Per-Second Billing Explained: Why Your Cloud Bill Shouldn't Be a Surprise

"The scariest email in a developer's inbox isn't an outage alert. It's an invoice."

Every few weeks a new horror story tops Hacker News: a side project racks up a five-figure cloud bill overnight — a runaway function, a DDoS'd endpoint, an egress fee nobody understood. The replies are always the same: "set up billing alerts," "you misconfigured it," "read the pricing docs." All of which quietly concede the point — on most clouds, the default outcome of a mistake is a bill.

How cloud billing became adversarial

Traditional cloud pricing has three traps built in. Coarse time units: pay for the hour or the month even if you used minutes. Unbounded post-paid spend: the meter runs first, the invoice arrives later, and there is no ceiling unless you build one. Invisible line items: egress, requests, IOPS, NAT-gateway hours — costs you can't see in the moment you incur them. None of this is accidental; unpredictability is profitable.

What per-second billing actually means

On Abasthan, every plan has a monthly price — say $7/month for a Standard web service — but that price is just a rate: it's divided down to a per-second cost, and you're charged only for seconds your service is actually running. Run a demo for two days? You pay for two days. Stop a worker at night? The meter stops with it. A cron job that runs 40 seconds a day costs… 40 seconds a day.

There's no rounding up to the hour, no "you provisioned it, you pay for the month," and no billing categories you need a certification to understand: the price on the pricing page is the whole price.

Prepaid credits: a hard ceiling on the worst case

The second half of predictability is prepayment. You top up a credit balance; usage draws it down per second. When credits run low, services pause — and resume automatically when you top up. That flips the default outcome of a mistake: instead of an unbounded invoice, the worst case is a paused app and a number you chose in advance.

For teams, that turns cloud spend into a budget instead of a forecast. For students and side projects, it means experimenting is safe: you cannot spend money you didn't load. And your first app doesn't even need credits — it's free.

Why most platforms can't offer this

A platform that rents its compute by the hour from a hyperscaler can't honestly bill you by the second — the mismatch comes out of its margin. Abasthan runs on bare-metal hardware we own, so our unit economics are steel and electricity, not someone else's price list. That's what makes per-second rates, prepaid ceilings and a real free tier possible at $1.5–$17/month prices.

A quick cost comparison

• Hobby API, running 24/7: Starter web service at $2/month. That's the whole bill — SSL, subdomain, deploys included.
• Demo for a client, up 3 days: Standard plan at $7/month ≈ $0.70 for the three days, then stop it.
• Nightly batch job, 10 minutes/day: Cron Job plan billed for ~5 hours a month of actual runtime — cents, not dollars.

FAQ

What happens the moment credits hit zero?

Services are suspended, not deleted — data and configuration stay intact. Top up and everything resumes automatically within minutes.

Can I set up automatic top-ups?

Yes — you can securely save a card and charge it from the billing page in one click when credits run low.

Are there egress or request fees?

No per-request or bandwidth line items on standard usage. The plan price is the price.