Cloudways vs Hostinger VPS for Self-Hosted Web Apps: The 52-Minute Decision
Published 5 September 2026. Cloudways pricing and included features are taken from their public pricing page on that date. The Hostinger figures are measured on a live KVM2-class server we operate. We have no Cloudways account and did not benchmark their platform; every performance number below is labelled with its source.
1. The Verdict
Cloudways Medium costs $54 per month for 4 GB of RAM and 2 vCPU. A Hostinger KVM2-class VPS costs roughly $9 per month for 8 GB and 2 vCPU. Six times the price for half the memory, which sounds like an easy decision until you price the thing Cloudways is actually selling.
They are not selling hardware. They are selling the hours you do not spend configuring it. The equation that decides this is short:
Monthly saving from self-managing:
D = 54.00 - 10.50 = $43.50 (Cloudways Medium vs VPS + backup storage)
Hours of your time that saving covers:
H = D / your_hourly_rate
At $50/hour: 43.50 / 50 = 0.87 hours = 52 minutes per month
If managing the server costs you MORE than 52 minutes a month,
Cloudways is cheaper in real terms. If less, the VPS is.Our own honest estimate of maintaining a VPS running several applications is about one hour a month: security updates, certificate checks, backup verification, and the occasional thing that breaks. That is above the 52-minute line. For a single site run by someone who bills their time, Cloudways is the rational choice on cost alone, and the case for a VPS is control rather than money.
The arithmetic inverts the moment the server carries more than one application, because compute is the only line that does not multiply. The machine these benchmarks ran on hosts five unrelated production sites on that same $9 tier. Five Cloudways applications on one Medium server is also permitted — they advertise unlimited applications — but five applications sharing 4 GB is a different proposition from five sharing 8 GB.
Decision Matrix
| If this describes you | Choose | Why |
|---|---|---|
| You bill your time and run one WordPress site | Cloudways | 52 minutes of your month costs more than the price difference |
| You are comfortable in a terminal and run several applications | Hostinger VPS | One server, many apps, and the marginal cost of app number four is memory rather than money |
| You need a staging environment and one-click restores | Cloudways | Both are included; building the equivalent yourself is an afternoon plus ongoing maintenance |
| Your stack is not WordPress, Magento, Laravel or plain PHP | Hostinger VPS | Root access runs anything: Node, Python, Go, Docker, a Django app beside a WordPress one |
| Nobody on the team reads an nginx error log at midnight | Cloudways | 24/7 support is included on every plan, and a VPS has none |
| You want to know exactly what runs on your machine | Hostinger VPS | Managed platforms are opaque by design; that opacity is the product |
2. Pricing, Side by Side
Cloudways figures below are their published Flexible tiers on DigitalOcean-class infrastructure, read from their pricing page on 5 September 2026, excluding tax. Hostinger figures are the tier of the machine we measured, at 24-month term pricing.
| Plan | Monthly | RAM | vCPU | NVMe | Bandwidth |
|---|---|---|---|---|---|
| Cloudways Micro | $14 | 1 GB | 1 | 25 GB | 1 TB |
| Cloudways Small | $28 | 2 GB | 1 | 50 GB | 2 TB |
| Cloudways Medium | $54 | 4 GB | 2 | 80 GB | 4 TB |
| Cloudways XL | $170 | 16 GB | 8 | 320 GB | 6 TB |
| Hostinger KVM2 (measured) | ~$9 | 7.8 GiB | 2 | 96 GB | 8 TB |
Two caveats on that last row, stated plainly. Hostinger’s headline price is a promotional 24-month term; renewal pricing is materially higher and we did not verify the current renewal rate, so check it before committing to the arithmetic. And the 7.8 GiB and 96 GB are what our machine actually reports, not what the plan page advertises — those are the numbers free and df return.
Cloudways bills hourly and lets you resize; the VPS is a term commitment. If your workload is genuinely spiky or short-lived, that flexibility is worth something the table cannot express.
