The U.S. is entering a period of rapid electricity demand growth, driven by power consumptive technologies such as AI, coupled with data center expansion and economic development, including new domestic manufacturing. Electricity demand is expected to grow 15-20% in the next decade and double by 2050. The DOE Office of Policy recently released the blog post - Clean Energy Resources to Meet Data Center Electricity Demand, highlighting DOE resources designed to assist data center developers in meeting electricity demands through clean energy solutions. DOE resources available to data centers include loans, grants, FOAs, tax credits, and technical assistance.

The purpose of this study was to gain a better understanding of the activities and use behaviors of AI users leveraging HPC-centric cloud resources. Key goals included creating a picture of user goals for AI integration, their current and planned methodologies, budget allocations, model lifecycle expectations, and preferred hardware and cloud platforms for their HPC-centric AI endeavors. This study also sought to gain a greater understanding of inferencing activities among organizations currently or planning to integrate AI into their workflow. Highlights from the study results include: •     Public cloud resources are considered a valuable asset in exploration and integration of AI into HPC or compute-intensive environments. •     Respondent organizations are leveraging a wide range of public cloud offerings. •     There are numerous architectures and device types currently used to meet inferencing needs. •     There is a plethora of desired qualities for the future of cloud computing expected by current and prospective AI users. •     Budgets among respondent organizations are expected to increase to meet training and inference needs, both in the cloud and on-premises.

The EuroHPC Joint Undertaking (JU) recently initiated two calls seeking interest from hosting sites to house AI Factories. These calls stem from the July 2024 appended JU regulation to expand the JU's AI objectives. Funding includes up to €400M for new or upgraded AI EuroHPC supercomputers and up to €180M for the establishment and operation of the AI Factories as well as the development and deployment of advanced experimental AI-optimized supercomputing platforms. The EU funds will be matched with an equal amount of budget by participating JU countries. The AI Factories are intended to bring together key elements required for AI success (computing power, data, and talent) and make the resources available to a broad range of European users, including start-ups, industry, and researchers.

A recently released US government report prepared by the US National Science and Technology Council, a cabinet-level council of advisors to the US President, presented a compelling argument for robust international cooperation in quantum information science and technology (QIST). The group called for dedicated funding mechanisms and improved US government interagency coordination to bolster global QIST collaboration. Key motivations cited for standing up a centralized US government cooperative QIST program included opportunities to advance QIST capability at a faster pace, to grow, attract, and engage international talent, to promote robust access to necessary QIST resources and their eventual markets, and to strengthen international engagement with nations considered to be aligned with US economic and national security interests. The report recognized that US government departments operate under an autonomous funding model with individual agencies having their own programmatic mechanisms, complicating the options for those agencies to participate in large-scale, multinational cooperative partnerships either individually or collectively.

With the recent political transition in the UK, national investment priorities to stimulate the economy appear to have shifted. The new leadership, through the Department for Science, Innovation, and Technology (DSIT), recently rescinded a commitment made by the prior government to spend £1.3 billion towards advancing AI-based research, including £800 million to develop an exascale supercomputer at the Edinburgh Parallel Computer Center (EPCC) and £500 million to extend and expand the AI Research Resource (AIRR) to support computing power for AI research. This decision does not impact £300 million already distributed for the AIRR, nor DSIT's funding for its AI Opportunities Action Plan.

The University of California Board of Regents recently approved the creation of a new school at UC San Diego aimed to provide students opportunities to engage directly with industry and government partners, and learn first-hand how data science can allow organizations to better address a broad range of societal problems such as climate change mitigation, social justice issues, technical challenges, and healthcare. The School of Computing, Information and Data Sciences (SCIDS) intends to combine the strengths of the San Diego Supercomputer Center (SDSC) and the Halicioğlu Data Science Institute (HDSI) to create a curriculum to advance data science and AI state-of-the-art.

For years, the AI cloud accelerator market has been dominated by just a few powerful GPU offerings, enabling the rapid advancement of artificial intelligence and machine learning. However, the growing demand for efficient and specialized hardware to power complex AI workloads has led to the rise of a trend: the development of custom AI chips by cloud service providers (CSPs). CSPs, such as AWS, Google, Microsoft, and Alibaba. They have recognized the need for more tailored hardware solutions to address the unique requirements of their cloud-based AI services and customers. By developing their custom AI chips, these CSPs aim to diversify their accelerator offerings to users, optimize the hardware-software stack for their cloud environments, and gain a competitive edge in the rapidly evolving AI cloud ecosystem.

