How to value unused cloud credits
There is no index price for cloud credits. Here are the factors that actually move what a buyer will pay, and how to reason about your own balance.
There is no public index price for unused cloud credits, no ticker, and no reliable rule of thumb. Any site that quotes you a fixed percentage before seeing your listing is guessing — and a guess that sounds precise is worse than no number at all, because you will anchor on it.
What does exist is a consistent set of factors that determine what a buyer will pay. This guide walks through them so you can form your own view before anyone makes you an offer.
Start from the buyer's position
The only reason a buyer pays for your credits is to reduce what they spend on compute they were going to buy anyway. That gives you the whole framework. A buyer is asking:
- Can I actually apply this to spend I already have?
- Can I consume it before it expires?
- What is the chance the provider voids it and I lose the money?
- How much work is this arrangement going to be?
Every factor below is one of those four questions in a different costume. If you can answer all four well, your balance is worth something. If you cannot answer any of them, the discount will be brutal or there will be no deal.
The factors that move the number
| Factor | Pushes value up | Pushes value down |
|---|---|---|
| Time to expiry | 12+ months, enough to migrate and consume | Under 90 days |
| Service scope | Unrestricted, applies to general usage | Single service, SKU or region |
| Exclusions | Covers Marketplace, commitments, licences | Excludes exactly what the buyer spends on |
| Transfer route | A documented, provider-sanctioned arrangement | An improvised handover of credentials |
| Balance size | Large enough to matter to a real buyer | Too small to justify the paperwork |
| Account cleanliness | Isolated account, nothing else entangled | Production data, other customers, shared IAM |
| Usage tier (AI APIs) | Tier supports the buyer's throughput | Low limits that throttle the buyer's workload |
| Evidence | Clear billing records and grant documentation | A single screenshot and an assertion |
Two of these are underrated. Account cleanliness decides whether an account transfer is even possible — a balance sitting on the account that also runs your production database is not a sellable asset, it is a hostage. And on OpenAI or Anthropic, usage tier can matter more than the balance, because a buyer who cannot get their throughput has no use for cheap tokens.
The factor that dominates everything
Transferability. Not "is it technically possible to give someone access" but "is there a route your agreement with the provider permits".
This dominates because it sets the risk of total loss. If the provider can void the balance on discovering the arrangement, the buyer is not buying credits at a discount, they are buying a lottery ticket, and they will price it that way. Most cloud and AI providers restrict transfer and resale of promotional credits, and some prohibit it outright — which is why the honest answer for many balances is that they are not sellable at any price.
Working out where your specific credit sits is the first thing to do, and it is provider-specific: see the pages for AWS, Azure, OpenAI, Anthropic and Google Cloud.
How to think about your own balance
Work through it in this order:
- Identify the programme. Not the platform — the programme. An Activate grant, a Founders Hub grant and a Visual Studio credit are three different assets on two clouds.
- Read expiry per credit line. Aggregate figures hide short-dated tranches. See when credits expire for where to look on each platform.
- Write down the scope. Services, SKUs, regions, and anything excluded.
- Assess the account. Could it be handed over cleanly, or is it entangled?
- Read your agreement. Before you talk to anyone, so you are not relying on a counterparty's characterisation of your own contract.
- Only then think about price. A number formed before steps one to five is not a valuation, it is a wish.
Why we do not publish a percentage
We get asked for one constantly, and the honest position is that we would be making it up. The spread between a clean, long-dated, unrestricted balance and a short-dated, service-locked one is enormous, and averaging across them produces a figure that misleads everyone it reaches.
What we will do is tell you what buyers are currently saying about your platform and your specific situation once we have reviewed the listing — including when the answer is that we do not think it can be sold. That is less satisfying than a number on a landing page, and considerably more useful.
Related reading
- Are AWS credits transferable? — the constraint that dominates value
- When do cloud and AI credits expire? — reading the date correctly
- A buyer's risk checklist — the other side of the table
- How we verify listings — what a review establishes, and what it cannot