 GB RAM) went from €3.99 to €5.99 a month, about $4.30 to $7.09 at September exchange rates, according to a pricing tracker at HostMath’s Hetzner price-increase breakdown. Hetzner folded a previously separate IPv4 charge into the new price, so the effective jump is closer to 30% once you account for what used to be an add-on, but it’s still the biggest single move any of the five providers in this comparison has made all year.
That price hike matters because Hetzner’s whole pitch has been “the same specs for less money.” A 30% increase doesn’t erase that advantage, but it narrows the gap with Vultr and Linode’s shared-CPU tiers enough that the old advice, just use Hetzner, it’s cheaper, no longer holds automatically. Specs stayed the same through the change: same vCPU count, same RAM, same SSD, and Hetzner still bundles a genuinely generous 20 TB of monthly traffic on European nodes, per Hetzner’s own documentation.
Meanwhile AWS Lightsail, DigitalOcean, Vultr, and Linode’s headline monthly prices have held steady, but the memory-chip shortage rattling the wider hardware market this year has pushed several providers toward smaller default storage allotments and paid add-ons for extra NVMe capacity. None of the five main providers here have publicly announced a DRAM-driven price hike as of September 22, 2026, but it’s worth watching, since RAM has gotten measurably more expensive across the hardware industry this year.
The five providers compared: DigitalOcean, AWS Lightsail, Vultr, Linode/Akamai, and Hetzner
DigitalOcean built its reputation on predictable pricing and documentation good enough that beginners rarely need a support ticket. Droplets are the compute product, billed hourly with a monthly cap, and DigitalOcean throws in a managed dashboard plus one-click apps for Postgres, Redis, and WordPress that the budget competitors mostly skip.
AWS Lightsail is Amazon’s answer to “EC2 is too complicated for a $5 server.” It bundles compute, a static IP, and a chunk of bandwidth into one flat monthly bill, and it plugs directly into the rest of AWS if you outgrow it. The catch is that Lightsail’s small print includes multiple bundle families (Linux/Unix general purpose plus separate WordPress and app-specific bundles) that aren’t always apples-to-apples with a plain VPS.
Vultr runs 32+ data center locations and sells itself on that geographic spread plus hourly billing down to the second. Its Cloud Compute Regular Performance line uses shared vCPUs on modern AMD EPYC hardware.
Linode, now operating as Akamai Cloud Compute after Akamai’s 2022 acquisition, kept its old plan names (Nanode, Linode 2GB, and so on) even as the backend infrastructure moved onto Akamai’s network. It’s the steadiest of the five: prices haven’t moved much in years, and Akamai’s global CDN footprint is a real edge for latency-sensitive apps.
Hetzner is the German provider that self-hosters treat like a cheat code: dedicated NVMe, big bandwidth allowances, and prices that undercut the U.S. hyperscalers by half or more, even after the August increase. The tradeoff is a smaller region list (Germany, Finland, the U.S., and Singapore) and a genuinely bare-bones control panel, no managed database add-ons, no one-click WordPress installer.
How this comparison was built
Every price in this article was pulled directly from a provider’s own pricing page where that page exposes static numbers, which is the case for DigitalOcean’s Droplet pricing, AWS Lightsail’s bundle pricing, and Hetzner’s Cloud pricing page. Vultr and Linode/Akamai both render pricing through JavaScript rather than static HTML, so those two providers’ figures come from live pricing trackers (Pikkly and HostMath) that scrape and timestamp the official pages daily, cross-checked against each other for agreement before inclusion here. Anywhere the two source types disagreed by more than a few cents, the more conservative (higher) figure was used, per this outlet’s fact-checking policy.
Benchmark numbers come from three separate third-party testing operations, not from any single provider’s marketing page, since self-reported cloud benchmarks are notoriously unreliable. None of the three benchmark sources used here (VPSchart, MyUnix, and AI Product Weekly) are affiliated with any of the five VPS providers being compared.