Add-ons change the comparison
| Cloudways add-on | Price | VPS equivalent |
|---|---|---|
| Offsite backup storage | $0.033/GB per server | ~$1.50/mo for 50 GB on S3-compatible storage |
| Cloudflare Enterprise CDN | from $4.99/domain/mo | Cloudflare free tier, or paid if you need the enterprise features |
| Site Manager | from $3/app/mo | Your own scripts, or nothing |
| Malware protection | from $4/app/mo | File integrity baseline via cron, free |
| Email mailboxes | $1.50/mailbox/mo | Same cost anywhere; a VPS IP cannot send reliably either |
| Advanced support | $100/mo (currently discounted) | Not available at any price |
| DNS with fast propagation | $0.50/domain/mo | Cloudflare DNS, free |
Read that table in both directions. Several add-ons have free equivalents a competent administrator sets up once. The last row of the previous section has no equivalent at all: you cannot buy 24/7 engineers for a VPS you administer yourself, and on the night something breaks that is either irrelevant or the only thing that matters.
3. What the Managed Layer Actually Buys
The most valuable thing bundled into a managed plan is the caching layer, and we can put a measured number on it because we built the equivalent by hand on the VPS and benchmarked both states.
| WordPress on our KVM2 | PHP executed per request | With an nginx FastCGI microcache | Delta |
|---|---|---|---|
| Throughput at concurrency 10 | 18.6 req/s | 749.6 req/s | 40.4x |
| Throughput at concurrency 50 | 19.9 req/s | 532.6 req/s | 26.7x |
| P95 latency at concurrency 50 | 2,692 ms | 132 ms | 20.4x lower |
| TTFB, median of 10 | 127 ms | 25 ms | 5.1x |
A factor of forty, from configuration that costs nothing and took about twenty minutes. Cloudways ships its own optimised stack with page caching, Redis and Object Cache Pro included — Object Cache Pro alone is licensed at $95 a month and is free on their 4 GB tiers and above. Their entire performance proposition is that this layer is already correct when you arrive.
The honest reading cuts both ways. If you would not have configured caching yourself, the managed plan is worth its price on this line alone, because an uncached WordPress site on any host ceilings around 20 requests per second. If you would have configured it — and section 4 contains the exact configuration — you are paying $45 a month for twenty minutes of work you already know how to do.
The full bundle, and what it costs to replicate
| Included with Cloudways | VPS equivalent | Effort to replicate |
|---|---|---|
| Optimised stack with page cache and Redis | nginx FastCGI cache + redis-server | ~20 min, measured at 40x above |
| Free Let’s Encrypt SSL | certbot --nginx | ~2 min, plus renewal monitoring |
| Automated backups, one-click restore | mysqldump + GPG + cron + S3 | ~1 h once, then a restore test quarterly |
| Staging environment and cloning | A second vhost and a database copy | ~1 h once, no one-click equivalent |
| Server-level firewall (Imunify360) | UFW + fail2ban | ~15 min, less capable |
| Vertical scaling without rebuilding | Provider resize, usually with a reboot | Depends on provider |
| Vulnerability scanner, auto-healing | No direct equivalent | You are the auto-healing |
| 24/7 support engineers | None | Cannot be replicated |
Roughly three hours of one-time setup replicates most of the technical bundle, and our own guides cover each piece: the caching and PHP-FPM configuration in the CodeCanyon deployment guide, the framework-level tuning in the Laravel production setup, and the firewall and certificate work in our Ubuntu hardening walkthrough. What three hours cannot replicate is the last row, and the ongoing attention.
4. Architecture: Where the Real Differences Sit
What you can run
Cloudways Flexible targets WordPress, Magento, Laravel and plain PHP. That covers a large share of the web and excludes everything else. Root access on a VPS is not a luxury feature here; it is the difference between a platform and a computer.
The machine these benchmarks ran on carries, simultaneously: a WordPress site, three Django applications, a Next.js application under PM2, PostgreSQL, MySQL, Redis, and Docker for the n8n and Cal.com deployments documented in our other articles. Each PHP site has its own PHP-FPM pool, so one application cannot exhaust another’s workers. That topology is not available on a managed PHP platform at any tier.
