Two Nvidia data center GPUs, one Hopper and one Turing: 80 GB HBM3 against 16 GB GDDR6, 11x the memory bandwidth on the H100, and NVLink 900 GB/s on the H100 where the T4 is PCIe only.
At a glance
Updated 4 minutes ago- Memory
- 80 GB
- HBM3 · 94 GB on NVL
- Memory bandwidth
- 3.35 TB/s
- 2,000 GB/s on PCIe · 3,938 GB/s on NVL
- From
- $1.50 /GPU/hr
- 53 providers
- Memory
- 16 GB
- GDDR6
- Memory bandwidth
- 300 GB/s
- From
- $0.15 /GPU/hr
- 11 providers
Price comparison
Cheapest listed price per GPU per hour by billing type, and the median across providers.
| Billing | H100 | T4 | Difference |
|---|---|---|---|
| On-demand | $1.50 | $0.15 | 10x more expensive on H100 |
| Spot | $1.00 | $0.06 | 18x more expensive on H100 |
| Reserved | $1.53 | $0.20 | 7.7x more expensive on H100 |
| Median | $3.39 | $0.64 | 5.3x more expensive on H100 |
Median: Median across providers, one vote each, so a large catalog does not outweigh a small one.
Price history
H100 vs T4 price history
Weekly median on-demand price per GPU per hour
Specs comparison
| Spec | H100 | T4 | Difference |
|---|---|---|---|
| Memory | |||
| Capacity | 80 GB HBM3 94 GB on NVL | 16 GB GDDR6 | 5x more on H100 |
| Bandwidth | 3,350 GB/s 2,000 GB/s on PCIe 3,938 GB/s on NVL | 300 GB/s | 11x higher on H100 |
| Bus width | 5,120-bit | 256-bit | 20x wider on H100 |
| Compute · dense | |||
| FP8 | 1,979 TFLOPS | Not supported | H100 only |
| FP16 / BF16 | 989.5 TFLOPS | 65 TFLOPS | 15x higher on H100 |
| INT8 | 1,979 TOPS | 130 TOPS | 15x higher on H100 |
| FP32 | 67 TFLOPS | 8.1 TFLOPS | 8.3x higher on H100 |
| FP64 | 34 TFLOPS | Not supported | H100 only |
| Platform | |||
| Interconnect | NVLink 900 GB/s NVLink 600 GB/s on PCIe, NVL | PCIe only | NVLink on H100 only |
| Board power | 700 W | 70 W | 10x higher on H100 |
| Released | Q3 2022 | Q3 2018 | 4 years newer on H100 |
| Precision | |||
| Supported formats | FP8FP16BF16TF32FP32FP64INT8 | FP16FP32INT4INT8 | |
Figures are for each model's headline configuration. Where a form factor carries a different number, it is named under the figure.
Compute: Vendor peak figures, dense and per GPU. Not measured throughput.
Source: H100 datasheet and T4 datasheet.
Where to rent them
Providers listing these GPUs, those with both first, so you can switch without moving clouds.
| Provider | H100 /GPU/hr | T4 /GPU/hr |
|---|---|---|
Vast.ai
|
On-Demand $1.73 | On-Demand $0.15 |
Lium
|
On-Demand $1.50 | -- |
HyperAI
|
On-Demand $1.80 | -- |
GPU.ai
|
On-Demand $1.98 | -- |
Koyeb
|
On-Demand $2.50 | -- |
Spheron
|
On-Demand $2.64 | -- |
Massed Compute
|
On-Demand $2.73 | -- |
Runcrate
|
On-Demand $2.75 | -- |
Zenlayer
|
-- | On-Demand $0.56 |
Theta EdgeCloud
|
-- | On-Demand $0.20 |
Hugging Face
|
-- | On-Demand $0.40 |
Google Cloud
|
-- | On-Demand $0.42 |
AWS
|
-- | On-Demand $0.53 |
Azure
|
-- | On-Demand $0.53 |
Cerebrium
|
-- | On-Demand $0.59 |
53 providers list the H100 and 11 the T4. Every listing is on the H100 page and the T4 page.
Similar GPUs
Other comparisons in the same class.
Our data for Nvidia H100 was last updated on Sept. 7, 2026, and for Nvidia T4 on Sept. 7, 2026.