GPU cloud providers, compared
Each provider is profiled from its own documentation, with the source recorded against every figure. The unit here is fixed by the hardware rather than the vendor: an H100 SXM 80GB is the same card everywhere, so the hourly rate for one is directly comparable — and every price names the card it was read for.
GPU prices by model
81 hourly prices from 8 clouds for the 16 GPU models more than one of them prices, from each cloud's own price list. Clouds are alphabetical within a model, not ranked: some print a rate per GPU and others a whole multi-GPU node, and the unit is on every row. The source and the date it was read are too.
B300 · B200 · H200 · H100 · A100 · L40S · L40 · RTX PRO 6000 · RTX 6000 Ada · A6000 · RTX 5090 · RTX 4090 · RTX 3090 · A5000 · A4000 · V100
One GPU, NVIDIA's spec sheet beside every cloud's price: NVIDIA H100
Compare two GPUs, NVIDIA's specs beside what each costs to rent: A100 vs H100 · H100 vs H200 · B200 vs H200 · B200 vs B300 · B200 vs H100 · A100 vs L40S
B300
| Cloud | As listed | VRAM | Per hour | Read |
|---|---|---|---|---|
| CoreWeave | NVIDIA HGX B300 | 270 GB | $35.84 for 8 GPUs | 23 Oct |
| Crusoe Cloud | NVIDIA B300 288GB HGX | 288 GB | $8.96 per GPU | 23 Oct |
| Hyperstack | NVIDIA B300 | 288 GB | $7.40 per GPU | 23 Oct |
| Nebius AI Cloud | NVIDIA HGX B300 | — | $9.50 per GPU | 23 Oct |
| RunPod | B300 | 288 GB | $8.99 per GPU | 23 Oct |
| Vast.ai | B300 | 288 GB | $5.90 per GPU | 23 Oct |
B200
H200
| Cloud | As listed | VRAM | Per hour | Read |
|---|---|---|---|---|
| CoreWeave | NVIDIA HGX H200 | 141 GB | $50.44 for 8 GPUs | 23 Oct |
| Crusoe Cloud | NVIDIA H200 141GB HGX | 141 GB | $4.29 per GPU | 23 Oct |
| Hyperstack | NVIDIA H200 SXM | 141 GB | $3.99 per GPU | 23 Oct |
| Nebius AI Cloud | NVIDIA HGX H200 | — | $5.40 per GPU | 23 Oct |
| RunPod | H200 | 141 GB | $5.29 per GPU | 23 Oct |
| Vast.ai | H200 NVL | 141 GB | $3.63 per GPU | 23 Oct |
| Vast.ai | H200 | 141 GB | $4.42 per GPU | 23 Oct |
H100
NVIDIA H100 specs from NVIDIA's own spec sheets, with these prices
| Cloud | As listed | VRAM | Per hour | Read |
|---|---|---|---|---|
| CoreWeave | NVIDIA HGX H100 | 80 GB | $49.24 for 8 GPUs | 23 Oct |
| Crusoe Cloud | NVIDIA H100 80GB HGX | 80 GB | $3.90 per GPU | 23 Oct |
| Hyperstack | NVIDIA H100 | 80 GB | $2.50 per GPU | 23 Oct |
| Hyperstack | NVIDIA H100 NVLink | 80 GB | $2.60 per GPU | 23 Oct |
| Hyperstack | NVIDIA H100 SXM | 80 GB | $3.20 per GPU | 23 Oct |
| Lambda | NVIDIA H100 PCIe | 80 GB | $3.29 per GPU | 23 Oct |
| Lambda | NVIDIA H100 SXM | 80 GB | $4.29 per GPU | 23 Oct |
| Nebius AI Cloud | NVIDIA HGX H100 | — | $4.50 per GPU | 23 Oct |
| Paperspace | H100 | — | $5.95 per GPU | 23 Oct |
| RunPod | H100 PCIe | 80 GB | $2.89 per GPU | 23 Oct |
| RunPod | H100 NVL | 94 GB | $3.19 per GPU | 23 Oct |
| RunPod | H100 SXM | 80 GB | $3.99 per GPU | 23 Oct |
| Vast.ai | H100 SXM | 80 GB | $1.73 per GPU | 23 Oct |
| Vast.ai | H100 NVL | 80 GB | $2.33 per GPU | 23 Oct |
A100
| Cloud | As listed | VRAM | Per hour | Read |
|---|---|---|---|---|
| CoreWeave | NVIDIA A100 | 80 GB | $21.60 for 8 GPUs | 23 Oct |
| Crusoe Cloud | NVIDIA A100 80GB PCIe | 80 GB | $2.00 per GPU | 23 Oct |
| Crusoe Cloud | NVIDIA A100 80GB SXM | 80 GB | $2.30 per GPU | 23 Oct |
