The hpc8a.96xlarge pairs 192 vCPU with 768 GiB of RAM, the 4.00 GiB-per-vCPU ratio that marks a general-purpose instance. In the name, "hpc" is high performance computing, "8" is the generation, "a" marks AMD processors and "96xlarge" is the size within the hpc8a family. It runs on AMD EPYC 9R45 Processor at 4.5 GHz, and AWS introduced it in February 2026. hpc8a is the current generation of the hpc family, so there is no newer AWS size to migrate to. Spot's node ladder tops out below this size, so hpc8a.96xlarge becomes 12 × gp.vs2.2xlarge (192 vCPU, 768 GiB combined). Spot delivers compute as managed Kubernetes, so a node pool of that size is the native shape of the answer rather than a workaround — provided the workload scales horizontally.
Match qualityScale-out match
hpc8a.96xlarge gives you 192 vCPU and 768 GiB in a single machine. The largest comparable Spot node, gp.vs2.2xlarge, is 16 vCPU and 64 GiB, so the equivalent here is 12 nodes rather than one.
Spot runs as managed Kubernetes, so a 12-node pool is the normal way to run things rather than a workaround. If your workload already sits behind a load balancer, a queue or a scheduler, it spreads across them without changes. If it needs all 192 cores and 768 GiB in one process, it does not.
Pricing by region
1 hr
Region
AWS on-demand
AWS Spot
1-yr Reserved
3-yr Reserved
Rackspace Spot
You save
US East, N. Virginia
$0.492
US Central, Chicago
no AWS region
$0.492
US Central, Dallas
no AWS region
$0.612
Hong Kong
$0.132
Sydney, Australia
$1.932
London, UK
$2.652
US West, San Jose
$1.812
US Central, Dallas 2
no AWS region
$0.012
US East, N. Virginia 2
$1.812
When to use it
Good for
●Backend services with no single dominant resource
●Queue workers and scheduled jobs
●Web and application servers with mixed CPU and memory demand
●Workloads that fit inside 768 GiB of RAM
●Parallel work that can occupy 192 vCPU
●Network-heavy services — AWS rates this size at 300 Gigabit
●Anything that shards cleanly across 12 nodes, which is how it maps onto Rackspace Spot
Avoid if
●The working set needs more than 768 GiB of RAM
●The work is single-threaded. Most of these 192 vCPU would sit idle.
●Local NVMe or instance-store disk is a hard requirement
●GPU or FPGA acceleration is needed
●Sustained full-core CPU load. A compute-optimized family costs less per unit of work.
●More than 4.00 GiB of RAM per vCPU is needed — that ratio is fixed for the whole family
Rackspace Spot sells spare capacity at market prices. You pay what the market decides, not a list price.
Spot sells spare capacity at market prices. It is the cheapest compute on the internet.
Prices are set by supply and demand. You pay what the market decides.
Compute is delivered as fully managed Kubernetes. All infrastructure is taken care of.
Egress costs $0.01/GB after the free allowance. AWS charges $0.09/GB.