Full specs comparison table
The table below lines up each provider’s closest match to a 2 vCPU / 4 GB entry point, plus the smallest and a mid-tier option, using specs pulled directly from official pricing pages and cross-checked against live pricing trackers current as of September 2026.
| Provider | Plan | vCPU | RAM | Storage | Bandwidth included | Storage type | Billing granularity | Regions | Managed add-ons |
|---|---|---|---|---|---|---|---|---|---|
| DigitalOcean | Basic 1 GiB | 1 (shared) | 1 GB | 25 GB | 1,000 GB | SSD | Hourly, monthly cap | 15 regions | Yes (DBs, Spaces, App Platform) |
| DigitalOcean | Basic 4 GiB | 2 (shared) | 4 GB | 80 GB | 4,000 GB | SSD | Hourly, monthly cap | 15 regions | Yes |
| AWS Lightsail | $5 bundle | 2 (burstable) | 0.5 GB | 20 GB | 1,000 GB | SSD | Hourly, monthly cap | AWS global regions | Yes (full AWS console access) |
| AWS Lightsail | $24 bundle | 2 (burstable) | 4 GB | 80 GB | 4,000 GB | SSD | Hourly, monthly cap | AWS global regions | Yes |
| Vultr | Cloud Compute 1GB | 1 (shared) | 1 GB | 25 GB | ~1,000 GB | NVMe | Hourly, per-second | 32+ locations | Limited (managed DBs separately) |
| Vultr | vc2-2c-4gb | 2 (shared) | 4 GB | ~80 GB | ~3,000 GB | NVMe | Hourly, per-second | 32+ locations | Limited |
| Linode (Akamai) | Nanode 1GB | 1 (shared) | 1 GB | 25 GB | ~1,000 GB | SSD | Hourly, monthly cap | 11 regions + Akamai edge | Yes (managed DBs, NodeBalancers) |
| Linode (Akamai) | Linode 4GB | 2 (shared) | 4 GB | ~80 GB | ~4,000 GB | SSD | Hourly, monthly cap | 11 regions + Akamai edge | Yes |
| Hetzner | CPX12 | 1 (shared) | 2 GB | 40 GB | 1,000-20,000 GB (region dependent) | NVMe | Hourly, monthly cap | Germany, Finland, US, Singapore | No (unmanaged) |
| Hetzner | CX23 | 2 | 4 GB | 40 GB | 20,000 GB (EU) / 1,000 GB (US) | SSD | Hourly, monthly cap | Germany, Finland, US, Singapore | No |
The bandwidth column is the one that trips people up. Hetzner’s 20 TB European allowance dwarfs everyone else, but that figure drops to 1 TB the moment you deploy in a U.S. location, and to 0.5 TB in Singapore, per Hetzner’s own documentation. AWS Lightsail’s bandwidth is generous relative to its price but Amazon still meters data transfer separately the moment you touch any other AWS service, like an S3 bucket or a CloudFront distribution, from that Lightsail box.
Pricing table: monthly cost at three RAM tiers
Here’s the direct monthly price comparison, sourced from each provider’s official pricing page (DigitalOcean, AWS Lightsail, Hetzner) and cross-referenced against the live pricing trackers Pikkly and HostMath for Vultr and Linode/Akamai figures, since those two vendors render pricing dynamically and don’t expose static numbers to standard scraping.
| Provider | ~1 GB RAM tier | ~2-4 GB RAM tier | ~8 GB RAM tier | Overage / extra bandwidth cost |
|---|---|---|---|---|
| DigitalOcean | $6/mo (1 GB) | $24/mo (4 GB) | $48/mo (8 GB) | $0.01/GB outbound |
| AWS Lightsail | $5/mo (0.5 GB) | $24/mo (4 GB) | $44/mo (8 GB) | $0.09/GB outbound (standard AWS rate outside bundle) |
| Vultr | $5/mo (1 GB) | $20/mo (4 GB) | $40/mo (8 GB) | $0.01/GB outbound |
| Linode (Akamai) | $5/mo (1 GB) | $24/mo (4 GB) | $48/mo (8 GB) | $0.02/GB outbound (varies by region) |
| Hetzner | ~$4.35/mo (ARM CAX11, 4 GB) | $7.09/mo (CX23, 4 GB, post-hike) | ~$14.33/mo (ARM CAX31, 16 GB) | ~$1.08/TB overage (EU), higher in US/Singapore |
Two things jump out. First, Hetzner’s ARM-based CAX line (built on Ampere Altra chips rather than x86) is dramatically cheaper per gigabyte of RAM than any x86 plan from any of the five providers, but only workloads that run cleanly on ARM64 can use it. Second, DigitalOcean and Linode land within a dollar of each other at every tier, which tells you the budget VPS market has largely converged on a shared price floor for x86 shared-CPU compute, somewhere around $24/month for 4 GB of RAM, with Vultr undercutting that floor by roughly 15-20%.