| Hyperstack | NVIDIA A100 | 80 GB | $1.35 per GPU | 23 Oct |
| Hyperstack | NVIDIA A100 NVLink | 80 GB | $1.40 per GPU | 23 Oct |
| Hyperstack | NVIDIA A100 SXM | 80 GB | $1.60 per GPU | 23 Oct |
| Lambda | NVIDIA A100 SXM | 40 GB | $1.99 per GPU | 23 Oct |
| Lambda | NVIDIA A100 PCIe | 40 GB | $1.99 per GPU | 23 Oct |
| Lambda | NVIDIA A100 SXM | 80 GB | $2.79 per GPU | 23 Oct |
| Paperspace | A100 | — | $3.09 per GPU | 23 Oct |
| Paperspace | A100-80G | — | $3.18 per GPU | 23 Oct |
| RunPod | A100 PCIe | 80 GB | $1.79 per GPU | 23 Oct |
| RunPod | A100 SXM | 80 GB | $1.79 per GPU | 23 Oct |
| Vast.ai | A100 PCIE | 80 GB | $0.34 per GPU | 23 Oct |
| Vast.ai | A100 SXM4 | 80 GB | $0.40 per GPU | 23 Oct |
L40S
| Cloud | As listed | VRAM | Per hour | Read |
|---|---|---|---|---|
| CoreWeave | NVIDIA L40S | 48 GB | $18.00 for 8 GPUs | 23 Oct |
| Crusoe Cloud | NVIDIA L40S 48GB | 48 GB | $1.50 per GPU | 23 Oct |
| Nebius AI Cloud | NVIDIA L40S with Intel CPU | — | $1.55 per GPU | 23 Oct |
| Nebius AI Cloud | NVIDIA L40S with AMD CPU | — | $1.82 per GPU | 23 Oct |
| RunPod | L40S | 48 GB | $1.09 per GPU | 23 Oct |
| Vast.ai | L40S | 48 GB | $0.43 per GPU | 23 Oct |
L40
RTX PRO 6000
| Cloud | As listed | VRAM | Per hour | Read |
|---|---|---|---|---|
| CoreWeave | NVIDIA RTX PRO 6000 Blackwell Server Edition (Standard Memory) | 96 GB | $9.56 for 8 GPUs | 23 Oct |
| CoreWeave | NVIDIA RTX PRO 6000 Blackwell Server Edition (High Memory) | 96 GB | $20.00 for 8 GPUs | 23 Oct |
| Hyperstack | NVIDIA RTX Pro 6000 SE | 96 GB | $1.85 per GPU | 23 Oct |
| Nebius AI Cloud | NVIDIA RTX PRO 6000 | — | $1.80 per GPU | 23 Oct |
| RunPod | RTX Pro 6000 | 96 GB | $2.49 per GPU | 23 Oct |
| Vast.ai | RTX PRO 6000 WS | 96 GB | $1.00 per GPU | 23 Oct |
RTX 6000 Ada
A6000
| Cloud | As listed | VRAM | Per hour | Read |
|---|---|---|---|---|
| Hyperstack | NVIDIA A6000 | 48 GB | $0.50 per GPU | 23 Oct |
| Lambda | NVIDIA A6000 | 48 GB | $1.09 per GPU | 23 Oct |
| Paperspace | A6000 | — | $1.89 per GPU | 23 Oct |
| RunPod | RTX A6000 | 48 GB | $0.59 per GPU | 23 Oct |
RTX 5090
RTX 4090
RTX 3090
A5000
| Cloud | As listed | VRAM | Per hour | Read |
|---|---|---|---|---|
| Paperspace | A5000 | — | $1.38 per GPU | 23 Oct |
| RunPod | RTX A5000 | 24 GB | $0.27 per GPU | 23 Oct |
A4000
| Cloud | As listed | VRAM | Per hour | Read |
|---|---|---|---|---|
| Hyperstack | NVIDIA A4000 | 16 GB | $0.15 per GPU | 23 Oct |
| Paperspace | A4000 | — | $0.76 per GPU | 23 Oct |
V100
| Cloud | As listed | VRAM | Per hour | Read |
|---|---|---|---|---|
| Lambda | NVIDIA Tesla V100 | 16 GB | $0.79 per GPU | 23 Oct |
| Paperspace | V100 / V100-32G | — | $2.30 per GPU | 23 Oct |
LLM APIs
GPU compute
| Provider | H100 SXM, per GPU-hour | Largest card offered (VRAM) |
|---|---|---|
| Baseten | $6.50 per hour for a dedicated H100 80GB. Baseten publishes $0.10833 per MINUTE — the hourly figure is that times 60, ours rather than Baseten's, and the page offers an hourly toggle. A100 80GB is $0.06667 a minute and B200 $0.16633. | 180GB per GPU on a B200, at $0.16633 per minute — $9.98 an hour. |
