p6-b300.48xlarge Instance Details | Amazon Web Services

$22.24

Spot PriceOregonUSA

$142.42

On Demand PriceOregonUSA

$148.62

Median On Demand Price

$154.82

On Demand PriceOregonUSA

$22.24

Spot PriceOregonUSA

On-Demand	Region	Location	Tenancy	OS
$142.42	us-west-2	USA, Oregon	Shared	Linux
$142.54	us-west-2	USA, Oregon	Shared	SUSE
$142.75	us-west-2	USA, Oregon	Shared	Ubuntu Pro
$144.26	us-west-2	USA, Oregon	Shared	RHEL
$146.28	us-west-2	USA, Oregon	Shared	Red Hat Enterprise Linux with HA
$150.96	us-west-2	USA, Oregon	Dedicated	Linux
$151.09	us-west-2	USA, Oregon	Dedicated	SUSE
$151.30	us-west-2	USA, Oregon	Dedicated	Ubuntu Pro
$152.80	us-west-2	USA, Oregon	Dedicated	RHEL
$154.82	us-west-2	USA, Oregon	Dedicated	Red Hat Enterprise Linux with HA

Provider

Amazon Web Services

CPU

5th Gen Intel Xeon Scalable (Emerald Rapids, 96 cores, custom for AWS)

vCPU

192 vCPUs

Architecture

x86_64

Vendor

Intel

Threads/Core

2 threads

Base Clock

0 GHz

Max Turbo

0 GHz

All-Core Turbo

0 GHz

L2/Core

0 MB

TDP

0 W

Memory

4,096 GB

GPU Count

8 GPUs

GPU VRAM

2,144 GB

Category

GPU Accelerated

Availability

Current

P6-B300 is AWS’s next-generation EC2 GPU instance type for large-scale AI workloads, accelerated by NVIDIA Blackwell Ultra B300 GPUs. A single p6-b300.48xlarge provides 8 B300 GPUs with 2.1 TB of high-bandwidth GPU memory, 192 vCPUs, 4 TB of system memory, 6.4 Tbps of EFA networking, 300 Gbps dedicated ENA throughput, and high-performance local NVMe storage. Compared to P6-B200, P6-B300 delivers 2x networking bandwidth and 1.5x GPU memory and TFLOPs (FP4, no sparsity), making it well-suited for training and serving trillion-parameter foundation models, Mixture-of-Experts architectures, and other extremely large, distributed AI workloads on EC2 UltraClusters.