- Sheet
- P-01
- Title
- Rate schedule
- Checked
- 2 Sep 2026
- Basis
- per GPU-hour
NVIDIA B300 rental pricing, sourced and dated
Every published B300 rate we could find, what it costs per GPU-hour, and whether we could confirm the capacity was actually available. Under the lowest rate we could confirm as bookable — the figure is set on the call. Last checked 2 September 2026.
- Lowest verified in stock
- $7.89
- per GPU-hour
- Median across tracked providers
- $7.85–7.87
- per GPU-hour
- Our on-demand quote
- Under the lowest rate we could confirm as bookable
- Quoted on the call · per GPU-hour
We are not claiming to be the cheapest
Listings at $6.50–7.50 exist and are in the table below. We could not book capacity at any of them. $7.89 is the lowest rate we could confirm as actually orderable, and that is the comparison we think is honest — so it is the one we hold our own quote against.
P-01.1 — Chart
Published rates, on one scale
Datum = lowest bookable $7.89
Under the lowest rate we could confirm as bookable · Quoted on the call
- 01
Neocloud listings, low end
Unverified
$6.50–7.50
0.82×
- 02
Lowest verified in-stock listing
Verified
$7.89
datum
- 03
Median on-demand
Aggregate
$7.85–7.87
0.99×
- 04
Oracle
Rate card
$15.00
1.90×
- 05
AWS on-demand
Rate card
$17.80
2.3×
The datum is the lowest rate we could confirm as in stock, so multiples are against something a buyer could actually have bought. Hatched columns are published prices we could not confirm — including the one under the datum. The shaded region under the datum is where our own quote falls; it is a band and not a line because the figure depends on the job. Ranges are shown at their top value with the range hatched beneath. All rates read 2 Sep 2026.
P-01.2 — Schedule
The table, in full
| Provider | USD / GPU-hour | vs. datum | Term | Status | Source | Checked | |||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Neocloud listings, low endMultiple providersThe cheapest published B300 numbers we found anywhere. We could not book capacity at any of them — out of stock, unconfirmable, or a sales process that never produced a slot. They are in the table because leaving them out would flatter us. | Rate$6.50–7.50 | vs. datum0.82× | TermOn-demand | Unverified / no stock | SourceOur price survey — unverified listings | Checked2 Sep 2026 | The cheapest published B300 numbers we found anywhere. We could not book capacity at any of them — out of stock, unconfirmable, or a sales process that never produced a slot. They are in the table because leaving them out would flatter us. | ||||||
| Lowest verified in-stock listingConfirmed orderableThe cheapest B300-class capacity we could confirm as actually available to buy. This is the benchmark the rest of this page is measured against, because it is the real alternative to us. | Rate$7.89 | vs. datum— | TermOn-demand | Verified in stock | SourceOur price survey — in-stock check | Checked3 Sep 2026 | The cheapest B300-class capacity we could confirm as actually available to buy. This is the benchmark the rest of this page is measured against, because it is the real alternative to us. | ||||||
| Median on-demandAcross providers we trackMidpoint of tracked published rates. Half of what we saw was higher. | Rate$7.85–7.87 | vs. datum0.99× | TermOn-demand | Aggregate of tracked listings | SourceOur price survey — median | Checked2 Sep 2026 | Midpoint of tracked published rates. Half of what we saw was higher. | ||||||
| OracleOracle Cloud Infrastructure | Rate$15.00 | vs. datum1.90× | TermOn-demand | Published rate card | SourceOracle Cloud price list ↗ | Checked2 Sep 2026 | |||||||
| AWS on-demandAmazon EC2 | Rate$17.80 | vs. datum2.3× | TermOn-demand | Published rate card | SourceAWS EC2 on-demand pricing ↗ | Checked2 Sep 2026 | |||||||
| AWS Capacity BlocksAmazon EC2, same B300 instanceCheaper than AWS on-demand above, and the lowest hyperscaler rate here. You book the block ahead and pay it up front, which is the lead time this page is about. | Rate$14.04 | vs. datum1.78× | TermReserved block | Published rate card | SourceAWS EC2 Capacity Blocks pricing ↗ | Checked3 Sep 2026 | Cheaper than AWS on-demand above, and the lowest hyperscaler rate here. You book the block ahead and pay it up front, which is the lead time this page is about. | ||||||
| Reserved / committedAcross providers we trackLong commitments buy a lower rate everywhere, including here. This is the market range, not a rate card of ours. Interruptible spot capacity is cheaper still — we have seen B300 spot from about $4.30 — but spot can be reclaimed mid-run, which is a different product from guaranteed capacity and is not what this row prices. | Rate$4.25–5.62 | vs. datum0.54× | Term24–60 month | Aggregate of tracked listings | SourceOur price survey — committed terms | Checked2 Sep 2026 | Long commitments buy a lower rate everywhere, including here. This is the market range, not a rate card of ours. Interruptible spot capacity is cheaper still — we have seen B300 spot from about $4.30 — but spot can be reclaimed mid-run, which is a different product from guaranteed capacity and is not what this row prices. | ||||||
Where we sit
We do not print a per-GPU-hour figure on this site. Capacity this small gets priced against the job — how many GPUs, for how long, on what term — and a headline number would be wrong for almost everyone who read it. What we will commit to in public is the position: our on-demand quote comes in under the lowest rate on this page that anyone could confirm was actually in stock, and committed terms price below that again.
