A standard computing system solves problems primarily by using serial computing. It divides the workload into a sequence of tasks and then runs the tasks one after the other on the same processor.

Parallel computing runs multiple tasks simultaneously on numerous computer servers or processors. HPC uses massively parallel computing, which uses tens of thousands to millions of processors or processor cores.

An HPC cluster comprises multiple high-speed computer servers networked with a centralized scheduler that manages the parallel computing workload. The computers, called nodes, use either high-performance multi-core CPUs or—more likely today—GPUs, which are well suited for rigorous mathematical calculations, machine learning (ML) models and graphics-intensive tasks. A single HPC cluster can include 100,000 or more nodes.

Linux is the most widely used operating system for running HPC clusters. Other operating systems include Windows, Ubuntu and Unix.

All the other computing resources in an HPC cluster—such as networking, memory, storage and file systems—are high speed and high throughput. They are also low-latency components that can keep pace with the nodes and optimize the computing power and performance of the cluster.

HPC workloads rely on a message passing interface (MPI), a standard library and protocol for parallel computer programming that allows users to communicate between nodes in a cluster or across a network.

As recently as a decade ago, the high cost of HPC, which involved owning or leasing a supercomputer or building and hosting an HPC cluster in an on-premises data center, put it out of reach for most organizations.

Today, HPC in the cloud—sometimes called HPC as a service or HPCaaS—offers a significantly faster, more scalable and more affordable way for companies to take advantage of HPC. HPCaaS typically includes access to HPC clusters and infrastructure hosted in a cloud service provider's data center, network capabilities (such as AI and data analytics) and HPC expertise.

Today, three converging trends drive HPC in the cloud.

Organizations across all industries increasingly depend on the real-time insights and competitive advantage of using HPC applications to solve complex problems. For example, credit card fraud detection—something we all rely on, and most have experienced at one time or another—relies increasingly on HPC to identify fraud faster and reduce annoying false positives, even as fraud activity expands and fraudsters' tactics change constantly.

Since the launch of technologies like ChatGPT, organizations have rapidly embraced the promise of generative AI (gen AI) to accelerate innovation and foster growth. This development has spurred an even greater demand for high-performance computing. HPC provides the high computational power and scalability to support large-scale AI-driven workloads. In a report from Intersect 360 Research, the total worldwide market for scalable computing infrastructure for HPC and AI was USD 85.7 billion in 2023, up 62.4% year-over-year, due predominantly to a near tripling of spending by hyperscale companies on their AI infrastructure.²

Remote direct memory access (RDMA) enables one networked computer to access another networked computer's memory without involving either computer's operating system or interrupting either computer's processing. This helps minimize latency and maximize throughput, reducing memory bandwidth bottlenecks. Emerging high-performance RDMA fabrics—including InfiniBand, virtual interface architecture and RDMA over converged Ethernet—make cloud-based HPC possible.

Today, every leading public cloud service provider, including Amazon Web Services (AWS), Microsoft Azure, Google Cloud and IBM Cloud®, offers HPC services. While some organizations continue to run highly regulated or sensitive HPC workloads on-premises, many are adopting or migrating to private-cloud HPC services provided by hardware and solution vendors.