What It Really Costs to Run a SaaS in 2026 (My Actual Monthly Bill)

The $90 Invoice for Nothing

There’s a specific kind of dread I want to describe, because if you’ve felt it you’ll know it instantly, and if you haven’t, this whole post is me trying to save you from it.

It’s the first of the month. Your card gets charged. You open your banking app — not to check a balance, just out of habit — and there it is: a little cluster of small charges that add up to about ninety dollars. Hosting. A database. An email tool. A domain. A “pro” plan for something you turned on once and forgot. Ninety dollars, gone, for a SaaS that made exactly zero dollars that month.

It wasn’t the ninety dollars — I could afford ninety dollars. It was what it meant: I was paying rent on a product nobody was using. The product was Clickly, my URL shortener, and every month on the first my card statement may as well have read Clickly — still nobody. I paid that invoice for the better part of a year before I got angry enough to actually take it apart.

So this is the honest version. Not an estimate, not a “here’s what it could cost” — my actual monthly bill for running a live SaaS in 2026, line by line, plus the dumb version I ran first. Let’s open the books.

---

My Actual Monthly Bill, Line by Line

Here’s what it costs me to keep a small, live, commercial product online today — deployed, payments switched on, a trickle of real users, the works:

Line item	Tool	What I pay
Frontend + light backend	Vercel Pro	$20/mo
Database	Neon (free tier)	$0
CDN + storage	Bunny	~$1/mo
Transactional email	ZeptoMail	~$2.50 per 10k
Marketing email	SendPulse (free)	$0
Analytics	PostHog + Vercel Analytics	$0
Error monitoring	Sentry (free)	$0
Business email inbox	Zoho Mail (free)	$0
Domain	~$12/year	~$1/mo
Payments	Paddle / Dodo	$0/mo*

Total: roughly $25 a month. And most of that is one line — the hosting.

*The asterisk on payments matters. A Merchant of Record like Paddle or Dodo charges you nothing monthly; they take a cut when you make a sale (Paddle is 5% + 50¢ a transaction). So at $0 MRR, payments cost you $0. The day you make money, they take their slice of that money. That’s the only fee in this entire stack that scales with your success instead of with your anxiety, which is exactly how it should be. (Which MoR to actually pick — and why the Stripe-owned option gets tricky if you’re in India — is its own rabbit hole: Paddle vs Lemon Squeezy, explained. And if you’re building from India specifically, here’s every payment option that actually works.) One line you’ll be tempted to add but almost certainly shouldn’t: a separate invoicing tool — your Merchant of Record already generates tax invoices for you, so don’t pay for invoice automation you already own.

Look at that table again. Eight of the ten lines are zero. The database that would’ve cost me $20-something on a managed provider is free on Neon until I actually have load. Analytics that companies pay hundreds for is free on PostHog up to a million events a month. The whole “SaaS is expensive to run” belief is mostly a story we tell ourselves before we’ve looked at the free tiers.

---

How I Used to Pay $90 for This

So where did the ninety dollars go? Clickly’s bill was a monument to my own paranoia:

• Vercel Pro, $20 — fine, I’ll get to why this one’s actually justified.
• A time-series analytics database, ~$30 — this was the crime. Clickly is a URL shortener, so I decided I needed a serious time-series database to handle “millions of click events.” Clickly was logging maybe forty clicks a week, most of them me testing redirects on three browsers. I was paying thirty dollars a month to analyze my own browsing. (That entire over-engineering bender — Timescale, ClickHouse, the lot — became its own database guide.)
• Mailchimp, ~$20 — for a newsletter I sent twice.
• Google Workspace, ~$8 — for one email address I checked once a week.
• A grab-bag of $5–$10 tools — an uptime monitor, a scheduling thing. Each one felt small. Together they were a car payment.

If you want the single dumbest line, it’s that database: real money, every month, to store data that only existed because I kept refreshing my own short links. I didn’t overpay because I was careless. I overpaid because I was scared, and because I didn’t know the free, equally-good option was sitting one search away. Knowledge is the discount.

---

Where Founders Quietly Overpay

If I audit my old bill against what I know now, the leaks are almost always the same four:

The managed database. You do not need a $20+ production Postgres plan on day one. Neon’s free tier will carry you for a genuinely long time — long enough that I run seven separate databases on it for $0 — and it scales to a pay-as-you-go model with no monthly minimum when you outgrow it — you pay for compute you actually use, not for a plan you might need. (If you’re stuck choosing between Neon, Supabase, and “Vercel Postgres,” I put all three head-to-head in the cheapest serverless Postgres comparison — spoiler: one of them isn’t real anymore. And I went deep on the whole database rabbit hole in the advanced database guide, including when you finally do need more than plain Postgres.)