# The isolation that makes multi-tenancy safe on one box:
# a dedicated pool per site, each with its own socket and worker budget.
ls /etc/php/8.3/fpm/pool.d/
# app-one.conf app-two.conf stackrecipes.conf www.conf
grep -H "^pm.max_children\|^listen =" /etc/php/8.3/fpm/pool.d/*.confWho owns the operating system
On a managed platform the provider decides when PHP is upgraded, which extensions exist, and what the kernel is. That is mostly a benefit: someone else tracks CVEs. It becomes a constraint the day you need an extension they do not ship, or a PHP version they have not rolled out, or a kernel parameter their stack does not expose.
On a VPS you own all of it, including the obligation. The relevant question is not which is better but whether you will actually do the work:
# The minimum a self-managed server needs, automated
apt-get install -y unattended-upgrades
systemctl enable --now unattended-upgrades
# Certificate renewal, proven rather than assumed, monthly from cron
0 4 1 * * certbot renew --dry-run || echo "certbot dry-run FAILED" | \
mail -s "cert renewal check" you@example.com
# Backup that has actually been restored, quarterly
# (a backup never restored is a hypothesis, not a backup)Scaling
Cloudways scales vertically on demand and bills hourly, so a traffic spike is a slider rather than a migration. A VPS on a 24-month term is a fixed box: growing means a resize with downtime, or moving. For predictable workloads the fixed box is cheaper; for genuinely spiky ones, hourly billing is a real feature and not marketing.
Cloudways also markets a separate Autonomous product for WordPress that autoscales for high-concurrency ecommerce and LMS traffic. That is a different architecture from the Flexible plans compared here and a different conversation; if your requirement is thousands of concurrent users, neither a single VPS nor a fixed Flexible server is the right shape.
Worth keeping in proportion, though: our measured 749 requests per second on cached WordPress is 2.7 million page views an hour on a $9 server. Most sites that believe they need autoscaling need a cache.
Three-year cost, with labour priced in
| Line | Cloudways Medium | Hostinger KVM2-class |
|---|---|---|
| Hosting | $54.00 / mo | ~$9.00 / mo |
| Offsite backup, 50 GB | $1.65 / mo (at $0.033/GB) | $1.50 / mo |
| Maintenance, 1 h/mo at $50 | $0 | $50.00 / mo |
| True monthly | $55.65 | $60.50 |
| Year 1 | $667.80 | $726.00 |
| Year 3 cumulative | $2,003.40 | $2,178.00 |
| Year 3, four applications on one server | $2,003.40 (same server) | $2,178.00 (same server) |
| Year 3, if you value maintenance at $0 | $2,003.40 | $378.00 |
The bottom two rows are the ones to argue about, and they are why this comparison has no universal answer.
Both platforms allow multiple applications on one server, so consolidating four sites does not change either column — it changes the per-site cost identically on both sides, and the specification difference reasserts itself: four applications sharing 8 GB behave differently from four sharing 4 GB.
The final row is the case almost every “cheap VPS wins” article silently assumes: that your own hours cost nothing. That is a legitimate position for someone learning, for a side project, or for anyone who enjoys the work and would spend the evening on a terminal regardless. It is not a legitimate position for an agency billing clients. Decide which you are before reading either column as a conclusion.
One more line the table cannot hold: the price of being wrong. A managed platform absorbs your mistakes — a bad plugin update rolls back in one click, and a misconfiguration gets a support ticket. On a VPS, a mistake is downtime measured in however long it takes you to notice and diagnose it. Weight that by how much an hour of downtime actually costs your business, and for some readers it dominates every other figure here.
5. Benchmarks: One Side Measured, Stated As Such
What we measured. A Hostinger KVM2-class VPS: AMD EPYC 9354P, 2 vCPU, 7.8 GiB RAM, 96 GB disk, Ubuntu 24.04.4 LTS, nginx 1.24.0, PHP 8.3.6, MySQL 8.0.46, Redis 7.0.15. Load generated with ApacheBench 2.3 on the same host, which competes for the same cores and makes every figure a conservative floor.
What we did not measure. Cloudways. We hold no account there and ran nothing on their infrastructure. Publishing invented numbers for the other column would make this article worthless, so that column does not exist. What follows is a baseline you can reproduce against whichever host you are evaluating, including theirs.