In a soon to be available Hyperion Research study focusing on HPC users leveraging cloud resources for AI tools and capabilities, respondents most often indicated that they are currently engaged in exploratory and experimental stages of AI adoption, even those who have already integrated AI capabilities and technologies into some portion of their production environments. In this study, respondents demonstrated a preference for cloud over on-premises options for a range of AI-centric activities.

US-based quantum computer (QC) maker, Infleqtion, announced that it will provide the first of seven prototype deployments made under the UK National Quantum Computing Centre's (NQCC) £30 m (US$38.9 million) program designed to establish QC testbeds based on different hardware technologies by March 2025. All of the system-level prototypes will be built by the quantum hardware companies at the NQCC facilities on the Harwell Campus in Oxfordshire. The stated goal of the testbed program is to support the NQCC's mission of accelerating the development of quantum computing capabilities and related infrastructure within the UK. The six other QC makers participating in the program, who likely will be announcing their deployments shortly, are Aegiq, ORCA Computing, Oxford Ionics, Quantum Motion, QuEra Computing, and Rigetti, a mix of US, EU, and Israeli-based QC suppliers.

At their annual HPE Discover conference, HPE introduced the HPE Private Cloud AI family of turnkey solutions as part of its NVIDIA AI Computing by HPE portfolio. The HPE Private Cloud AI solutions integrate NVIDIA AI computing, networking, and software with HPE's AI storage and HPE Greenlake Cloud, with complete on-premises solutions fully tested with all the software preloaded. Beyond the infrastructure elements HPE is bringing to this partnership, they are also contributing business leadership in areas such as sales teams and channel partners, training, and a global network of system integrators to the joint go-to-market (GTM) efforts that aim to help enterprises expedite time-to-value from their investments in AI inferencing.

The HPC User Forum was established in 1999 to promote the health of the global HPC industry and address issues of common concern to users. In April 2024, the 84th HPC User Forum took place in Reston, Virginia. This update summarizes a presentation from that conference given by John Speakman, Assistant Vice President for Research and Education Information Technology at NYU Langone. He leads the provision of information and technology for the academic (scientific and educational) missions of NYU Langone Health, advancing the vision of the institute's digital transformation. His discussion at the April 2024 meeting covered the HPC resources built and leveraged at the university and healthcare system as well as details about their newly integrated AI tools.

With a theme of Reinventing HPC, ISC24 provided a venue for the global HPC community to come together and share knowledge and experiences relative to the latest market developments and technology innovations for the evolution of advanced technical computing. That said, it was clear that this was the year that AI took over the conference and was the key topic of discussion. Issues at the forefront centered on new hardware and software to support the rapidly expanding demand for AI training and inference, the promise and concerns with AI 's abilities to accelerate scientific and engineering workloads, apprehensions with the relative slowdown in performance progress made in the traditional HPC space, the growing presence and influence of cloud-based capabilities and their associated providers, especially in the areas of AI, and the issues, if not angst, of new AI processor availability, performance, and cost. Indeed, the approximately 3,200 attendees reflected a wide array of sentiment towards the convergence and adoption of AI within the worldwide scientific and research community, ranging from examples of compelling use cases and success stories to lessons learned from associated challenges, to "… AI will soon be inflicted upon us … ."

Two leading government-funded Japanese research facilities recently announced plans to procure two on-premise quantum computing (QC) systems, both from US suppliers, to support the exploration of new capabilities in QC as well as hybrid quantum/classical capabilities. QuEra Computing announced it has been awarded a 6.5 billion JPY contract (approx. $41M USD) by Japan's National Institute of Advanced Industrial Science and Technology to deliver one of its gate-based neutral-atom quantum computers in 2025. IBM announced an agreement with Riken, a Japanese national research laboratory that is home to Fugaku, one of the fastest classical HPC in the world, to procure an IBM System 2, powered by one of IBM's most advanced superconducting quantum processors, the 133-qubit Heron.

This Hyperion Research study presents the latest five-year forecast (2023-2028) for HPC on-premises technical servers. Worldwide revenue for the HPC technical server market in 2023 was $15.0 billion, representing a slight decline (-2.7%) over 2022 revenues. Hyperion Research now predicts that the HPC technical server market will grow at an 8.2% CAGR between 2023 and 2028 to reach $22.2 billion in 2028 (see Figure 1). Compared with the previous version of the forecast, there have been some downgrades to the revenues for 2023 through 2027, primarily due to rescheduling of some large exascale systems and ongoing supply chain delays.