Business email. Everyone reflexively buys Google Workspace at $8.40 a user just to get you@yourbrand.com. But Zoho Mail has a genuinely-free plan for up to five users on your own domain, and their paid tier is a dollar a month if you need the extras. That’s ~$100 a year saved on literally one email address. (And no, don’t try to save that dollar by self-hosting your own mail server — I did, and it cost me a weekend to save nothing.)

Marketing email. Mailchimp quietly gutted its free plan in 2026 down to 250 contacts and 500 sends a month — you’ll blow past that before you have an audience. SendPulse gives you a free 15,000 emails a month and scales by subscriber count. (The full transactional-vs-marketing breakdown is in the complete email guide.)

Analytics and monitoring. People pay for these before they have traffic. Don’t. PostHog is free to a million events a month, Vercel Analytics is free to 50,000, and Sentry’s free tier catches plenty of errors for a small project. Turn them on, pay nothing, revisit when you have real volume. (The full setup — and why self-hosting PostHog to save nothing is a trap — is the $0 analytics stack.) And if a friend tells you to “just use Datadog,” read why Datadog’s pricing eats broke founders alive first — observability doesn’t need to be a line on this bill at all.

My take: every one of those is a case of buying the plan you’re scared you’ll need instead of the one you actually use. Bootstrap the fear out of it. Pay when it hurts not to.

---

The Vercel Hobby Trap

Now the one line I don’t cut, and the gotcha nobody warns you about.

Vercel’s free Hobby plan is glorious. It’s also, per their own terms, for non-commercial personal use only. The moment your project is a real business — you’re charging money, it’s a company thing — you’re supposed to be on Pro at $20 a user a month. A lot of founders run a commercial SaaS on Hobby and don’t realize they’re offside until traffic spikes and Vercel comes knocking. (And the day you are on Pro, set a spend cap so a spike bills you a pause instead of a four-figure invoice.)

So budget the $20. Or — and this is the move I actually make — put the heavy, always-on parts somewhere with a flat fee. A small Railway box is $5 a month and won’t surprise-bill you when a cron job goes wild; Koyeb gives you a real always-on instance for a predictable monthly fee. Frontend on Vercel, the expensive workloads on a flat-fee host. (I ranked every option in the deployment tier list, broke down when to switch between Vercel, Railway, and Fly, and wired the whole split together — caps and all — in the full deployment guide.)

Hosting is the one place the “everything’s free” fantasy breaks. Accept it, pick the cheapest legitimate option, and move on. (And it makes zero difference to this bill whether you built the frontend in Next.js or Astro — both ride the same free tier.)

---

Getting to (Near) $0

If you’re pre-launch — a side project, no paying customers, no commercial use yet — you can run the whole thing for basically nothing:

• Frontend on Vercel Hobby (fine while it’s genuinely a personal project) — $0
• Database on Neon free — $0
• Bunny for any assets — the $1 minimum
• Email on ZeptoMail’s free signup credit and SendPulse’s free tier — $0
• Analytics, monitoring, inbox — all free

That’s a dollar or two a month for a real, deployed, working product. The costs only switch on when reality switches on: commercial use flips Vercel to $20, real load eventually flips Neon to pay-as-you-go, real email volume costs you a couple of dollars per ten thousand sends. Every one of those is a cost that arrives because something good happened — you launched, you got traffic, you’re sending real mail to real people.

People always ask what the bill looks like at 100 or 1,000 users. My honest answer: I mostly can’t tell you from my own receipts, because I rarely kept a product alive long enough to find out — and that’s a distribution problem, not a cost one (a whole other war). But the shape is knowable. Neon and PostHog stay free for absurdly long. Email creeps up a few dollars per ten thousand sends. Hosting is the line that grows first — and the day it starts to hurt is the day you’ve finally earned the right to care about it, because by then people are paying you.

That’s the mental reframe that fixed me. A cost that scales with your success is an investment; a cost you pay at $0 MRR is just a tax on your own hope. Get everything you can into the first bucket. (If “MRR” is a new acronym, here’s what it means and why it’s the number that decides everything.)

---

The Bottom Line

The honest answer to “what does it cost to run a SaaS in 2026” is: almost nothing until you have users, and about $25 a month once it’s a real, live product. Not the hundreds you’re bracing for. The tools got cheap. The free tiers got generous. The only thing that stayed expensive is not knowing they exist.

But here’s the part I actually want you to keep. The ninety-dollar invoice never hurt because of the ninety dollars. It hurt because of the story it told me every single month. Cut that bill to twenty-five, and you don’t just save sixty-five dollars — you take away the monthly reminder that nobody’s using your thing yet, and you buy yourself the one resource a broke founder actually runs out of: the will to keep going. (For the whole survival playbook this fits into, start with the broke solopreneur’s guide.)

This is the Broken Engineer Guide — I optimized my infra bill down to a rounding error while my revenue stayed a rounding error too. Learn from the first part. Don’t repeat the second. Now go make something worth paying for.