Cheapest GPU node to train for a week.
prices as of · re-ranked daily · 63 qualifying machines
A week on a multi-GPU node is the unit a real training run is bought in, and the gap between the on-demand price and the interruptible one is the whole decision. We count only machines with two GPUs or more, each with 40 GB or more, and 63 of the 156 rentals we read qualify. The cheapest of them is $198 per week at 2A6000.20V · 2x RTX A6000 48GB, which rents for $0.592/GPU-hr. They are ranked by what the work costs rather than by the hourly rate, because a faster card at twice the price can still be the cheaper way to buy the job.
The cheapest machine that can do this job is 2A6000.20V · 2x RTX A6000 48GB at Verda (DataCrunch): $198 per week, read 17 Sept 2026.
Cheapest that fits
2A6000.20V · 2x RTX A6000 48GB
Verda (DataCrunch)$198per week
$0.592/GPU-hr · $99 on spot
lowest cost per week
Most memory for the money
8A6000.80V · 8x RTX A6000 48GB
Verda (DataCrunch)$796per week
$0.592/GPU-hr · $398 on spot
the largest instance in the first twenty-five
What actually matters here
- More than one GPU
- Training anything worth a week of compute is split across GPUs. One card is a fine-tune, not a training run, which is why this list starts at two.
- 40 GB a board or more
- Below that the batch size collapses and the run spends its week moving data instead of learning. It is the floor the training boards of every generation clear.
- What links the GPUs
- NVLink and NVSwitch inside a node, InfiniBand between them. Where a provider states it, the row says so; where it does not, nothing here guesses.
- Checkpoint if you take the spot price
- An interruptible machine can be taken back mid-run. The saving is real and so is the restart, so the two columns are shown side by side rather than blended into one number.
Cheapest 10 per week
- 1Verda (DataCrunch)
2A6000.20V · 2x RTX A6000 48GB
$198 per week · 2x RTX A6000 · 48 GB each · 20 vCPU · 120 GB RAM · $1.18/hr for the machine · $99 at its spot rateDedicated Hardware Instance · storage billed separately · 50GB/s between GPUs
✓ 48 GB per GPU✓ 2 GPUs✓ spot rate published
$0.592/GPU-hr
- 2
Lambda2x NVIDIA A6000
$366 per week · 2x RTX A6000 · 48 GB each · 28 vCPU · 200 GB RAM · $2.18/hr for the machinePriced per GPU-hour, billed by the minute · storage included in the instance
✓ 48 GB per GPU✓ 2 GPUs
$1.09/GPU-hr
- 3Verda (DataCrunch)
2RTX6000ADA.20V · 2x RTX 6000 Ada 48GB
$368 per week · 2x RTX 6000 Ada · 48 GB each · 20 vCPU · 120 GB RAM · $2.19/hr for the machine · $184 at its spot rateDedicated Hardware Instance · storage billed separately · 50GB/s between GPUs
✓ 48 GB per GPU✓ 2 GPUs✓ spot rate published
$1.09/GPU-hr
- 4Verda (DataCrunch)
4A6000.40V · 4x RTX A6000 48GB
$398 per week · 4x RTX A6000 · 48 GB each · 40 vCPU · 240 GB RAM · $2.37/hr for the machine · $199 at its spot rateDedicated Hardware Instance · storage billed separately · 50GB/s between GPUs
✓ 48 GB per GPU✓ 4 GPUs✓ spot rate published
$0.592/GPU-hr
- 5Verda (DataCrunch)
2L40S.40V · 2x L40S 48GB
$484 per week · 2x L40S · 48 GB each · 40 vCPU · 120 GB RAM · $2.88/hr for the machine · $242 at its spot rateDedicated Hardware Instance · storage billed separately · 50GB/s between GPUs
✓ 48 GB per GPU✓ 2 GPUs✓ spot rate published
$1.44/GPU-hr
- 6
Vultrvcg-l40s-32c-375g-96vram
$561 per week · 2x L40S · 48 GB each · 32 vCPU · 375 GB RAM · $3.34/hr for the machineCloud GPU · $2245.82/mo if kept a whole month
✓ 48 GB per GPU✓ 2 GPUs
$1.67/GPU-hr
- 7Verda (DataCrunch)
2A100.44V · 2x A100 SXM4 80GB
$575 per week · 2x A100 80GB SXM · 80 GB each · 44 vCPU · 240 GB RAM · $3.42/hr for the machine · $287 at its spot rateDedicated Hardware Instance · storage billed separately · 100 GB/s between GPUs
✓ 80 GB per GPU✓ 2 GPUs✓ spot rate published
$1.71/GPU-hr
- 8Verda (DataCrunch)
2RTXPRO6000.60V · 2x RTX PRO 6000 96GB
$618 per week · 2x RTX PRO 6000 · 96 GB each · 60 vCPU · 180 GB RAM · $3.68/hr for the machine · $309 at its spot rateDedicated Hardware Instance · storage billed separately · 1597 GB/s between GPUs
✓ 96 GB per GPU✓ 2 GPUs✓ spot rate published
$1.84/GPU-hr
- 9Lambda
2x NVIDIA A100 PCIe
$669 per week · 2x A100 40GB PCIe · 40 GB each · 60 vCPU · 450 GB RAM · $3.98/hr for the machinePriced per GPU-hour, billed by the minute · storage included in the instance
✓ 40 GB per GPU✓ 2 GPUs
$1.99/GPU-hr
- 10Lambda
4x NVIDIA A6000
$732 per week · 4x RTX A6000 · 48 GB each · 56 vCPU · 400 GB RAM · $4.36/hr for the machinePriced per GPU-hour, billed by the minute · storage included in the instance
✓ 48 GB per GPU✓ 4 GPUs
$1.09/GPU-hr
A week is 168 hours of the whole instance at its published hourly rate, with no commitment discount, because a committed rate is not the standard recurring price this index carries. The interruptible column is the same 168 hours at the spot rate, which assumes the run is never preempted; a real run on spot costs somewhere between the two. The price beside each row is the on-demand rate for one GPU for one hour, so the two numbers answer different questions: $0.592/GPU-hr is what the GPU costs, and $198 per week is what the work costs.
Frequently asked
- What is the cheapest GPU to train for a week?
- 2A6000.20V · 2x RTX A6000 48GB at Verda (DataCrunch), $198 per week. That is 2x RTX A6000 with 48 GB of memory each, renting for $0.592/GPU-hr. Read on 17 Sept 2026.
- How is the cost worked out?
- A week is 168 hours of the whole instance at its published hourly rate, with no commitment discount, because a committed rate is not the standard recurring price this index carries. The interruptible column is the same 168 hours at the spot rate, which assumes the run is never preempted; a real run on spot costs somewhere between the two.
- Why do only 63 machines qualify?
- The job needs two GPUs or more, each with 40 GB or more. The other 93 rentals we read are still in the index; they just cannot do this job as described. Serverless rates are left out entirely, because they buy a container that runs when it is called rather than a machine you hold.
- Can I pay less on spot?
- 41 of the 63 qualifying machines publish an interruptible rate, and the column beside each row shows what the same work would cost on it if the run were never preempted. It can be, so treat the spot figure as a floor rather than a quote.
- How current are these prices?
- Prices are read daily from 10 providers' own public pages and keyless APIs, and this ranking recomputes with them. Last refresh: 17 Sept 2026.