A worked example: real first-year cost for a small SaaS API
Sticker prices rarely match what actually lands on a card statement. Take a common scenario: a small SaaS API running on a 4 GB RAM instance, with 2.5 TB of monthly outbound transfer (busier than a hobby project, lighter than a media-heavy app), plus weekly automated backups.
| Provider | Base monthly (4 GB tier) | Backup add-on | Bandwidth overage (2.5 TB used) | Realistic monthly total | Year-one total |
|---|---|---|---|---|---|
| DigitalOcean | $24.00 | ~$4.80 (20%) | $0 (within 4,000 GB) | ~$28.80 | ~$345.60 |
| AWS Lightsail | $24.00 | ~$4.80 (est., 20%) | $0 (within 4,000 GB) | ~$28.80 | ~$345.60 |
| Vultr | $20.00 | ~$4.00 (20%) | $0 (within ~3,000 GB) | ~$24.00 | ~$288.00 |
| Linode (Akamai) | $24.00 | ~$4.80 (20%) | $0 (within ~4,000 GB) | ~$28.80 | ~$345.60 |
| Hetzner (CX23, US region) | $7.09 | ~$1.50 (est., per-GB snapshot) | ~$1.62 (1,500 GB over 1,000 GB cap at ~$1.08/TB) | ~$10.21 | ~$122.52 |
Even after paying a U.S.-region bandwidth overage, Hetzner comes out roughly a third of the cost of the four other providers for this workload, which is the clearest illustration of why it remains the default recommendation for cost-sensitive teams willing to manage their own patching and skip the managed dashboard. Deploy that same Hetzner instance in an EU region instead, and the bandwidth overage line disappears entirely, since 2.5 TB sits comfortably under the 20 TB EU allowance, pushing the realistic monthly total down closer to $8.59.
Benchmark data: what you actually get for the money
Price alone doesn’t tell you whether a box is fast. Three independent benchmark trackers give a consistent enough picture to draw real conclusions.
VPSchart’s benchmark database, which runs Geekbench, fio, and iperf3 against comparable 2 vCPU / 2 GB plans, put Hetzner’s CX32 ahead on raw CPU score at 9,200, with Vultr at 8,900, DigitalOcean at 8,800, and Linode’s Nanode 2GB at 8,700. The spread between the top and bottom of that pack is under 6%, close enough that CPU generation alone won’t decide your choice.
A separate fio and iperf3 benchmark run published by MyUnix found a bigger gap on disk performance: Vultr posted 87,400 random-read IOPS versus Linode’s 51,200 and DigitalOcean’s 44,800 on comparable plans, using fio 3.36 with a 4 KB random read/write mix against a 4 GB test file, per MyUnix’s VPS benchmark methodology writeup. Vultr also led on internal network throughput (9.4 Gbit/s versus DigitalOcean’s 8.7 Gbit/s), which lines up with Vultr’s marketing claim of NVMe storage across its Regular Performance line.
A third data point, a seven-provider speed test published on AI Product Weekly’s Substack, measured raw disk I/O and same-region latency and found Hetzner posting the highest disk throughput of the group tested (1,150 MB/s) but also the highest latency to a U.S.-East test point at 95ms, a predictable result given Hetzner has no U.S.-East region and routes from Ashburn to its nearest node. Vultr and Linode both stayed under 15ms in that same test, while AWS Lightsail trailed on raw disk I/O at 720 MB/s despite AWS’s overall infrastructure scale.
The pattern across all three sources: Vultr punches above its price on disk and network throughput, Hetzner wins on CPU-per-dollar but loses on latency outside Europe, and DigitalOcean and Linode sit in a steady middle tier where you’re paying more for polish, documentation, dashboard, managed services, than for raw hardware speed.