Storage, measured
| Operation | Result | Method |
|---|---|---|
| Sequential write, 1 MiB blocks | 1.7 GB/s | dd oflag=direct, 1 GiB |
| Sequential read, 1 MiB blocks | 1.3 GB/s | dd iflag=direct, cache dropped |
| 4 KiB synchronous write | 6.3 MB/s ≈ 1,575 IOPS | dd oflag=direct,dsync, 32 MiB |
The third row governs database behaviour and is the one to compare between hosts. Sequential throughput is irrelevant to an InnoDB commit; synchronous 4 KiB latency is what caps transactions per second, and it is the figure marketing pages never publish.
Application throughput, measured
| Workload | Concurrency | Req/s | P50 | P95 |
|---|---|---|---|---|
| WordPress, uncached | 10 | 18.6 | 516 ms | 736 ms |
| WordPress, FastCGI microcache | 10 | 749.6 | 13 ms | 17 ms |
| WordPress, FastCGI microcache | 50 | 532.6 | 78 ms | 132 ms |
| Laravel 13, OPcache tuned | 10 | 190.8 | 49 ms | ~105 ms |
| Laravel 13, OPcache disabled | 10 | 16.4 | 589 ms | 789 ms |
Those five rows contain the argument of this entire article. The spread between the best and worst configuration on identical hardware is a factor of forty for WordPress and a factor of twelve for Laravel. Hardware differences between hosts are small compared with configuration differences on the same host, which is precisely why a managed platform that gets the configuration right can outperform a larger unmanaged server.
Run this against any host before you commit
# 1. Storage. The third command is the one that matters for databases.
dd if=/dev/zero of=/tmp/bench.tmp bs=1M count=1024 oflag=direct
sync; echo 3 > /proc/sys/vm/drop_caches
dd if=/tmp/bench.tmp of=/dev/null bs=1M count=1024 iflag=direct
dd if=/dev/zero of=/tmp/b4k.tmp bs=4k count=8192 oflag=direct,dsync
rm -f /tmp/bench.tmp /tmp/b4k.tmp
# 2. What CPU did you actually get? Shared tiers vary between neighbours.
lscpu | grep -E "^Model name|^CPU\(s\)"
# 3. Application throughput. Run from a SEPARATE machine if you can;
# a co-located generator competes for the same cores.
apt-get install -y apache2-utils
ab -n 200 -c 10 -k https://your-site.example.com/
# 4. Is the cache actually serving? Managed hosts differ in what they cache.
for i in 1 2 3; do
curl -s -o /dev/null -D - https://your-site.example.com/ \
| grep -iE "x-cache|x-fastcgi-cache|cf-cache-status|age:"
doneCheck number four is the one that settles a managed-hosting evaluation. If the response is served from cache, you are measuring the cache and not the host, and two providers will look identical. Compare uncached response times to compare infrastructure; compare cached ones to compare their caching layer. Both are worth knowing, and conflating them is how hosting benchmarks become meaningless.
6. Operational Reality on Each Side
What actually goes wrong on a self-managed VPS
Not exotic failures. The same handful, repeatedly, and all of them are avoidable with the checks in the last section of this article.
# Certificate expired because renewal was never tested
certbot renew --dry-run
# Disk filled with logs or database growth, and PostgreSQL went read-only
df -h /
du -sh /var/log/* | sort -h | tail -5
# A saturated PHP-FPM pool queues instead of erroring: the error-rate
# dashboard stays green while users wait seconds. Alert on P95, not on 5xx.
grep -c "server reached pm.max_children" /var/log/php8.3-fpm.log
# The backup was never restored, so nobody knew it was empty
zcat /var/backups/site/db-*.sql.gz | grep -c "^INSERT INTO"None of these requires deep expertise. All of them require somebody to look. That is the recurring cost the 52-minute equation is measuring, and underestimating it is the most common error in this comparison.
What you give up on a managed platform
Debugging depth. When something is slow and the platform will not give you the slow query log or the FPM status page, you are limited to the diagnostics the vendor chose to expose. On a VPS every layer is readable.
Colocation. Running a cron worker, a queue consumer or a small Go service beside your site is trivial with root and impossible on a PHP-only platform. Many businesses discover this requirement after the migration.