In April 2024, the 84th HPC User Forum took place in Reston, Virginia. This update summarizes a presentation from that conference given by Stephen Ezell, Vice President for Global Innovation Policy at the Information Technology and Innovation Foundation (ITIF) and Director of ITIF's Center for Life Sciences Innovation. He also leads the Global Trade and Innovation Policy Alliance. Ezell's discussion at the April 2024 meeting covered an update on the CHIPS and Science Program as well as some issues surrounding the funding and development of the program's goals.

The HPC User Forum was established in 1999 to promote the health of the global HPC industry and address issues of common concern to users. In April 2024, the 84th HPC User Forum took place in Reston, Virginia. This update summarizes a presentation from that conference given by Buddy Rizer, Executive Director for economic development in Loudoun County, Virginia. He leads the agency responsible for encouraging growth and developing relationships with local business communities in both commerce and agriculture. His discussion at the April 2024 meeting covered the economic and growth strategies of Loudoun County, VA and their economic partners, as well as some of the holistic advantages of the area which ultimately led to the term "Data Center Alley."

US-based quantum computer (QC) supplier, PsiQuantum announced this week that it has entered into an agreement with the Australian Commonwealth and Queensland Government to build what they are calling the world's first utility-scale quantum computer, to be located at a new site in Brisbane, Australia. Funding consists of a mixed financial package of equity, grants, and loans totaling $940M AUD ($620M USD). The project schedule calls for the site to be operational by the end of 2027, with successive generations of PsiQuantum QC systems to follow.

The American Meteorological Society (AMS) recently hosted its 104th Annual Meeting in Baltimore, Maryland with the theme of "Living in a Changing Environment". The event consisted of 47 meetings and symposia on various topics involving the science, policy, and technology of climate change, oceanography, and meteorology. Among these was the 10th Annual Symposium on High Performance Computing (HPC) for Weather, Water, and Climate. HPC is considered essential to the efforts of weather, water, and climate organizations worldwide to provide accurate forecasts and advance scientific understanding of Earth's systems. A panel discussion including David Michaud (NWS), Laurie Carriere (NASA), Thomas Hauser (NCAR), Frank Indiviglio (NOAA), and Tiago Quintino (ECMWF) shed light on the shared challenges and opportunities faced by these organizations in leveraging HPC for weather prediction and climate modeling.

Established in November 2023, the DAOS Foundation was formed under the auspices of the Linux Foundation to advance the governance and development of the Distributed Asynchronous Object Storage (DAOS) project. The broad range of founding members includes Argonne National Laboratory (ANL), Enakta Labs, Google Cloud Platform (GCP), Hewlett Packard Enterprise (HPE), and Intel. Accordingly, the DAOS Foundation aims to accelerate next-generation technical computing for a broad range of workloads, including both traditional HPC modeling and simulation (mod/sim) as well as modern Artificial Intelligence and Machine Learning (AI/ML).

The EuroHPC JU recently issued a call seeking interest in identifying a consortia of private partners and hosting entities for the procurement of industrial high performance computers. The selection of a hosting site and identification of private partners is a precursor to subsequently implementing the procurement and operation of the machine, which may be an independent system or a partition of one of the EuroHPC supercomputers. The EuroHPC JU will provide up to 35% of the acquisition costs.

The HPC User Forum was established in 1999 to promote the health of the global HPC industry and address issues of common concern to users. In September 2023, the 83rd HPC User Forum took place in Tucson, Arizona. This update summarizes a presentation from that conference given by Dan Stanzione, Executive Director of Texas Advanced Computing Center (TACC) and Associate Vice President for Research at University of Texas-Austin. Stanzione presented TACC perspectives on datacenter sustainability including infrastructure improvements and incentivizing users to improve software.

Addressing multiple challenges facing the global HPC community, Microsoft has expressed intent to invest US$3.46B (€3.2B) in Germany over the next two years. Expanding its cloud region in Frankfurt am Main and planning new infrastructure in North Rhine-Westphalia, Microsoft expects the investment to double its existing cloud capacities in Germany, accelerate benefits of AI to German scientists, researchers, and engineers, and train more than 1.2 million individuals in digital skills.

According to a recent Japanese press release from the Institute for Molecular Science (IMS), a research institute corporation under Japan's National Institutes of Natural Sciences, a diverse group of Japanese commercial and academic quantum computing (QC) developers are forming a company, tentatively called the Commercialization Study Platform, to bring to market a next-generation high-speed quantum computer based on research developed at IMS's Omori Laboratory. Ten companies, including the Development Bank of Japan, Fujitsu, Hitachi, NEC, and Hamamatsu Photonics, are participating along with a number of smaller Japanese QC software firms. Each participant is expected to invest in the new company as well as provide human resources and technology. The stated mission of the company is the development of neutral atom quantum computer with intentions of producing a prototype in 2026 and targeting a commercial QC system by 2030. The final company name and terms of investment from corporate backers are not yet finalized.

