Finding a VPS under four dollars used to mean settling for 512 MB of RAM, no IPv4, or uptime promises written in disappearing ink. Not anymore. The market shifted in the past two years, and a handful of providers now offer legitimate 1 GB plans with real IP addresses and SLAs that matter.
I've spent the last month testing entry-level VPS plans from providers who consistently stay under the $4 mark. This isn't a race to the bottom—it's about finding hosts who can deliver baseline reliability without charging shared-hosting prices for virtualized resources.
What "cheap" actually means in 2026
Cheap is relative. A $3.50 VPS that reboots twice a week isn't cheap—it's expensive downtime. When I say "cheapest," I mean plans that meet three non-negotiable criteria: at least 1 GB of RAM, a dedicated IPv4 address, and a published uptime SLA of 99.5% or better.
Why 1 GB? Anything less struggles to run a modern Linux distribution plus a web server and database without constant swapping. You'll spend more time troubleshooting OOM kills than building your project.
IPv4 matters because most tutorials, firewall rules, and monitoring tools still assume you have one. IPv6-only plans are fine for experimentation, but they add friction when you're trying to connect a domain, configure Cloudflare, or SSH in from a coffee shop.
The SLA filters out resellers who vanish after three months. A written uptime guarantee means the provider has invested in infrastructure monitoring and has enough margin to stand behind it.
1. Vultr: $3.50/month, KVM, global locations
Vultr's entry plan sits at $3.50 monthly. You get 1 GB RAM, a single CPU core, 25 GB SSD, 1 TB transfer, and a dedicated IPv4. The control panel is straightforward—pick a datacenter, choose an OS, and you're SSH-ing in two minutes later.
They run KVM virtualization, which matters if you need full kernel access or want to test custom modules. I've run production email relays and small Django apps on these instances without issues. Disk I/O is the bottleneck, as expected at this price, but sequential writes are fine for logs and app data.
Vultr's SLA promises 100% uptime with credits for any outage. In practice, the last twelve months saw two brief maintenance windows in the Singapore datacenter, both announced 72 hours ahead.
2. Hetzner Cloud: €3.79/month, Falkenstein and Helsinki
Hetzner's CX11 plan converts to roughly $3.80 at current exchange rates. You get 1 vCPU, 2 GB RAM—yes, two gigabytes—20 GB SSD, and 20 TB traffic. That's double the RAM of most competitors at this price.
The catch? Datacenter choice is limited to Germany and Finland. Latency to North America adds 90-120 ms compared to a US East host. For backend services, dev environments, or European audiences, that's irrelevant. For a customer-facing site serving US traffic, test your tolerance first.
Hetzner publishes live status pages and maintains 99.9% uptime across their cloud fleet. Provisioning is instant, and the API is clean if you're scripting infrastructure. I've had a CX11 running Postfix, Dovecot, and Roundcube for eighteen months with zero unplanned reboots.
3. Contabo: €3.99/month, more CPU, controversial reputation
Contabo's VPS S plan runs €3.99 (about $4.10, so technically over the line, but close enough to mention). You get 4 vCPUs, 8 GB RAM, 100 GB NVMe, and 32 TB traffic. Those specs are absurd for the price.
The tradeoff is noisy neighbors and occasional slow-disk complaints on Reddit. Contabo oversells aggressively, and you'll notice it during European business hours when I/O wait spikes. For batch processing, cron jobs, or services that don't demand consistent sub-10ms disk latency, it's fine.
Their SLA is 99.9%, and in two years of running backup storage on a Contabo instance, I've seen one four-hour outage after a network switch failure in Munich. Support took 36 hours to respond to the ticket, which aligns with the budget-tier experience.
4. Racknerd: $3.18/month via annual prepay
Racknerd specializes in cheap annual deals. Their entry VPS costs $38.20 per year, which breaks down to $3.18 monthly. You get 1.5 GB RAM, 1 vCPU, 21 GB SSD, 1500 GB transfer, and one IPv4.
The billing structure means you're committing to a year upfront. That's fine if you're running a side project or personal VPN, less ideal if you're testing a business concept and might pivot in three months.
I've used Racknerd instances for offsite backups and jump hosts. Uptime has been consistent, provisioning is semi-automated (you get a setup email within 24 hours), and the control panel is SolusVM, which is functional but dated. They don't advertise an SLA percentage, but downtime in the past year totaled under six hours across all my instances.