Exit cost. A VPS is a filesystem and a database dump; moving is a weekend. A managed platform’s convenience features — its staging, cloning and deployment tooling — are the things that do not come with you, and the more you rely on them the more the migration costs.
7. Frequently Asked Questions
Is Cloudways worth it compared to a cheap VPS?
Above 52 minutes of monthly maintenance, yes. Cloudways Medium is $54 against roughly $10.50 for a VPS plus backup storage, a $43.50 difference that buys 0.87 hours at a $50 hourly rate. Our honest estimate for maintaining a VPS is about an hour a month, which puts a single site on the Cloudways side of the line.
How much RAM do I get for the money?
Cloudways Medium gives 4 GB and 2 vCPU for $54 monthly. A Hostinger KVM2-class VPS gives 8 GB and 2 vCPU for around $9 on a 24-month term — six times cheaper for twice the memory. You are not paying for hardware on a managed plan; you are paying for the configuration, the backups and the support desk.
Is a managed host actually faster than a VPS?
Usually, because its caching layer is already correct. We measured WordPress on one machine at 18.6 requests per second uncached and 749.6 with an nginx microcache, a 40-fold spread from configuration alone. Managed platforms ship that layer configured. Configure it yourself and the hardware difference between hosts becomes the small term.
Can I run non-PHP applications on Cloudways?
Not on Flexible plans, which target WordPress, Magento, Laravel and plain PHP. A VPS with root access runs anything: our benchmark machine carries Django, Next.js under PM2, PostgreSQL, Redis and Docker containers alongside WordPress. If your roadmap includes a queue worker or a small service in another language, that constraint matters more than the price.
How do I benchmark two hosts fairly?
Measure 4 KiB synchronous disk writes rather than sequential throughput, since that is what caps database commits. Run the load generator from a separate machine. Then check the response headers: if you are hitting a cache, you are comparing caching layers rather than infrastructure. Measure both states deliberately and label which is which.
8. Decision Checklist
Answer these before you compare prices. The price difference is the smallest term in this decision.
# --- Is a VPS viable for you at all? ---
# Answer honestly, in writing:
# 1. What is your hourly rate, and how many hours per month will this cost?
# Under 52 min at $50/h, the VPS wins on money. Over, it does not.
# 2. Who fixes it at 02:00 on a Sunday? If the answer is "nobody",
# buy the support desk.
# 3. Will you run anything that is not PHP within 12 months?
# If yes, a managed PHP platform is a dead end.
# 4. How many applications will share the machine?
# One favours managed; four favours a VPS decisively.
# --- If you choose the VPS, these are non-negotiable ---
# Caching. Worth 40x on WordPress, measured. Without it you bought
# a bigger server to run slower software.
curl -s -o /dev/null -D - https://your-site/ | grep -i x-fastcgi-cache
# Backups that have actually been restored
zcat /var/backups/site/db-*.sql.gz | grep -c "^INSERT INTO"
# Certificate renewal proven, monthly from cron
certbot renew --dry-run
# Firewall enforcing, database and cache on loopback only
ufw status verbose
ss -tlnp | grep -E ':(3306|5432|6379)' # expect 127.0.0.1 only
# Unattended security upgrades armed
systemctl is-active unattended-upgrades
# External monitoring, because a check on the box reports healthy
# right up until the box is the problem
# --- If you choose managed, verify what you are buying ---
# Which layer is actually caching your responses?
curl -sI https://your-site/ | grep -iE "x-cache|cf-cache-status|age:"
# Can you reach the slow query log and the FPM status page?
# Does the backup restore, and have you tested it on their platform?
# What is the exit path if you leave in two years?The first block is the whole decision and it has nothing to do with hosting. Answer question two truthfully. Most people who choose a VPS on price answer it with an optimistic assumption about their own future availability, and the invoice for that assumption arrives at an inconvenient hour.
If you run the benchmark commands in section 5 against Cloudways or any other managed host, send us the raw output with your plan tier and we will publish the comparison with proper attribution. We would rather extend this article with measurements somebody actually took than fill the second column with estimates.