🚀 B200 starting at $3.75/hr. The best price you'll find. DC in US West → (Access it from button on top after login).

Get Your B200 →
Start Building
In stock · Available now

NVIDIA Blackwell, 192 GB HBM3e, SXM

NVIDIA B200Blackwell. 192 GB. The frontier.

The NVIDIA B200 is the Blackwell-generation flagship with 192 GB HBM3e memory and 8 TB/s of bandwidth, the largest and fastest GPU available. It runs 405B models at FP16 natively and delivers 2.25 PetaFLOPS of FP4 compute. Available on packet.ai from $3.75/GPU-hour.

From $3.75/GPU-hr~36% below market

Dynamic from $3.75/hr · Dedicated from $5.90/hr

192GB
HBM3e memory
8TB/s
Memory bandwidth
2.25PFLOPS
FP4 compute
1.8TB/s
NVLink bandwidth
Architecture

Blackwell HBM3e — the frontier.

The B200 pairs 5th-gen Tensor Cores with 192 GB HBM3e and 1.8 TB/s NVLink for the largest training and inference jobs.

5th-gen Tensor Cores + FP4

FP4 and FP8 compute. 2.25 PFLOPS FP4 for the largest transformer models at maximum throughput.

192 GB HBM3e at 8 TB/s

192 GB handles 405B at FP16 natively. 8 TB/s bandwidth, the fastest memory access of any GPU.

NVLink 5.0 (1.8 TB/s)

1.8 TB/s NVLink for tight multi-GPU coupling. Scale to 1,024+ GPUs with full bandwidth coherence.

Transformer Engine 2.0

Native FP4 mixed precision with micro-scaling. 2× the throughput of H100 on transformer inference.

Technical specs

NVIDIA B200 specifications.

SpecificationValueGreat for
GPU architecture
NVIDIA Blackwell
FP4 Tensor Cores — the frontier standard.
GPU memory
192 GB HBM3e
405B at FP16 natively.
Memory bandwidth
8 TB/s
Fastest token generation of any GPU.
FP4 compute
2.25 PFLOPS
Frontier model inference at maximum throughput.
NVLink
5.0 · 1.8 TB/s
Scale to 1,024+ GPUs.
Power
700W TDP
Liquid-cooled data centre deployment.
Pricing

Three ways to run B200.

Dynamic, dedicated, or multi-node clusters.

DedicatedHourly · Single-tenant
$5.90 /GPU-hr

Full B200 reserved exclusively for you. 99.99% SLA, zero noisy-neighbour risk.

Deploy now →
Multi-node Cluster
From 8 GPUs

Scale frontier training and inference across multiple B200 nodes with NVLink 5.0 and InfiniBand interconnect.

  • 8–512 GPUs per cluster
  • NVLink + InfiniBand
  • Provisioned in <1 hr
Get a wholesale quote →
Use cases

What the B200 is built for.

Frontier model training

2.25 PFLOPS FP4 and NVLink 5.0. The fastest training GPU available.

  • FP4 mixed precision
  • 1.8 TB/s NVLink
  • 1,024+ GPU clusters

405B inference at FP16

192 GB runs 405B at FP16 natively. No sharding, no quantisation.

  • 192 GB HBM3e
  • 8 TB/s bandwidth
  • Single-GPU 405B

Pre-training at scale

InfiniBand clusters with 1,024+ B200s for frontier pre-training runs.

  • Cluster from 8 nodes
  • Wholesale pricing
  • Named TAM
FAQ

NVIDIA B200, answered.

For anything else, reach help@packet.ai.

What is the NVIDIA B200?

Blackwell flagship: 192 GB HBM3e, 8 TB/s, 2.25 PFLOPS FP4. The largest and fastest GPU available.

How much does the B200 cost?

From $3.75/GPU-hour dynamic, $5.90/hr dedicated.

Can B200 run 405B models natively?

Yes. 192 GB HBM3e is enough for Llama 3.1 405B at FP16 on a single card.

How does B200 compare to H100?

B200 has 2.4× the memory (192 vs 80 GB), 2.4× the bandwidth (8 vs 3.35 TB/s), and 3× the FP8 compute. It is the current frontier GPU.

Is B200 available now?

Yes. Join the waitlist on packet.ai to secure allocation. Capacity is limited.

Run the B200. 192 GB Blackwell from $3.75/hr.

The frontier GPU — $3.75/hr dynamic, $5.90/hr dedicated.

On-demand · hourly billing · US & EU regions

NVIDIA B200from $3.75/GPU-hr In stock
Deploy now →