Uptime, reliability, and the numbers that don’t move much
VPSchart’s rolling 12-month uptime tracking put Hetzner’s CX32 at 99.98%, DigitalOcean’s Basic 2GB at 99.97%, Vultr’s Cloud 2GB at 99.96%, and Linode’s Nanode 2GB at 99.96%. Those numbers are close enough (a spread of 0.02 percentage points) that uptime alone shouldn’t be a deciding factor between any of the four, all four sit comfortably above the 99.9% threshold most SLAs promise, and the real-world differences show up more in how fast support responds to an incident than in raw availability percentages.
Real-world use cases: which provider fits which job
Side project or portfolio site. DigitalOcean’s $6/month Basic Droplet or Vultr’s $5/month Cloud Compute plan both cover a small Node.js or Django app comfortably. DigitalOcean’s one-click app marketplace saves setup time if you’d rather not hand-configure Nginx and a process manager.
Self-hosted media server or home lab extension. Hetzner’s CX line is the default recommendation in most self-hosting communities specifically because of the bandwidth allowance. 20 TB in EU regions covers a Plex or Jellyfin server streaming to a handful of users without triggering overage charges that would hit fast on AWS or DigitalOcean.
Startup already committed to AWS. Lightsail makes sense when you want a cheap staging box or a small internal tool that still needs an IAM role, a VPC peering connection, or access to RDS. Paying a small premium over Vultr or Hetzner buys you one fewer vendor to manage and one fewer set of credentials to rotate.
Latency-sensitive global app. Linode/Akamai’s edge network and Vultr’s 32+ location spread both beat Hetzner here, since Hetzner has no presence in South America, Africa, or most of Asia outside Singapore. If your users are spread across continents, region count matters more than the per-GB price.
CI/CD runners or batch jobs. Vultr’s per-second billing and fast boot times make it a good fit for ephemeral compute you spin up and tear down repeatedly, since you’re not paying for a full hour of an instance that only ran for six minutes.
Agency hosting client sites. A web agency juggling a dozen small WordPress or Laravel client sites benefits from DigitalOcean’s or Linode’s managed load balancers and snapshot tooling, since the time saved on infrastructure babysitting across a dozen accounts usually outweighs the per-instance price gap versus Hetzner or Vultr.
Student or hobbyist learning DevOps. Vultr and Linode both offer the lowest entry price at $5/month for a full Linux box with root access, and Linode in particular has a long-running library of community guides that makes it a common first stop for anyone working through a Kubernetes or Docker tutorial for the first time.
Scaling path: what happens when you outgrow the base tier
Every provider here lets you resize a running instance to a bigger plan within the same product line, usually with a short reboot, but the ceiling and the next step look different depending on where you started. DigitalOcean Droplets scale up to dedicated CPU plans with up to 48 vCPUs before you’d need to look at a different product entirely. AWS Lightsail’s upgrade path is the most distinctive of the five: once a workload outgrows the largest Lightsail bundle, you can migrate the instance directly into EC2 and keep the same AWS account, IAM roles, and VPC configuration, without re-architecting from scratch.
Vultr and Linode both support in-place resizing across their shared-CPU tiers and offer separate dedicated-CPU and high-memory product lines for workloads that need guaranteed (not burstable or shared) compute. Hetzner’s resize path moves a CX or CPX instance up through the Regular Performance line up to the CPX62 (16 vCPU, 32 GB), beyond that, Hetzner’s dedicated root server product is a separate purchase rather than a resize.
The practical takeaway: if you expect meaningful growth within 12-18 months, weigh the ceiling of each provider’s compute line as heavily as the entry price, since a mid-migration provider switch costs real engineering hours that erase any savings from the cheaper starting tier.
Migration guide: moving a workload between providers
Moving a VPS workload is more mechanical than moving a managed database, but it’s still easy to get wrong, and the risk is almost always downtime or a silently broken firewall rule rather than lost data. Here’s the process that avoids both for a typical Linux app server.
- Spin up the destination instance first and keep the source running. Don’t tear anything down until the new box is verified working end to end.
- Match OS and kernel versions. Hetzner and Vultr both default to newer kernel builds than some DigitalOcean base images, which can matter for containerized workloads relying on specific cgroup behavior.
