EXECUTIVE SUMMARY
This Intersect360 Research report presents the 2022 total market size and five-year forecast for the overall market for public cloud services used for High Performance Computing (HPC) and artificial intelligence (AI), segmented into cloud categories: raw cycles (infrastructure as a service, IaaS), storage, environment hosting (platform as a service, PaaS), and application hosting (software as a service, SaaS). The forecast horizon is from 2023 through 2027, with compound annual growth rates (CAGRs) using 2022 as a base.
This report presents multiple views of the market for infrastructure for HPC and AI. The “traditional HPC” market view represents HPC user budgets. In most cases, these environments have evolved to include machine learning as part of their overall workloads, and HPC budgets have generally increased because of machine learning; they are still essentially HPC budgets, for systems that would have existed regardless of machine learning. The “expanded HPC-AI” view incorporates the growing effect of spending on scalable infrastructure for AI workloads that is not associated with an HPC budget, i.e., spending that only exists for the sake of machine learning, not HPC. This usually relates to the training of AI models, although in some cases we find scalable clustered systems used for AI inference.
Intersect360 Research defines HPC as the use of servers, clusters, and supercomputers—plus associated software, tools, components, storage, and services—for scientific, engineering, or analytical tasks that are particularly intensive in computation, memory usage, or data management. Additionally, this report tracks similar infrastructure for machine learning and deep learning.
Intersect360 Research reports available in this series include the following segmentations:
- Products and Services: servers, storage, networks, software, service, cloud, other
- Vertical markets: academia, national security, national research labs, national agencies, state or local governments, bio sciences, chemical engineering, consumer product manufacturing, electronics, energy, financial services, large product manufacturing, media and entertainment, retail, transportation, other
- Regions: North America, EMEA, Asia-Pacific, Latin America
- Server class (HPC server revenue): entry-level, midrange, high-end, supercomputer
- Cloud categories (HPC cloud revenue): raw cycles, cloud storage, application hosting (SaaS), infrastructure hosting (IaaS, PaaS), other (traditional HPC model only)
- Software categories (HPC software revenue): operating environments, developer tools, middleware, storage software, transfer costs, application software, other (traditional HPC model only)
- Services categories (HPC services revenue): maintenance and repair, system engineering, system integration, training, programming services, other (traditional HPC model only)
- HPC server market shares (total on-premises HPC and AI, current year only, not forecast)
Worldwide HPC-AI Market Size
This forecast represents a return to steady growth following the recent pandemic-affected years. While some HPC users reported that supply chain issues did affect their ability to spend their complete budgets within the calendar year, the tradtional HPC market revenue grew to $38.1 billion ($38,085 million) in 2022, up 8.1% from 2021.
The expanded HPC-AI market grew to $44.4 billion ($44,431 million) in 2022, up 11.3% from 2021. Compared to the traditional HPC view above, this includes $6.3 billion ($6,346 million) in dedicated AI budgets not associated with HPC. This figure does not include the $17.9 billion spent by the Hyperscale market on AI and HPC infrastructure, which dwarfs the rest of the market.
Cloud computing has been one of the major trends affecting the HPC-AI market. It underwent two consecutive years of growth over 75% per year during the COVID-19 pandemic, and in 2022, it had its eighth straight year of double-digit growth, up 25.3% year-over-year in the traditional HPC view, and up 26.2% in the expanded HPC-AI view. Cloud computing has a much higher penetration overall in the expanded HPC-AI view, due to the proclivity of “pure AI” spending, in the absence of HPC, to leverage public cloud.
In either view, growth was roughly evenly distributed between the segments, with the exception that environment hosting (PaaS) declined in the traditional HPC view. Because this corresponded with a growth in “other” cloud, and PaaS did not decline among pure AI initiatives, we conclude this has to do mostly with the term falling out of favor with technical users. Cloud computing will approach its maximum penetration of the HPC-AI market over the course of the forecast, with growth concentrated in the application hosting (SaaS) category.
By the end of the decade, the “pure AI” market will begin to diminish, or else the distinction between scalable, high-performance infrastructure for HPC and machine learning will become less relevant. In the long run, Intersect360 Research expects the two views of the market to merge back together.