AWS PCS

AWS Parallel Computing Service

Managed high performance computing at virtually any scale

Get started with AWS PCS

Why AWS PCS?

AWS Parallel Computing Service (AWS PCS) runs Slurm-based high performance computing (HPC) workloads on AWS without the operational overhead. AWS PCS runs and maintains the cluster control plane on your behalf, so you can focus on running your workloads rather than managing infrastructure. It monitors cluster health and fixes common issues automatically. You keep full control of the Amazon EC2 instances your cluster uses, and you can connect other AWS services for compute, storage, networking, and visualization. Your teams work in a familiar HPC environment and use the tools they already know.

Benefits

Run clusters without the upkeep

AWS PCS sets up and operates your HPC clusters for you, handling provisioning, patching, and day-to-day maintenance, so your team spends its time on research and results instead of running infrastructure.

See cluster operations

Schedule and manage jobs your way

Give your users a managed Slurm scheduler with the semantics your HPC teams already know. Automate job submission, connect your own tools, and account for usage across teams.

See scheduler features

Pay only for the compute you use

Clusters scale up as work arrives and back down as it finishes, so you are not paying for idle machines. Choose the Amazon EC2 instances and purchase options that fit your budget.

See compute options

See how your clusters are running

Keep an eye on cluster health, performance, and job activity, and send that data to the monitoring tools your teams already use.

See monitoring features

Run every kind of HPC workload

PCS supports tightly-coupled MPI, high-throughput independent jobs, GPU and Trainium-accelerated compute, and interactive human-in-the-loop workflows, all from a single cluster.

See use cases

Use cases

Run engineering simulations

Design and validate physical products with computational fluid dynamics (CFD), structural and crash analysis, thermal analysis, and electromagnetics. Run parallel MPI applications and sweep parameters across many design variants at once.

Train and fine-tune AI and ML models

Train new models, fine-tune existing ones, and run reinforcement learning at scale on GPU-based instances and AWS Trainium. Reserve accelerated capacity when the run demands it, and connect training nodes with low-latency networking.

Accelerate scientific research

Advance work in life sciences, physics, chemistry, and earth science: genomics analysis, protein structure prediction, molecular dynamics, Cryo-EM, weather and climate modeling, and seismic and reservoir simulation. Iterate on runs interactively as results come in.

Run financial risk and quantitative analysis

Deliver risk analytics and pricing on the timelines your desk needs. Run Monte Carlo simulations, market risk calculations, derivative pricing, and portfolio analytics across many independent workers, scaling out for overnight batches and end-of-quarter reporting.

Design semiconductors and hardware (EDA)

Move chip designs from RTL to tapeout with electronic design automation: logic simulation and verification, synthesis, place-and-route, and timing analysis. Scale out to thousands of cores for peak demand and scale back when the queue clears.

Customers

Aionics

Aionics moved its computational chemistry clusters from AWS ParallelCluster to AWS PCS in August 2025. It runs separate CPU and GPU Slurm clusters so its team can focus on battery electrolyte research instead of infrastructure.

Read the blog

Tune Therapeutics

“Using AWS PCS, we can make decisions far more rapidly, which is critical for a biotech company.”

Harry Winters, Senior Software Engineer, Tune Therapeutics

Read the case study

Daiichi Sankyo

“AWS Parallel Computing Service gives us a sustainable, flexible HPC environment for drug discovery with far less administrative overhead, and it frees our researchers to focus on the science rather than managing infrastructure.”

Source: How Daiichi Sankyo modernized drug discovery using AWS Parallel Computing Service | AWS HPC Blog

Read the blog

Shell

“AWS PCS has changed how we approach high-performance computing. We spend less time managing infrastructure and more time innovating, spinning up capacity the moment we need it and delivering results 3-5x faster than before.”

Watch Shell's re:Invent 2025 talk

Amazon Leo

“We migrated our EDA chip-design workloads from a managed LSF cluster to AWS PCS with Slurm, and it was more straightforward than we expected. Our biggest problem was compute power and scaling with so many jobs running — and that problem is completely gone. PCS just scales up when we need it, and we've cut costs by 30%, improved performance by 4x, and reduced operational overhead by 90%.”

Ekjot Basra, Amazon Leo Silicon Development