- Copy application code and dependencies via rsync or a git deploy, not a full disk image, unless you’re moving between the same hypervisor family (most of these providers run KVM, so raw disk images often work, but always test first).
- Export and re-import your database separately from the app files. Use pg_dump or mysqldump for a clean logical export rather than copying data directory files across providers with different storage backends.
- Recreate firewall rules manually. Cloud firewall syntax isn’t portable: DigitalOcean Cloud Firewalls, AWS security groups, Hetzner Cloud Firewalls, and Vultr’s firewall groups all use different rule formats.
- Update DNS with a low TTL set at least 24 hours before cutover, so the switch propagates fast when you flip the A record.
- Run both instances in parallel for 24-48 hours, mirroring traffic if possible, before decommissioning the old box.
- Cancel the old instance only after confirming backups of the new one are working, not just configured.
The most common mistake in a VPS migration isn’t the server move itself, it’s forgetting that bandwidth-included allowances reset per provider. A workload that stayed comfortably under DigitalOcean’s 4 TB allowance can blow past Hetzner’s 1 TB U.S. limit if you deploy in an American region instead of a European one, so check the region-specific transfer cap before you commit, not after the first invoice.
Data center locations: where each provider actually runs
Region coverage decides latency more than any other variable in this comparison, and it’s the spec most people skip until after they’ve already committed. DigitalOcean runs 15 data center regions spread across North America, Europe, and Asia-Pacific. AWS Lightsail inherits the full reach of AWS’s global region footprint, the broadest of the five by a wide margin, though Lightsail itself isn’t available in every AWS region.
Vultr’s 32+ locations is the largest footprint among the pure budget-VPS players, stretching into markets like Mumbai, Seoul, and Tel Aviv that DigitalOcean and Linode don’t cover directly. Linode/Akamai runs 11 traditional compute regions but layers Akamai’s much larger edge and CDN network on top, which helps static content and API response times even in cities without a full Linode compute region. Hetzner remains the outlier: just Germany, Finland, the U.S. (Ashburn and Hillsboro), and Singapore. That’s a real constraint for any product with a meaningfully global user base, and it’s the single biggest reason teams choose Vultr or Linode over Hetzner despite the price gap.
Pros and cons of each provider
DigitalOcean
Pros: best documentation in the budget tier, one-click apps and managed databases, predictable flat pricing, large community of tutorials. Cons: priced roughly 15-20% above Vultr and Hetzner for equivalent specs, bandwidth allowance smaller than Hetzner’s.
AWS Lightsail
Pros: flat, predictable pricing inside a notoriously unpredictable AWS billing universe, straightforward upgrade path to EC2 and other AWS services, strong global region coverage. Cons: weakest raw disk throughput of the five in third-party benchmarks, bundle pricing gets confusing once you compare it against plain EC2 or other Lightsail bundle families.
Vultr
Pros: strongest benchmark results on disk and network throughput relative to price, 32+ regions, per-second billing. Cons: fewer managed add-ons than DigitalOcean, smaller community and fewer third-party tutorials.
Linode (Akamai)
Pros: most stable pricing history of the five, Akamai’s edge network is a genuine latency advantage, solid managed database and load balancer options. Cons: fewer regions than Vultr, brand identity muddled since the Akamai rebrand confuses some longtime users.
Hetzner
Pros: lowest price-per-core of any provider here even after the August 2026 hike, best bandwidth allowance by a wide margin in EU regions, all-NVMe storage on the Regular Performance line. Cons: just four region clusters, no managed database or one-click app ecosystem, U.S. and Singapore bandwidth allowances far smaller than EU, recently raised prices 30%+.
Bandwidth, snapshots, and the hidden costs that change the real price
The advertised monthly price is rarely the full story on a budget VPS. Snapshots and backups are billed separately by every provider here except in cases where they’re bundled: DigitalOcean charges roughly 20% of the Droplet’s monthly cost for automated backups, Vultr charges a similar percentage, and Hetzner’s snapshot pricing is billed per GB of stored snapshot rather than as a percentage of the instance cost, which can work out cheaper for small instances and more expensive for large ones.