In September 2023, the 83rd HPC User Forum took place in Tucson, Arizona. This update summarizes a presentation from that conference given by Ti Leggett, Deputy Project Director & Deputy Director of Operations at Argonne National Laboratory (ANL). Leggett presented the sustainability efforts of the Argonne Leadership Computing Facility (ALCF), including areas of success and challenges in the procurement, operations, and decommissioning processes.

The HPC ecosystem continues to evolve, providing endless avenues to explore and research. Items contributing to these market research opportunities include the rapid introduction and adoption of large language models (LLMs) and other AI-related innovations, increased adoption of HPC resources in the cloud, and heightened emphasis and awareness of sustainability and energy efficiency, to name just a few. The team of analysts at Hyperion Research strive to provide thoughtful, insightful, critical analysis across these topics, as well as the HPC market's many other dimensions. These dimensions include, but are not limited to, market data, technology, innovations, exascale deployments, and vertical application use cases. 2023 proved to be quite eventful, and 2024 is expected to be just as significant.

Issued in August of 2023, the EuroHPC JU's call for proposals for Innovation Action in Low Latency and High Bandwidth Interconnects closed on January 31, 2024. Intended to support and provide synergies with other EuroHPC I/O, processor, and accelerator programs, proposals were to: Develop a roadmap for European scalable inter-node interconnects targeting HPC exascale and post-exascale systems; Include the inter-node interconnect hardware, address design, development, testing and tape-out, as well as outline the integration into testbeds; Encompass the software, installation, configuration, and management tools for the developed interconnect, driven by the needs of relevant HPC workflows and application requirements; Address a broad range of system requirements (e.g., high bandwidth, low latency, power efficiency, virtualization, scalability, reliability, security).

In the past month, two quantum computing (QC) suppliers have announced acquisitions of specialized technology firms that provide components critical to the development and production of their respective QC product offerings, in this case both centered on photonic-based systems. US-based Infleqtion, a provider of both quantum technologies and commercial quantum computer hardware and software, announced it had acquired two integrated silicon photonics companies, SiNoptiq Inc. and Morton Photonics Inc. UK-based ORCA, developer of full-stack photonic QC systems, announced the acquisition of the US-based Integrated Photonics Division of GXC.

The North Atlantic Treaty Organization (NATO) recently released a public version of its first ever quantum strategy, outlining NATO's plans to foster a secure, resilient, and competitive quantum ecosystem that is able to respond to the fast pace of technological competition in the quantum sector. Specific goals included supporting development of key quantum technologies to maintain NATO's technological edge, implementing frameworks for hardware and software to enhance operability, and standing up a Transatlantic Quantum Community to strategically engage with government, industry, and academia across the NATO innovation ecosystem.

For 2024 and beyond, Hyperion Research makes these predictions for the worldwide HPC market:

4. Interest in procuring on-premises quantum computing (QC) systems will grow, augmenting but not replacing access to QC through the cloud.

Surpassing pre-pandemic attendance and achieving new attendance and exhibitor participation records, SC23 in Denver, CO did not disappoint. With over 14,000 on-site attendees and over 435 exhibitors, SC23 provided a vibrant, high-energy environment for the global HPC community to exchange ideas while establishing or extending collaborative partnerships and relationships. As is our custom, the Hyperion Research team of analysts has compiled its primary takeaways and perspectives from the event.

The HPC User Forum was established in 1999 to promote the health of the global HPC industry and address issues of common concern to users. In September 2023, the 83rd HPC User Forum took place in Tucson, Arizona. This update summarizes a presentation from that conference given by Doug Kothe, Chief Research Officer and Associate Labs Director for Advanced Science & Technology at Sandia National Laboratories. In addition to providing updates on the Exascale Computing Project (ECP) and key insights regarding the Frontier Supercomputer, he discussed the application portfolio, challenges, and goals for ECP.

The HPC User Forum was established in 1999 to promote the health of the global HPC industry and address issues of common concern to users. In September 2023, the 83rd HPC User Forum took place in Tucson, Arizona. This update summarizes a presentation from that conference given by Doug Kothe, Chief Research Officer and Associate Labs Director for Advanced Science & Technology at Sandia National Laboratories. In addition to providing updates on the Exascale Computing Project (ECP) and key insights regarding the Frontier Supercomputer, he discussed the application portfolio, challenges, and goals for ECP.