| CoreWeave | $6.16 per GPU-hour for NVIDIA HGX H100 (80GB) is CoreWeave's inference-platform price, printed in its "Inference Single GPU Price" column, which it says applies only to inference platform customers. The on-demand instance price is $49.24/hour for the 8-GPU HGX H100 instance, spot $19.71/hour (North America). | 279GB per GPU on GB300 NVL72, which is Contact sales only. HGX B300 is next at 270GB, with no on-demand price and a $35.84/hour spot price for the 8-GPU instance. The most VRAM with a published on-demand price is GB200 NVL72 at 186GB per GPU: $42.00/hour instance price for 4 GPUs, or $10.50 per GPU-hour on the inference platform. |
| Crusoe Cloud | $4.50 per GPU-hour on demand for NVIDIA H100 80GB HGX. Crusoe's VM docs list the H100 only as an 8-GPU VM (h100-80gb-sxm-ib.8x), so the smallest H100 rental is eight GPUs. H200 141GB is $5.40 per GPU-hour; spot prices are not printed (Contact sales). | 288GB per GPU on NVIDIA HGX B300, at $9.90 per GPU-hour on demand; AMD MI355X also has 288GB but is Contact sales. NVIDIA B200 180GB is $8.50 and AMD MI300X 192GB $3.45 per GPU-hour; GB200 186GB is Contact sales. |
| DeepInfra | $2.20 per GPU-hour for a dedicated H100 80GB — below every dedicated cloud in this dataset. A100 80GB is $0.89, H200 $2.69, B200 $3.69. Billed in minute granularity, invoiced weekly. | 270GB per GPU on a B300, at $4.89 per GPU-hour. |
| Fireworks AI | $7.00 per hour for an H100 80GB on demand, or $8.00 on the shorter commitment shown beside it. The same rate applies to an H200 141GB. | 180GB per GPU on a B200, at $10.00 an hour on demand. |
| Hyperstack | $3.20 per hour for H100 SXM 80GB on demand. The H100 NVLink is $2.60, and the row the page lists as plain "NVIDIA H100" is $2.50. | 288GB per GPU on B300 at $7.40 an hour; B200 at 192GB is $6.00. |
| Lambda | $4.29 per GPU-hour for H100 SXM 80GB on a single-GPU instance, falling to $3.99 on an 8-GPU node. The PCIe variant is $3.29. | 180GB per GPU on B200 SXM6, from $6.69 per GPU-hour on an 8-GPU node to $6.99 on a single GPU. |
| Nebius AI Cloud | $4.50 per GPU-hour on demand for NVIDIA HGX H100, listed with 16 vCPUs and 200 GB RAM per GPU. H200 is $5.40. | 288GB HBM3e per GPU on NVIDIA B300 (Blackwell Ultra, SXM); B200 has 180GB and H200 141GB. |
| Novita AI | $3.39 per GPU-hour for an on-demand H100 SXM 80GB, or $1.70 spot. | 80GB per GPU on an H100 SXM, the largest card it lists; below it are an L40S 48GB, RTX 5090 32GB and RTX 4090 24GB. |
| Paperspace | $5.95 per hour for a single H100 80GB machine (20 vCPU, 250 GB RAM); H100x8 is $47.60 an hour. paperspace.com's pricing footnote calls $5.95 a special promo price but names no other on-demand rate. | 80GB per GPU on H100 (HBM3) and A100-80G (HBM2); H100x8 carries 640GB across eight GPUs. |
| RunPod | $3.99 per hour for H100 SXM 80GB, single card, on Secure Cloud (the page's default view). The PCIe variant is $2.89 and H100 NVL is $3.19. | 288GB per GPU on B300; B200 at 180GB is $7.99 an hour and H200 at 141GB is $5.29, on Secure Cloud (the page's default view). |
| TensorDock | From $2.25 per hour for an on-demand H100 SXM5 VM (1 to 8 GPUs), depending on the CPU and RAM allocated; $1.30 as a spot instance. Quoted on TensorDock's undated H100 page; live marketplace prices are shown only after sign-in. | — |
