A2 family · Accelerator Optimizedx86_64 · Intel/AMD

a2-highgpu-4g

The a2-highgpu-4g machine type has 48 vCPUs and 340 GB of memory. Pricing for this instance starts at $14.69 per hour and $10726.28 monthly in the us-central1 region.

Updated June 27, 2026
vCPUs
48
Memory
340 GB
Network
Up to 50 Gbps
Storage
Local SSD

Save up to $5364/mo on every a2-highgpu-4g

Connect Google Cloud — we'll find the optimal pricing model for your workload.

All pricing options

us-central1 · USD · monthly = 730 hrs
Prices exclude local taxes
TermHourlyMonthlySavings
On-Demand
$14.6935$10726.28 Baseline
1-Year Committed Use Discount (CUD)
$9.2568$6757.48 -37%
3-Year Committed Use Discount (CUD)
$5.1428$3754.27 -65%
Preemptible / Spot
$7.3454$5362.14 -50%
You could save $5364.14/mo on this instance with the right pricing model.
Find my savings

Monthly cost estimator

Estimate your spend based on actual usage.
730 h
1h365h730h (24/7)
1
Estimated monthly
$10726.28
730 hrs × $14.6935/hr × 1

Compute

x86_64 · Intel/AMD
vCPUs i48
Memory i340 GB
Physical processorIntel Cascade Lake
Nested VirtualizationNot supported
Sole TenantNot supported
GPU4 GPU
Within A2 family
a2-highgpu-1g
12 vCPU
a2-ultragpu-1g
12 vCPU
a2-highgpu-2g
24 vCPU
a2-ultragpu-2g
24 vCPU
a2-highgpu-4g
48 vCPU
a2-ultragpu-4g
48 vCPU
a2-highgpu-8g
96 vCPU
a2-ultragpu-8g
96 vCPU
a2-megagpu-16g
96 vCPU

Lock in this rate across your fleet — typically save 30–40%

Connect your Google Cloud account to apply this pricing logic to every running instance automatically.

Networking

Max egress bandwidth50 Gbps
Tier 1 bandwidthN/A
Enhanced networking iYes (gvnic/virtio)
IPv6 supportYes (Dual-stack VPC)

Storage

Max persistent disks128
Max disk size512 TB
Local SSDsSupported (Scratch Disk)

Compare with another instance

vs.
Speca2-highgpu-4ga2-highgpu-1gΔ
vCPUs4812-75%
Memory340 GB85 GB-75%
Hourly Price$14.6935$3.6734-75%
Monthly Price$10726.28$2681.57-75%
Cost optimization

GCP recommender alerts can cut this bill ~30-40%.

Connect your GCP account. We'll identify every underutilized a2.* instance and show exactly where downsizing, cleaning up orphaned disks, or shifting to CUDs saves money — with zero code changes.