The Cloudways invoice for September came in at ₹38,412 — about ₹4,000 more than August, with no traffic change worth speaking of. That is the month most clients first ask whether they should move. The bandwidth meter ticked up because someone uploaded a 14 MB hero video to the homepage and forgot the CDN was on the per-GB plan, but the founder doesn't know that yet. He just knows the number was supposed to be ₹22,000 when they signed up two years ago, and now it isn't.
So we sit down and actually cost it. Three years. Both sides. Line items, not vibes.
What the Cloudways line item actually buys you
For a store doing somewhere between ₹80 lakh and ₹1.2 crore in annual GMV — call it the ₹1 crore band — the Cloudways bill typically sits on a DigitalOcean 4GB or 8GB droplet, with a couple of add-ons that the pricing page does not put on the front of the brochure. The DO 2GB plan on Cloudways runs about ₹2,400/month against ₹1,000/month for the raw droplet from DigitalOcean directly. The DO 8GB — which is where a ₹1 crore store usually lives once the catalogue crosses 800 SKUs and there's a Redis object cache running — is around ₹8,500/month on Cloudways versus roughly ₹4,200 raw. That's a markup of about 2x on the compute, which is the headline number everybody quotes.
The headline number is also the wrong one. The actual bill, in our experience, is:
- The droplet itself — ₹8,500 give or take.
- CloudwaysCDN, which is billed per 25 GB block. ₹85 or so per block, and a moderately busy WooCommerce store with image-heavy product pages will burn through 6–10 blocks a month without trying.
- Application-level backups beyond the included server backup — ₹0.0275/GB/day, which sounds trivial until you have three staging clones sitting on the same server and the backup volume drifts past 40 GB.
- The Elastic Email SMTP add-on, because transactional mail through the default server IP gets binned by Gmail roughly the day after you go live. ₹85/month for the first 1,000 emails, then more.
- Bandwidth overage at — and we'd have to check the exact rate, it changed sometime in 2023 — somewhere around ₹0.16/GB once you cross the included quota. This is the line that creeps. Nobody notices it until it's ₹3,000 of a ₹38,000 bill.
What you actually buy with that markup is the on-call rota you don't have to run. On 11 August 2024 a client's Cloudways-hosted store went read-only for 47 minutes during a Varnish purge after a WooCommerce Subscriptions plugin update — the kind of incident where the managed-hosting premium is supposed to earn its keep. It partially did. Their support team caught it inside 12 minutes, identified the stuck purge, and restarted the cache. We picked up the WhatsApp ping from the client at 23:47 IST, by which point the site was back. If that had been a self-managed Ubuntu 22.04 droplet, the ping would have come to us first, and somebody — probably one of us — would have been logging in from a phone hotspot in a Gurgaon flat trying to remember whether varnishadm wanted ban.list or ban.url at that hour.
So the Cloudways monthly, for this profile of store, lands somewhere in the ₹28,000–₹42,000 band depending on traffic, backup retention, and how many staging environments are lying around. Over three years, call it ₹12–14 lakh.
The self-managed stack, costed honestly in engineer-hours
Now the other side. Same store, same traffic, moved onto a self-managed DigitalOcean 8GB droplet in BLR1 with a managed database alongside — or, more often than not, MariaDB on the same box because the founder doesn't want to pay for the managed DB until the read load forces it.
The raw infrastructure is comfortable:
- DO 8GB droplet: ₹4,200/month.
- Snapshots and a weekly off-box backup to Spaces or S3: ₹400–₹600/month.
- Bunny CDN or BunnyCDN-equivalent for image and static delivery: ₹400–₹900/month for this traffic profile.
- Transactional email via Amazon SES or Postmark: ₹200–₹600/month.
- Uptime monitoring, log aggregation (we use a small self-hosted Loki, but BetterStack at ₹0 to ₹1,500/month is fine): call it ₹500/month average.
That's around ₹6,000–₹7,500/month in actual money out the door. Against ₹35,000 on Cloudways for the same store, the spreadsheet says you save ₹27,000 a month. ₹3.2 lakh a year. Over three years, just shy of ₹10 lakh. Why would anyone not move?
Because the spreadsheet is lying by exactly the cost of a senior engineer.