IPv4 addresses are the other line item people forget. Hetzner folded its former €0.60/month IPv4 surcharge into the new base price as part of the August restructuring, but AWS charges separately for a static IP not attached to a running instance, and several providers are moving toward IPv6-first defaults as IPv4 exhaustion pushes wholesale address prices up industry-wide. If your app needs a dedicated IPv4 address for a TLS certificate or a legacy integration, add that line item before comparing headline prices.
Support is the last hidden variable. Baseline ticket support is free across all five providers, but faster response times generally cost extra. AWS gates anything beyond basic support behind its Developer, Business, or Enterprise support plans, which are priced as a percentage of total monthly AWS spend rather than a flat fee, meaning support costs on a single small Lightsail instance can look disproportionately high if you opt into a paid tier. DigitalOcean and Linode both offer flat-fee premium support add-ons, while Hetzner and Vultr lean more heavily on community forums and documentation for anything outside a priority-severity outage.
Developer tooling: APIs, CLIs, and infrastructure-as-code
All five providers publish a REST API and a first-party or community-maintained Terraform provider, which matters more than it sounds once you’re managing more than one or two servers by hand. DigitalOcean’s API and doctl CLI are widely considered the most polished of the group, with response times and documentation quality that consistently show up as a selling point in developer surveys. AWS Lightsail’s API is really a thin wrapper over a subset of AWS’s broader API surface, which means anyone already comfortable with the AWS CLI or the AWS provider for Terraform has effectively zero learning curve.
Vultr and Linode both maintain official Terraform providers and reasonably complete CLIs (vultr-cli and linode-cli), though neither has the third-party tooling ecosystem DigitalOcean has accumulated over a decade of being the default budget-VPS recommendation. Hetzner’s hcloud CLI and Terraform provider are actively maintained and cover the full feature set, but Hetzner’s API rate limits are tighter than the other four, worth knowing if you’re planning to provision or tear down instances programmatically at high frequency, such as for ephemeral CI runners.
# Example: minimal Terraform resource block for each provider's cheapest x86 tier
# DigitalOcean
resource "digitalocean_droplet" "web" {
image = "ubuntu-24-04-x64"
name = "web-1"
region = "nyc3"
size = "s-1vcpu-1gb"
}
# Vultr
resource "vultr_instance" "web" {
plan = "vc2-1c-1gb"
region = "ewr"
os_id = 2284
}
# Hetzner
resource "hcloud_server" "web" {
name = "web-1"
image = "ubuntu-24.04"
server_type = "cx23"
location = "nbg1"
}Security and compliance considerations
None of these five providers manage OS-level security for you on a standard compute instance, that’s true of AWS, DigitalOcean, Vultr, Linode, and Hetzner alike. You’re responsible for patching, firewall configuration, and SSH hardening regardless of which one you pick. Where they differ is in the tooling layered on top: DigitalOcean and Linode both offer managed Cloud Firewalls with a web UI, AWS Lightsail inherits AWS’s security group model plus optional AWS Shield for DDoS mitigation, and Hetzner’s Cloud Firewall is functional but sparse compared to the other four.
For workloads touching regulated data, AWS Lightsail is the only one of the five with a straightforward path to formal compliance certifications (SOC 2, HIPAA-eligible services) inherited from the broader AWS compliance program, since Lightsail resources can sit inside the same AWS account as services already covered by a Business Associate Addendum. DigitalOcean, Vultr, and Linode all publish SOC 2 reports, but HIPAA workloads generally require a more deliberate architecture review on those platforms. Hetzner’s compliance posture is oriented around GDPR and EU data protection law rather than U.S. frameworks, which makes it a strong choice for EU-first products but a weaker fit for U.S. healthcare or government workloads.
Verdict: which cheap cloud VPS should you actually pick
There’s no single winner here, and anyone claiming otherwise is skipping past your actual workload. But the data supports a few clear defaults.
If price-per-core is your only variable and you can deploy in Europe, Hetzner wins clearly even after its 2026 increase, at roughly a third of DigitalOcean’s cost for comparable specs. If you need U.S. presence with that same bargain pricing, Hetzner’s advantage shrinks to roughly parity with Vultr once you account for the smaller U.S. bandwidth allowance.
If you want the best balance of raw performance and price without sacrificing region coverage, the three independent benchmark sources point to Vultr. It beat DigitalOcean and Linode on disk I/O and network throughput in two of three benchmark sets while staying roughly 15-20% cheaper at every RAM tier.
