Credits: the one unit in software that does not get cheaper in bulk
AI tools increasingly sell credits rather than capacity, and credits behave unlike every other unit we have measured. Storage gets substantially cheaper per terabyte as you climb; monitoring gets cheaper per monitor; credits stay almost exactly the same price. That changes what a tier choice is — a forecast of usage rather than a bet on value — and it sits alongside the smallest annual discount in our entire record.
A credit is a unit the vendor invents. It has no external meaning, its exchange rate into actual work is set by the seller, and it can be redefined without any price changing. That makes credit-priced products hard to compare with each other and easy to compare within themselves, which is what makes the one ladder in our record that publishes both a price and a credit allowance at every tier worth working through carefully.
On that ladder the entry tier includes seven hundred and fifty credits a month and the top tier forty-two and a half thousand. The price between those two points rises by nearly fifty-three times; the credit allowance rises by nearly fifty-seven. Divide one into the other and the cost per credit falls by about six per cent from bottom to top. That is the whole volume discount available on this product: six per cent, for a fifty-three-fold increase in commitment.
It is worth pausing on how unusual that is. Every other per-unit ladder in our records rewards scale substantially. Cloud storage on one vendor's plans costs less than half as much per terabyte at the top of the ladder as at the bottom. Uptime monitoring on another falls by a factor of more than two and a half per monitor. Those are real bulk discounts and they change the arithmetic of which tier to buy. Credits do not behave that way at all: the unit price is close to flat, so climbing the tiers buys you more credits rather than cheaper ones.
The practical consequence is that a credit tier is a forecast rather than a value judgement. With a genuine bulk discount there is a case for buying above your needs, because the unit rate improves and the surplus is cheap. With a flat unit rate that case disappears entirely — surplus credits cost the same per unit as the ones you use, so buying a larger tier than you need is simply paying for credits you will not spend. Estimate your monthly consumption honestly and buy the smallest tier that covers it, and let the vendor's upgrade path handle the rest.
The same ladder carries the smallest annual discount we have recorded anywhere. Three of its four tiers price a year at just under two per cent below twelve monthly payments, and the fourth prices a year at exactly twelve monthly payments — no discount whatsoever. For comparison, a common shape elsewhere is two months free, which is nearly seventeen per cent. A commitment of a full year for a saving under two per cent is a poor trade in most circumstances, and the top tier's zero makes the point plainly.
Credits also resist comparison between vendors in a way that ordinary units do not. A terabyte is a terabyte and a monitored endpoint is roughly a monitored endpoint, but one vendor's credit and another's are different quantities of different work, and no amount of care in reading the record can make them commensurable. Our records hold credit allowances for several products and the counts are real, but a table of credits per pound across vendors would be arithmetic on units that do not share a definition. We do not publish one, and you should distrust any that you find.
Two further shapes are worth recognising. Some vendors publish a credit allowance with no price attached to the tier at all, so the allowance is a feature-list entry rather than something you can cost. And on one page the struck-through figure beside the effective one carries the same 'billed annually' qualifier as the price it is struck through against — meaning the reduction is a promotional discount on the annual rate rather than a reward for choosing annual over monthly, with no monthly figure published anywhere to compare. A crossed-out number in a credit product is not necessarily telling you what you think.
So the checks for a credit-priced tool are these. Convert to cost per credit at every tier and see whether it actually falls; if it does not, buy the smallest tier that covers your usage. Read what a credit buys, in the vendor's own words, and expect that definition to be able to change. Treat any cross-vendor credit comparison as meaningless unless someone has done the work of translating both into a common task. And check the annual saving as a percentage rather than accepting that a year's commitment must be worth taking.
The cost per credit barely moves across a whole ladder
Price up nearly fifty-three times, credit allowance up nearly fifty-seven, cost per credit down about six per cent. That is the entire bulk discount on offer, and it is close enough to flat that buying above your needs has no economic case.
Every other unit we measure does get cheaper
Storage falls to under half its entry rate per terabyte at the top of one ladder; monitoring falls by a factor of more than two and a half per monitor. Credits are the exception, and the difference changes which tier you should buy.
A credit tier is a usage forecast, not a value decision
With a flat unit rate, surplus credits cost exactly what used ones cost, so a larger tier is simply prepayment for unspent units. Estimate consumption and buy the smallest tier that covers it.
It carries the smallest annual discount in our record
Just under two per cent on three tiers and exactly nothing on the fourth, against a common shape elsewhere of two months free, which is nearly seventeen per cent. A year's commitment for under two per cent is rarely worth making.
Credits are not comparable between vendors
One vendor's credit and another's buy different amounts of different work. Our records hold the counts because the counts are printed; we publish no cross-vendor credit comparison, because it would be arithmetic on units with no shared definition.
A struck-through credit price may not be an annual saving
On one page both the crossed-out figure and the effective one carry the same 'billed annually' qualifier, so the reduction is promotional rather than a reward for annual billing — and no monthly figure exists anywhere to compare it with.
Questions people actually search
How do I work out whether a credit plan is good value?
Divide each tier's price by its credit allowance and compare the results. If the cost per credit falls meaningfully as you climb, there is a case for buying ahead of your needs. If it is flat, as it is on the ladder examined here, buy the smallest tier that covers your expected usage, because surplus credits cost the same as used ones.
Can I compare credits between two vendors?
Not directly, and we would treat any table that claims to with suspicion. A credit is defined by the vendor selling it, and two vendors' credits buy different amounts of different work. The only honest comparison is to price the same real task at both — a minute of rendered video, a generated image at a given size — which is work no pricing page does for you.
Why do credit prices not fall in bulk when storage does?
We can only report that they do not on the ladders we have read, rather than explain the vendors' reasoning. One plausible reading is that the underlying cost of serving a credit is largely variable — compute spent per generation — where storage and monitoring carry heavy fixed costs that spread across larger plans. That is an inference and we present it as one.
Is a two per cent annual discount ever worth taking?
Rarely, and it is worth doing the comparison explicitly. A year's commitment costs you flexibility that is worth something in a fast-moving category, and under two per cent is a thin return for it — particularly where the top tier of the same ladder offers exactly nothing for the same commitment.
Do you earn a commission from any vendor discussed here?
No. We hold no affiliate relationship with any of the credit-priced vendors referred to on this page, and this page carries no affiliate links. Our method is on the method page.
This page publishes no absolute prices. Every quantity in it is a count, a ratio or a percentage computed from figures the vendors print, because the credit-priced products we hold were each read in a single market while this guide appears in three editions, and several hold their prices in fields our pages cannot resolve into figures at all. Records were read on 2026-09-03. Credit allowances are quoted as the vendors print them; the cost-per-credit comparisons and the annual percentages are our arithmetic on their published pairs. Comparisons between storage, monitoring and credit ladders are drawn from separate vendors in separate categories and are about the shape of each ladder rather than about relative value. This page carries no affiliate links. Related: the price is not the bill, what free actually means and the annual discount you cannot check. The method →
Found a price that no longer matches the vendor’s page? Tell us — corrections are dated and stay on the page.