Prepaid, subscription or pay-as-you-go: where each billing model wins

Tutorials Published 2026-09-22 Updated 2026-09-27 72 views API Compass
The three billing models on offer look like a price question. They are really a cash-flow and risk question, and the right answer changes with how predictable your usage is.

The three models

ModelHow it worksTypical offer
Prepaid creditYou buy balance up front, it is consumed as you callA rate discount for larger top-ups
SubscriptionA fixed monthly fee for a quota or a capA rate discount, sometimes unlimited within limits
Pay-as-you-goNo upfront balance; billed against usage at period endHigher unit rate

At relay services, prepaid is by far the most common, subscription exists at the larger operators, and pay-as-you-go is rare because it requires the operator to carry your credit risk.

The real difference is who holds the money

This is the frame that makes the decision obvious.

  • Prepaid means you lend the operator money, interest-free and unsecured, and draw it down over time.
  • A subscription is the same arrangement with a fixed term, so the operator knows the money is there.
  • Pay-as-you-go means the operator carries your risk instead of you carrying theirs.

Everything else — discount sizes, quotas, expiry rules — is a variation on that theme. So the question is not "which is cheapest", it is "how much of my money do I want sitting on somebody else's balance sheet, and what am I paid for leaving it there".

Where prepaid wins

When your usage is highly predictable. If you know you spend roughly $200 a month, prepaying $200 is nearly free: you get the discount and the money would have been spent anyway.

When the discount is real and the operator is established. A 15% rate discount for an annual prepayment is a reasonable return if the risk is genuinely low.

When you need the low barrier. No credit check, no contract, payment in CNY. For individual developers and small teams this is often the whole attraction.

Where prepaid loses

When usage is unpredictable. Money sitting as unused balance is money not doing anything else. If you prepay a year and use 40% of it, the effective discount is much smaller than the headline.

When the operator's longevity is uncertain. This is the one that matters. Prepayment converts the operator's business risk into your loss. A 30% discount does not compensate for the possibility of losing the remaining balance plus the work of migrating.

When you have several providers. Prepaying at two or three services multiplies the dormant balance across accounts.

The practical rule: prepay roughly one month of usage, not one year. You give up some discount and you keep the ability to leave.

Where subscription wins

When the quota realistically matches your usage. A subscription is a good deal if you would consume the quota anyway, and a bad one if you buy headroom you do not use.

When the operator is large and stable enough to be worth the commitment. A subscription is only better than prepaid if the counterparty risk is genuinely lower, which usually means a bigger, more established operator.

Watch for quota cliffs. Some subscriptions charge a much higher per-unit rate once you exceed the quota, or stop entirely. Estimate your peak month, not your average one.

Where pay-as-you-go wins

When you cannot forecast usage at all. A new product, a pilot, an unpredictable workload. Paying a higher unit rate is a reasonable premium for not tying up capital.

When you have compliance constraints on prepayment. Some finance teams will not approve large prepaid balances to a small supplier. Pay-as-you-go solves that.

When you are testing a provider. A month of pay-as-you-go gives you real data on quality and stability without committing balance. There is usually no better way to evaluate a service.

A decision rule

Ask two questions:

  1. How predictable is my monthly usage? Within 20%, or genuinely unknown?
  2. How would I feel about losing the balance? Annoyed, or materially harmed?
Predictable usageUnpredictable usage
Loss is an annoyancePrepay one to three months for the discountPay-as-you-go if offered; otherwise prepay monthly
Loss would hurtPrepay one month, keep a second provider fundedPay-as-you-go, or prepay the minimum

The combinations in the bottom-right are where people get into trouble: unpredictable usage plus a prepayment they cannot afford to lose is exactly the configuration that a service collapse turns into a real problem.

The one rule that survives all of this

Never prepay more than you would be willing to write off.

Not because any particular operator is going to collapse, but because the discount you are being offered is compensation for exactly that risk. If the discount is 30% and the risk is 5%, it is a good trade. If the discount is 30% and you cannot afford the 5%, it is not a trade you should be making regardless of the maths.

Related reading

  • Six signals that a relay service is about to disappear (guides)
  • Corporate procurement of a relay service: invoices, contracts, compliance and risk control (guides)
  • From list price to your invoice: a worked cost conversion (guides)

Was this useful? If you have run into something the article does not cover, tell us and we will verify it and update the text.

Related articles

Video generation behind a relay: three API shapes, four billing traps, one acceptance checklist

2026-10-19

MCP through a relay: what works, what never will, and how to test it in ten minutes

2026-10-19

A relay selection decision tree: working back from your scenario to a site

2026-10-07

Logging and reconciliation: attributing token spend to the project level

2026-10-06

Migrating suppliers gradually: replacing an API site without interrupting service

2026-10-06

Build your own probe: twenty lines of script to monitor your API suppliers

2026-10-06

Read on: a few of the services we list

This article does not point at a specific model family, so here are a few of the highest-scoring services we list, in case you want to read on. They are ordered by score and uptime and are not a recommendation.

See the full list in the directory, or browse by model family.