If your priority is minimizing operational overhead rather than minimizing the invoice, DigitalOcean’s documentation and managed add-ons are worth the premium, especially for a small team without a dedicated ops person. And if you’re already an AWS shop, Lightsail’s convenience usually outweighs its middling benchmark numbers, since the real cost of a second cloud vendor is rarely visible on a pricing page.
The broader lesson from lining up five providers side by side is that the headline monthly price answers a narrower question than most buyers assume. Bandwidth caps, backup fees, region coverage, and support tiers move the real cost by 20-40% in either direction depending on the workload, and the worked example above shows how quickly that math shifts once a single variable, region, in that case, changes. Treat the number on the pricing page as a starting bid, not a final answer, and rerun it against your own traffic and backup needs before committing a production workload to any of the five.
Frequently asked questions
What’s the cheapest cloud VPS in September 2026?
On a pure price-per-resource basis, Hetzner’s ARM-based CAX11 plan (2 vCPU, 4 GB RAM) at roughly $4.35/month is the cheapest of the five compared here, though it requires ARM64-compatible software. For x86 workloads, Vultr’s $5/month Cloud Compute 1GB plan and Linode’s $5/month Nanode 1GB are the lowest entry points.
Did Hetzner really raise prices in 2026?
Yes. Hetzner’s shared x86 CX line increased by roughly 30-50% depending on the plan in August 2026, with the CX23 (2 vCPU, 4 GB) moving from €3.99 to €5.99 a month. Hetzner folded a previously separate IPv4 charge into the new price, which softens the effective increase to around 30% on a like-for-like basis.
Is DigitalOcean or Vultr better for beginners?
DigitalOcean generally has the edge for beginners because of its documentation depth and one-click application marketplace. Vultr is comparably easy to use but has a smaller library of community tutorials and fewer managed one-click installs.
Does AWS Lightsail use the same infrastructure as EC2?
Lightsail runs on the same underlying AWS infrastructure as EC2 but with a simplified, bundled pricing model. You can upgrade a Lightsail instance to a full EC2 deployment later, which is one of its main advantages over the other providers in this comparison.
Why did Linode become Akamai?
Akamai acquired Linode in 2022 and has since rebranded the compute product as Akamai Cloud Compute, though the legacy plan names (Nanode, Linode 2GB, and so on) are still used in billing and documentation. The infrastructure now benefits from Akamai’s global CDN and edge network.
Which provider has the best bandwidth allowance?
Hetzner offers by far the largest bandwidth allowance at up to 20 TB per month on European instances, though that drops sharply to 1 TB on U.S. instances and 0.5 TB in Singapore. For U.S.-based deployments, DigitalOcean and AWS Lightsail’s 4 TB allowances at the 4 GB RAM tier are more generous in absolute terms.
Can I run production workloads on a $5-6/month VPS?
Yes, for low-traffic sites, APIs, or internal tools. A 1 GB RAM instance comfortably runs a small Node.js, Python, or PHP application behind Nginx, but you’ll want to add swap space, monitor memory usage closely, and plan an upgrade path before traffic grows meaningfully.
Do these providers charge extra for backups?
Most do. DigitalOcean and Vultr both charge roughly 20% of the instance’s monthly price for automated backups. Hetzner bills snapshots per GB stored rather than as a percentage, which can be cheaper for small instances. Always check backup pricing separately from the headline compute price.
Should I pick an ARM or x86 instance?
ARM instances, like Hetzner’s CAX line, are meaningfully cheaper per gigabyte of RAM, but they only make sense if your stack compiles cleanly for ARM64. Most modern language runtimes (Node.js, Python, Go, Rust) support ARM64 natively, and most current Docker base images publish multi-arch builds, so the compatibility risk in 2026 is lower than it was even two years ago. Older or vendor-locked binaries are the main thing to check before switching.
Can I switch providers later without much pain?
Yes, if your infrastructure is defined as code (Terraform, Ansible, or even a well-documented setup script) rather than configured by hand through a dashboard. The migration steps outlined above take a few hours for a typical single-server app, and the real cost is usually DNS propagation time, not the technical move itself.