Multiples in the “vs. datum” column are the published rate divided by $7.89 — the lowest rate in this table we could confirm as in stock — computed at render. Where a row is a range, the multiple uses the low end, which is the comparison least favourable to the datum.
Cheaper listings exist. Bookable ones mostly do not.
There are published B300 rates in the $6.50–6.95 range. We went looking for them. What we found was capacity that was out of stock, listings that could not be confirmed, or a sales process that never produced a slot you could actually take. The lowest rate we could verify as in stock was $7.89, and that is the number we hold ourselves against.
So the argument is not that some number of ours is the smallest on the internet. It is that the alternative you can actually buy today costs what the table below says it costs, it takes weeks to get, and we come in under it on a date you can hold us to. Every rate below carries where we got it and when we checked, which is not a courtesy — it is the only way you can tell the difference.
We are taking inquiries on 1–5 year leases.
On-demand is quoted on the call. A lease is the other conversation: a named term, a named rate, and capacity that does not get re-priced when the market moves. One year is a lease. Five years is a lease. Anything in between is a lease. We quote it against the job rather than against a published ladder, because a two-GPU evaluation and four nodes for five years are not the same purchase.
An inquiry is not a signed term. Rate, start date and remedies are agreed on the call and written into the contract. Nothing on this page is a published lease rate.
P-01.3 — Load calculation
Price the job at the market
Rate × GPUs × hours · no other terms
Published rates × your GPU count × your hours. This is what the alternatives cost for the run you have in mind, with no assumptions of ours in the number — change any input. Ours is the one line that is not here, because it is set on the call.
One row per node · capped at 48 because that is the fleet
730 h ≈ one month running continuously
$7.89 × 8 × 730 h
$46,078
The lowest total anyone could actually book on the day we checked
Quoted on the call
Ours comes in under that line. The figure is set against your job, on the call.
Straight multiplication of the published rates above. Not a quote, and it excludes anything a provider bills separately.
- Neocloud listings, low end$37,960–$43,800-$8,118
$6.50–7.50 on-demand · Our price survey — unverified listings · 2 Sep 2026
- Lowest verified in-stock listing$46,078
$7.89 on-demand · Our price survey — in-stock check · 3 Sep 2026
- Median on-demand$45,844–$45,961-$234
$7.85–7.87 on-demand · Our price survey — median · 2 Sep 2026
- Oracle$87,600+$41,522
$15.00 on-demand · Oracle Cloud price list · 2 Sep 2026
- AWS on-demand$103,952+$57,874
$17.80 on-demand · AWS EC2 on-demand pricing · 2 Sep 2026
- AWS Capacity Blocks$81,994+$35,916
$14.04 reserved block · AWS EC2 Capacity Blocks pricing · 3 Sep 2026
- Reserved / committed$24,820–$32,821-$21,258
$4.25–5.62 24–60 month · Our price survey — committed terms · 2 Sep 2026
Baseline for the difference column is $46,078 — the same job at the lowest rate anyone could confirm as in stock
Methodology
How this table was built
- 01What counts as a rate
- Published per-GPU-hour pricing for B300-class capacity. Most rows are on-demand; where a row is a reserved product the term column says so, because reserving a block ahead and taking capacity now are not the same purchase. Where a provider prices by instance, we divide by the number of GPUs in that instance and say so.
- 02What counts as verified
- We got to a checkout or quote step and capacity was orderable. A price on a marketing page with no bookable stock behind it is listed here as unverified, because that is what it is.
- 03Why some rows are unnamed
- Named rows link to the provider's own public rate card. Unnamed rows are aggregates from our own survey, or listings we are not republishing as an endorsement. We would rather show you the number without the logo than not show you the number.
- 04Why our own rate is not a row
- Because a single headline figure would be wrong for almost every reader. A two-GPU evaluation and a two-year commitment on all six nodes are not the same purchase and should not carry the same number. We quote against the job on the call. What we will say in public is where the quote lands: under the lowest rate in this table that we could confirm was in stock.
- 05What we do not claim
- That this is the cheapest capacity available. Published numbers below the verified floor exist and are in the table; we could not book them, and neither could you on the day we looked.
- 06How often this changes
- Rates move. Every row carries the date we read it. If a figure here is stale, it is stale with a visible timestamp rather than quietly wrong.
48 B300s, configured as six 8-GPU nodes. Take-or-pay, 1–5 years. Updated 2 September 2026.
B300 capacity in days, not months.