| Vast.ai | Median $2.40 per hour for H100 SXM 80GB, with a floor of $1.73 from the cheapest host. Vast is a marketplace — the floor is one host's offer at one moment and the median is what the market is actually charging, so both move without anyone announcing anything. This figure is a snapshot, not a rate card. | 288GB per GPU on B300 Blackwell Ultra, from $9.93 an hour with a median of $9.93. B200 at 192GB is from $6.25 with a median of $9.13. |
Which to consider
Teams that want per-token APIs and dedicated GPUs from one vendor with no plan fee — minute-granularity billing means a model that runs for ten minutes costs ten minutes.
Training runs that want a whole HGX node and move large datasets — free egress and free I/O operations remove the charge that usually dominates a data-heavy job.
Multi-GPU training at published per-GPU rates on NVIDIA and AMD: H100 $4.50, H200 $5.40 and MI300X $3.45 per GPU-hour, L40S at $1.50, with free egress, a 99.5% per-VM uptime SLA, and SOC 2 Type II and ISO 27001 across the platform.
Running open-weight models at the lowest published per-token rates here — DeepSeek-V4-Flash at $0.06 per million input is a fraction of what the lab itself charges — with dedicated H100s at $2.20 an hour if throughput demands it.
Running open-weight models without operating GPUs — the same DeepSeek and Qwen checkpoints the labs publish, served per token, with a path to dedicated hardware if throughput demands it.
Steady European workloads that can commit — reserved H200 at $2.79 against $3.99 on demand is the clearest published discount in this set, and egress is free.
Renting exactly the instance shape you need — Lambda publishes 1x, 2x, 4x and 8x plans for the same card, so a single H100 is a real product rather than a fraction of a node.
Per-second GPU rental at published on-demand rates, from L40S at $1.55 to B300 at $9.50 per GPU-hour (H100 $4.50), with free egress, nine public regions in Europe, Israel and the US, and SOC 2 Type II and ISO 27001 in its Trust Center.
Breadth — 200+ models under one API with a cache-read rate published next to most of them, and an H100 SXM at $3.39 an hour if the same vendor should run the dedicated side too.
Existing Paperspace users and light notebook work: the Gradient Free plan includes free M4000 GPU notebooks, machines start at $0.45 an hour on an M4000, and ingress and egress are free.
Short experiments and bursty inference — per-second billing and a $0.27 entry card mean an idle hour costs nothing and a small job is genuinely small.
Low-cost GPU servers from a marketplace of independent hosts, billed per second from a $5 deposit with no egress fees; its H100 page lists on-demand H100s from $2.25 an hour and an eight-GPU 3-year rate of $1.50 per GPU-hour.
Price-sensitive training and batch inference that can tolerate a marketplace — the median H100 SXM rate sits below what every managed cloud here lists for the same card, and interruptible capacity goes lower still.
GPU compute in the news
- Nvidia B200 Price Hits $8.01/Hr as GPU Costs Diverge - shattered.io
- Nebius Hikes AI Cloud Prices 21%, Memory Jumps 41% [2026] - shattered.io
- Google Puts $12/Hr Price on Ironwood TPU vs Nvidia [2026] - shattered.io
- H100 Rental Prices Jump 22% in a Month as GPU Shortage Hits Firms and Labs - Seoul Economic Daily
Comments are hosted by EchoThread, which sets cookies and offers social sign-in. Like the analytics on this site, it does not load until you accept — see the privacy page.