5. BuyVM: $3.50/month, OpenVZ and KVM options
BuyVM's KVM slice costs $3.50 for 1 GB RAM, 1 CPU, 20 GB SSD, and 1 TB transfer. They operate datacenters in Las Vegas, Miami, New York, and Luxembourg.
What sets them apart is optional block storage. You can attach extra disks at $1.25 per 256 GB/month, which makes them a strong pick if you need to store backups, mail archives, or media files without paying per-GB transfer fees.
BuyVM has a cult following in the LowEndBox community. They've been around since 2010 and prioritize transparency over marketing. Their status page lists every incident with root-cause analysis. Uptime over the last year hovered around 99.8%, with one six-hour outage caused by a fiber cut in Las Vegas.
What to watch out for at this price
Cheap VPS hosting has failure modes you won't see on $20/month plans.
First, expect slow support. You're not getting 15-minute response times. Budget providers triage by revenue, and a $3.50 plan puts you near the bottom. Open a ticket for a stuck provisioning job or broken reboot button and wait 12-48 hours.
Second, disk I/O is shared and unpredictable. If your neighbor decides to run a file indexer that hammers the SAN, your write latency spikes. This matters for databases and busy WordPress installs. Test with dd and ioping after provisioning to set expectations.
Third, network routes aren't optimized. You might get 500 Mbps port speed but terrible peering to certain regions. Run MTR tests to your target audience's location before migrating production traffic.
Fourth, provider churn is real. Budget hosts fold or get acquired. Keep offsite backups and exportable configs so you can migrate in an afternoon.
Which one should you pick?
If you need maximum RAM and don't care about latency to the US, go Hetzner. Two gigs at this price is unmatched.
If you want global datacenter choice and a polished control panel, pick Vultr. It's the safest bet for first-time VPS users.
If you're running storage-heavy or CPU-heavy batch jobs and can tolerate variability, Contabo's specs win.
If you're committing to a year and want the absolute lowest monthly cost, Racknerd's annual prepay is hard to beat.
If you need cheap block storage or prefer a provider with a long track record, BuyVM fits.
I rotate between Vultr and Hetzner depending on the project. Vultr gets frontend services that need US reach; Hetzner handles backend APIs and internal tools where the extra RAM helps.
Initial setup checklist
Once you've provisioned a cheap VPS, lock it down before doing anything else.
Disable root SSH login. Create a sudo user and move your public key into place:
adduser deploy
usermod -aG sudo deploy
mkdir -p /home/deploy/.ssh
cp /root/.ssh/authorized_keys /home/deploy/.ssh/
chown -R deploy:deploy /home/deploy/.ssh
chmod 700 /home/deploy/.ssh
chmod 600 /home/deploy/.ssh/authorized_keys
Then edit /etc/ssh/sshd_config and set PermitRootLogin no and PasswordAuthentication no. Restart SSH with systemctl restart sshd.
Install fail2ban to block brute-force attempts:
apt update && apt install fail2ban -y
systemctl enable fail2ban
Enable automatic security updates. On Ubuntu or Debian:
apt install unattended-upgrades -y
dpkg-reconfigure -plow unattended-upgrades
Set up a firewall. If you're only running a web server:
ufw allow 22/tcp
ufw allow 80/tcp
ufw allow 443/tcp
ufw enable
Check your hostname and timezone:
hostnamectl set-hostname yourhostname
timedatectl set-timezone America/New_York
Finally, snapshot the fresh install if your provider supports it. Vultr, Hetzner, and BuyVM all allow free snapshots of stopped instances. Boot it back up, and you have a clean baseline to revert to if an experiment goes wrong.
Final checks before you deploy
Cheap VPS hosting works when you match the plan to the workload. Don't expect $20/month reliability from a $3.50 instance, but also don't assume it's junk. I've run email servers, DNS resolvers, monitoring agents, and CI runners on budget VPS plans for years without major issues.
Read the provider's acceptable use policy before hosting anything sensitive. Some budget hosts prohibit outbound SMTP on port 25 to prevent spam, which breaks self-hosted mail. Others block cryptocurrency miners or tor exit nodes. Know the rules.
Monitor your VPS from the outside. Use UptimeRobot or a similar service to ping your instance every five minutes and alert you if it goes down. Budget providers won't always notify you of maintenance windows.
Keep your kernel updated. Run apt update && apt upgrade weekly or enable unattended upgrades. A cheap VPS that's compromised costs you more in cleanup time than you saved on hosting.
The cheapest VPS isn't the one with the lowest monthly price. It's the one that stays online, responds to your SSH sessions, and gets out of your way so you can build.
