Two Nvidia data center GPUs, one Blackwell and one Ada Lovelace: 180 GB HBM3e against 48 GB GDDR6, 8.9x the memory bandwidth on the B200, and NVLink 1,800 GB/s on the B200 where the L40S is PCIe only.
At a glance
Memory
- B200
- 180 GB
- L40S
- 48 GB
3.8x more on B200
Memory bandwidth
- B200
- 7,700 GB/s
- L40S
- 864 GB/s
8.9x higher on B200
Cheapest on-demand
- B200
- $3.75 /GPU/hr
- L40S
- $0.55 /GPU/hr
6.8x more expensive on B200
Price comparison
Median price per GPU per hour by billing type, with the cheapest listing under it.
| Billing | B200 | L40S | Difference |
|---|---|---|---|
| On-demand | $6.79 /GPU/hr from $3.75 | $1.64 /GPU/hr from $0.55 | 4.2x more expensive on B200 |
| Reserved | $5.10 /GPU/hr from $3.75 | $1.11 /GPU/hr from $0.60 (12 mo) | 4.6x more expensive on B200 |
| Spot | $4.22 /GPU/hr from $3.12 | $1.11 /GPU/hr from $0.74 | 3.8x more expensive on B200 |
Heads up: A provider's own page may quote a different figure, for example on tax, region or a promotion. A monthly-only plan shows a derived hourly rate. Medians are across providers, one vote each, so a large catalog does not outweigh a small one. Verify before provisioning. More on how we price.
Price history
B200 vs L40S price history
Weekly median on-demand price per GPU per hour
Specs comparison
| Spec | B200 | L40S | Difference |
|---|---|---|---|
| Memory | |||
| Capacity | 180 GB HBM3e | 48 GB GDDR6 | 3.8x more on B200 |
| Bandwidth | 7,700 GB/s | 864 GB/s | 8.9x higher on B200 |
| Bus width | 384-bit | ||
| Compute · dense | |||
| FP4 | 9,000 TFLOPS | Not supported | B200 only |
| FP8 | 4,500 TFLOPS | 733 TFLOPS | 6.1x higher on B200 |
| FP16 / BF16 | 2,250 TFLOPS | 362.1 TFLOPS | 6.2x higher on B200 |
| INT8 | 4,500 TOPS | 733 TOPS | 6.1x higher on B200 |
| FP32 | 75 TFLOPS | 91.6 TFLOPS | 1.2x higher on L40S |
| FP64 | 37 TFLOPS | Not supported | B200 only |
| Platform | |||
| Interconnect | NVLink 1,800 GB/s | PCIe only | NVLink on B200 only |
| Board power | 1,000 W | 350 W | 2.9x higher on B200 |
| Released | Q1 2024 | Q3 2023 | 7 months newer on B200 |
| Precision | |||
| Supported formats | FP4FP6FP8FP16BF16TF32FP32FP64INT8 | FP8FP16BF16TF32FP32INT4INT8 | |
Compute: Vendor peak figures, dense and per GPU. Not measured throughput.
Source: B200 datasheet and L40S datasheet.
Where to rent them
Providers listing these GPUs, those with both first, so you can switch without moving clouds.
| Provider | B200 /GPU/hr | L40S /GPU/hr |
|---|---|---|
Packet·ai
|
On-Demand $3.75 | On-Demand $0.92 |
GPU.ai
|
On-Demand $5.40 | On-Demand $0.80 |
Lium
|
On-Demand $5.50 | -- |
Runpod
|
On-Demand $6.79 | -- |
Vast.ai
|
On-Demand $6.88 | -- |
Spheron
|
On-Demand $7.20 | -- |
Oblivus
|
On-Demand $8.40 | -- |
EmpirioLabs AI
|
On-Demand $9.00 | -- |
Novita
|
-- | On-Demand $0.55 |
Sesterce
|
-- | On-Demand $0.97 |
Massed Compute
|
-- | On-Demand $0.97 |
Runcrate
|
-- | On-Demand $0.97 |
UpCloud
|
-- | On-Demand $1.19 |
Lyceum
|
-- | On-Demand $1.19 |
35 providers list the B200 and 34 the L40S. Every listing is on the B200 page and the L40S page.
Similar GPUs
Other comparisons in the same class.
Our data for Nvidia B200 was last updated on Sept. 7, 2026, and for Nvidia L40S on Sept. 7, 2026.