Instances GCP Compute Engine a3-highgpu-1g
a3-highgpu-1g
a3-highgpu-1g is a GCP Compute Engine machine type with 26 vCPUs, 234 GiB memory, 1 GPU (Intel Xeon (Emerald Rapids) or Intel Xeon (Sapphire Rapids)). In Iowa it costs $11.061 per hour on demand ($8,074.75 per month) for Linux; a 1-year committed use discount costs $7.703 per hour, 30% less. Spot capacity is about $6.588 per hour. It is sold in 19 regions.
USD per hour, Linux. Committed use prices are billed monthly. Cheapest on-demand first.
Region On-demand Spot 1-yr committed use 3-yr committed use Iowa us-central1 $11.061 $6.588 $7.703 $4.903 Northern Virginia us-east4 $11.061 $6.642 $7.703 $4.903 Oregon us-west1 $11.061 $6.588 $7.703 $4.903 Mumbai asia-south1 $11.50 $6.162 $8.008 $5.096 Finland europe-north1 $12.171 $1.142 $8.474 $5.391 Belgium europe-west1 $12.171 $5.641 $8.474 $5.391 Columbus us-east5 $12.448 $4.998 $8.666 $5.512 Las Vegas us-west4 $12.456 $5.562 $8.674 $5.52 London europe-west2 $12.598 $2.713 $8.77 $5.578 Taiwan asia-east1 $12.795 $7.677 $8.907 $5.665 Paris europe-west9 $12.818 $4.39 $8.923 $5.675 Frankfurt europe-west3 $13.037 $7.832 $9.075 $5.771 Delhi asia-south2 $13.269 $4.264 $9.236 $5.873 Sydney australia-southeast1 $13.806 $3.164 $9.609 $6.109 Netherlands europe-west4 $14.059 $8.437 $9.784 $6.219 Seoul asia-northeast3 $14.181 $7.533 $9.869 $6.273 Singapore asia-southeast1 $14.277 $6.308 $9.935 $6.315 Melbourne australia-southeast2 $14.355 $1.347 $9.99 $6.35 Tokyo asia-northeast1 $15.806 $2.727 $10.997 $6.986
Compute Value vCPUs 26 Architecture x86_64 Processor Intel Xeon (Emerald Rapids) or Intel Xeon (Sapphire Rapids)
Memory Value Memory (GiB) 234
Accelerators Value GPUs 1 GPU NVIDIA H100 80GB
Networking Value Network Up to 25 Gbps
Storage Value Storage 2 x 375 GiB NVMe local SSD Max persistent disks 128
Instance details Value Instance type a3-highgpu-1g Family a3 Category Accelerator optimized
Other Value Series A3 Specs source Compute Engine API
Prices updated 2026-11-12 from the Google Cloud Billing Catalog API. Compare all GCP Compute Engine sizes .