NVIDIA B200 vs NVIDIA A100 40GB: which is better for AI?

Choose NVIDIA B200 if you need 180GB+ on a single GPU (140GB more than NVIDIA A100 40GB). Choose NVIDIA B200 if you need FP16 tensor throughput (621% faster than NVIDIA A100 40GB).

Can I deploy NVIDIA B200 or NVIDIA A100 40GB on Runcrate today?

NVIDIA B200 is available as reserved capacity on request. NVIDIA A100 40GB is on demand in 5 regions from $1.99/hr, billed by the second on most GPUs.

How much VRAM does each GPU have?

NVIDIA B200 has 180GB HBM3e. NVIDIA A100 40GB has 40GB HBM2.

GPU comparison

NVIDIA B200 vs NVIDIA A100 40GB.

Specs, performance and live cloud availability, side by side.

Launch A100 40GB Get a reserved quote

The short answer

Choose NVIDIA B200 if you need 180GB+ on a single GPU (140GB more than NVIDIA A100 40GB).
Choose NVIDIA B200 if you need FP16 tensor throughput (621% faster than NVIDIA A100 40GB).

Compare another pair

Blackwell

NVIDIA B200

Reserved capacity · on request

Memory: 180 GB HBM3e
FP16 tensor: 2250 TFLOPS
Bandwidth: 8.0 TB/s

Reserve capacity Deploy H200 instead

Full pricing details

Ampere

NVIDIA A100 40GB

from$1.99/hr

Available now · 5 regions

Memory: 40 GB HBM2
FP16 tensor: 312 TFLOPS
Bandwidth: 1.6 TB/s

Launch A100 40GB Get a reserved quote

Full pricing details

At a glance

Spec by spec. The longer bar is the better number.

Memory

NVIDIA B200180 GB

NVIDIA A100 40GB40 GB

Memory bandwidth

NVIDIA B2008.0 TB/s

NVIDIA A100 40GB1.6 TB/s

FP16 tensor

NVIDIA B2002250 TFLOPS

NVIDIA A100 40GB312 TFLOPS

FP32

NVIDIA B20090 TFLOPS

NVIDIA A100 40GB19.5 TFLOPS

Full specs

The complete spec sheet.

Spec	NVIDIA B200	NVIDIA A100 40GB
Architecture	Blackwell	Ampere
Memory	180 GB	40 GB
Memory type	HBM3e	HBM2
Memory bandwidth	8.0 TB/s	1.6 TB/s
FP32	90 TFLOPS	19.5 TFLOPS
FP16 tensor	2250 TFLOPS	312 TFLOPS
INT8	3600 TOPS	624 TOPS
TDP	1000 W	up to 400 W
GPUs per node	8	8
Hourly price (live)	Reserve on request	from $1.99/hr

Questions

B200 or A100 40GB. What buyers ask first.