For a WooCommerce store on Ubuntu, the work that actually has to happen, every month, looks like: patching the OS and the PHP-FPM pool (Ubuntu 22.04 with PHP 8.2 from Ondrej's PPA is the common shape), verifying that nightly mysqldump backups restored cleanly on a scratch droplet at least once that month, rotating the Let's Encrypt certs that should auto-renew but which broke last quarter because somebody added a CNAME for a marketing subdomain and forgot to extend the cert SAN list, watching New Relic or the equivalent for slow query patterns, updating the WAF rules (we usually run a basic ModSecurity OWASP CRS profile), and once a quarter doing the upgrade dance — kernel reboots, Redis 6 to 7, MariaDB minor versions, the PHP point releases.
In our experience that's 6–10 hours a month for a competent senior — somebody who can read an strace output without panicking. Fully loaded cost of that hour, in a Gurgaon or Bengaluru studio billing rate, is ₹2,500 to ₹4,000 depending on who's doing it. Take the middle: 8 hours at ₹3,200 = ₹25,600/month. Suddenly the saving against Cloudways is not ₹27,000 a month, it's ₹1,400.
And that's the calm months. It does not include the 03:14 IST night in February when unattended-upgrades on the same client's secondary droplet rebooted the box and MySQL refused to come back up, because we'd resized the droplet down from 16GB to 8GB three months earlier to save money and nobody had reset innodb_buffer_pool_size from 12G. MySQL OOM'd on start. The site was down for 90 minutes while one of us, half asleep, worked out why systemctl status mariadb was telling us the unit was active but the socket file wasn't appearing. That was, by itself, ₹15,000 of engineering time and a refund to the client for the SLA breach. Plus the relationship cost, which doesn't go on a spreadsheet.
The migration tax cuts both ways, by the way. A Cloudways→self-managed move for a moderately complex WooCommerce store — Redis object cache, two staging environments, half a dozen cron jobs running WP-CLI scripts for stock sync to Unicommerce or Vinculum, a Let's Encrypt setup that has to be rebuilt because Cloudways manages it internally — is roughly 12 to 20 hours of careful work. What breaks: the cron paths (Cloudways uses /home/master/applications/<app>/public_html, a self-managed Nginx setup usually puts things under /var/www), the wp-cli wrapper that assumed the Cloudways PHP binary path, the Redis socket vs TCP configuration (Cloudways defaults to socket, most self-managed installs default to TCP on 127.0.0.1:6379), and the certbot renewal hook that needs to know to reload Nginx and not Apache. Going back the other way is faster — 6 to 10 hours — but you lose all the customisations you'd built on the self-managed side, which is usually why people don't go back.
Where the crossover sits, and why most stores miss it by a year
If you put the three-year totals next to each other, the crossover lands somewhere between ₹35,000 and ₹40,000 of monthly Cloudways spend. Below that, the engineering hours required to run a self-managed equivalent cost more than you save. Above it, the markup over the underlying droplet starts to outweigh those hours — but only if your senior engineer is genuinely going to do the work. If they're going to do five hours instead of eight because they're busy on a different client, you lose the verified-backup discipline and you pay for it the first time a database corrupts.
The honest line we draw with clients: stores under ₹50 lakh annual GMV almost never recover the engineering hours. The hosting bill is too small, the traffic patterns are too forgiving, and the senior time is better spent on conversion work. Stores above ₹2 crore GMV almost always do, because the Cloudways bill at that traffic shape has crossed ₹45,000 and the engineering work is happening anyway as part of normal operations. The ₹50 lakh to ₹2 crore band is where every founder asks the question and where the answer is genuinely uncertain.
Most stores miss the crossover by about a year, in both directions. The ones who should move stay on Cloudways for a year longer than they should, because the bill is annoying but the migration weekend is scarier. The ones who shouldn't move yet leave too early, because their developer convinced them it'd be fine, and then the developer leaves for a product job in six months and the founder is left looking at an Ubuntu droplet they cannot SSH into.
Anyway. The thing we still cannot predict, on the first call with a new client, is which of them will actually use the engineering hours they save by staying on managed hosting, and which will simply pocket those hours as not my problem any more. The first group makes the Cloudways premium worth every rupee. The second group is paying ₹38,000 a month for a support ticket queue they only open twice a year. That variable — not the hosting bill, not the crossover line, not the markup percentage — is what actually decides whether the move pays off. If you have a cleaner heuristic for sorting clients into those two buckets before the migration weekend, send it over. We've been looking for one for about four years and haven't found it yet.
