Say what is in stock: GPU availability on a pricing page
A rate for a GPU nobody can rent is a rate nobody can pay. An availability line beside the price tells buyers whether the card exists for them today.
A price is not a promise of a card
A GPU pricing page should say, next to each rate, how a buyer gets that card: available now on demand, available on request, waitlist, or reserved capacity only. A listed rate tells a buyer what they would pay. It does not tell them whether they can pay it today, and a buyer who signs up for a card that is never free blames the price page for the wasted afternoon.
We should be plain about our side of this. This site records listed rates and the page each was read from. It does not measure availability, and nothing on it says which provider has cards free.
Why buyers ask once the price looks right
The sequence is easy to picture. A buyer finds a rate that fits the budget, then tries to start an instance, and a refusal at that point can end the evaluation. A card shown at an attractive rate and then refused at launch is worse for the provider than a card not shown at all, because the buyer has spent effort and has a story to tell.
The question is sharpest for the newest cards. The count of sellers with a listed B200 rate is 22, and anyone looking at those rates will want to know whether the rate is for a card they can start now or for a future allocation. A listed rate cannot say, which is why the page has to.
The same holds for the GB200 NVL72, another Blackwell product. Where a rate for one is listed, the page should say how one is actually obtained: a form, a call or a contract.
What an availability line needs to say
The mode of access first. "On demand, self-serve" means a buyer can start an instance from the console. "On request" means they must talk to sales. "Waitlist" means they cannot have it yet. "Reserved only" means the rate applies to a term and there is no on-demand pool. Each is a fair offer; each is a different promise.
Then the size. A buyer who needs one card and a buyer who needs a full node of eight are asking different questions, and a provider may honestly answer yes to the first and no to the second. Say the largest configuration a buyer can start without speaking to anyone.
Quota is part of the same answer. Where a new account is limited to a small number of cards until it is reviewed, the buyer needs that before they plan a run, not after their first request is refused.
Then the date the line was last checked. Availability changes faster than rates, and a line that was true last quarter is the one that causes the bad afternoon.
What not to write
Words like "limited" or "high demand" with no detail. A buyer cannot plan around either, and both read as a way of saying no without saying it. If capacity is short, saying "on request for more than one node" is more useful and less likely to lose the sale.
A live counter that is not live. If a page shows a number of free cards, the number must come from the system that allocates them. A static figure dressed as a live one, once caught, puts the rest of the page in doubt, including the price.
And do not hide the rate for a card on request. Buyers comparing the 51 cards with listed on-demand rates on the GPU index will compare yours too if it is published, and a sales conversation that starts with a known rate is shorter than one that starts with a question.
Where this sits next to the rest of the page
Availability belongs on the same row as the card and the rate, for the same reason the commitment label does: a rate copied into a spreadsheet takes its row with it and nothing else. A separate status page linked from the footer will not be read by a buyer comparing prices.
A claimed listing on this site lets a vendor answer the questions asked of its category with a source, which is the route for a provider that wants buyers here to see how its capacity is offered. It does not turn availability into a measured figure, and we will not present it as one.
Older cards raise the opposite question
For the newest cards the buyer asks whether the card exists yet. For older ones, such as the V100 with 16 GB or the T4 with 16 GB, the question is whether it still does. A rate for an older card can stay on a pricing page after the last machine was retired, and a buyer who finds the rate and then cannot start the instance has met the same failure from the other end.
The count of sellers with a listed V100 rate is 4, and for the T4 it is 4. Each is a published rate read from a page, and none says whether the card can be started today. A provider that has stopped offering a card should remove the row. If it is offered only on request, the row should say so.
A rate that disappears from the source page is not shown here once it is more than 21 days old. That keeps stale rates out of the comparison, but only the provider can keep a live rate for a card that cannot be had off its own page.
What we could not answer
- ⚑We do not measure availability, so this cannot say how often a listed GPU rate refers to a card buyers cannot start.
- ⚑The access modes named here are ones provider pages commonly use; we hold no count of how many providers use each.
What this rests on
infratoapp, “Say what is in stock: GPU availability on a pricing page”, read 27 October 2026. cards with a listed on-demand rate: 51. https://www.infratoapp.com/blog/gpu-availability-on-your-pricing-page
Questions people ask about this
Does a listed GPU price mean the GPU is available?
No. A listed rate says what a provider charges for a card, not whether one is free to start today. Many cards are offered on request, on a waitlist or only as reserved capacity, and a buyer should check the access mode before planning around the rate.
What should a GPU availability line on a pricing page include?
The mode of access, such as self-serve on demand, on request, waitlist or reserved only; the largest configuration a buyer can start without speaking to sales; and the date the line was last checked. Vague words like limited or high demand give a buyer nothing to plan with.
Does infratoapp track GPU availability?
No. The site records listed GPU rates with the page and date each was read from. It does not measure whether any card is in stock, and nothing it publishes says which provider has capacity free.
These render as FAQPage structured data · answers are filled from the same live figures
Read next
Before choosing an interruptible GPU a buyer needs the reclaim warning, what happens to the job and how the rate is set, not just the size of the discount.
Buyers who have paid to move data out of a cloud ask about egress first. A GPU page that is silent on it is read as hiding a charge, whether or not one exists.
A GPU pricing page buyers can compare names the card exactly, quotes per GPU per hour, labels the commitment and states egress, regions and access mode.