The use of clouds for HPC workloads over the last 12-24 months has undergone a decided shift. Conversations had largely centered around whether cloud-based resources could be reliably utilized at scale at performance levels required for complex HPC workloads. They have now evolved to dialogues seeking understanding of which workloads are capable of being migrated to the cloud to meet a user's cost, performance, security, and support requirements. This is inclusive of traditional HPC modeling and simulation (mod/sim) workloads and modern AI/HPDA/LLM workloads.    

This Hyperion Research study presents the latest five-year forecast (2022-2027) for HPC on-premises technical servers by vertical. Overall, the HPC server market is expected to grow at a 7.7% CAGR to reach $22.3 billion in 2027 as the market recovery continues and exascale system purchase and installations continue to ramp up in the Government Lab vertical.

Innovations are constantly occurring across all aspects of HPC infrastructure. Much of that innovation has been focused on the computing infrastructure:

• CPU performance once driven by faster clock speeds have evolved to integrating more cores.
• GPUs have emerged to advance performance of AI-related workloads.
• Application-specific accelerators are now being introduced to address unique elements of AI applications.

Introduced at Intel Innovation 2022 as a resource to private beta users, the Intel Developer Cloud was opened to the public as a beta resource at Intel Innovation 2023. The goal is to make new and future hardware platforms available for pre-launch development and testing to give developers, customers, and researchers early and efficient access to Intel technologies from a few months up to a full year ahead of product availability.

HPC in the cloud-related data and insights CSP Market Shares October 2023

HPC in the cloud-related data and insights related to Cloud vs On-Prem Workload Trends October 2023

HPC in the cloud-related data and insights related to Motivators and Limitations for HPC Cloud Users

This Hyperion Research study presents the latest five-year forecast (2022-2027) for HPC on-premises technical servers by competitive segment.

This Hyperion Research study presents the latest five-year forecast (2022-2027) for HPC on-premises technical servers by region.

This paper explores the evolution of HPC workloads and how cloud computing has evolved to become a viable tool to complement on-premises capabilities for running large-scale scientific workloads. It looks at the challenges driven by advanced, large-scale scientific workloads, the importance of their resultant research, how cloud-based resources are capable of meeting HPC requirements, and examples of results provided by running large-scale scientific workloads in the cloud.

The Historical View of the HPC On-Premises Server Market in Processors Installed, 2017-2022 is a summary of key insights gleaned from our QView database, the market tracking database that profiles HPC server sales and related information that goes back the past 20 years.

The Historical View of the On-Premises HPC Server Market by Competitive Segment, 2017-2022 is a summary of key insights gleaned from our QView database, the market tracking database that profiles HPC server sales and related information that goes back the past 20 years.

The Historical View of the HPC Server Accelerators Market, 2017- 2022 is a summary of key insights gleaned from our QView database, the market tracking database that profiles HPC server sales and related information that goes back the past 20 years.

The Historical View of the HPC Server Market by Vendor, 2017- 2022 is a summary of key insights gleaned from our QView database, the market tracking database that profiles HPC server sales and related information that goes back the past 20 years.

CIQ, Oracle, and SUSE have partnered to form the Open Enterprise Linux Association (OpenELA), which focuses on community-driven source code for Enterprise Linux. The trade association aims to encourage the development of Linux distributions that are both compatible with Red Hat Enterprise Linux (RHEL) and can uphold open community standards. This arrangement formalizes what the members consider to be the existing de facto standards of transparency, community involvement, security, free access, open redistribution, and expedient updates.

The EuroHPC Joint Undertaking (EuroHPC JU), the EU's main governmental body charged with overseeing the EU's HPC ecosystem, has issued a call to engage with small and medium enterprises (SMEs) offering support for competitiveness and innovation. The primary objective of this endeavor is to empower SMEs with advanced HPC capabilities to drive innovation, enhance competitiveness, and overcome challenges in the digitization of R&D and business processes.

atNorth has made the 2023 shortlist for two categories of the UK National Sustainability Awards. This is the second annual National Sustainability Awards, which recognizes organizations within all sectors that are innovating to create a more sustainable and better future. atNorth made the shortlist for the 'Energy Innovation' and the 'Building of the Year' categories. The Energy Innovation Award nomination was for the bespoke direct liquid cooling (DLC) system developed for the atNorth SWE01 datacenter facility in Stockholm, Sweden, in collaboration with CoolIT, a Canadian-based DLC supplier. The DLC and warm water combined system at SWE01 allows for higher rack density and higher peak performance, all while using less power.

Recently inaugurated at the NCAR-Wyoming Supercomputer Center (NWSC) in Cheyenne, WY, the Derecho supercomputer is a result of broad collaboration between and support from the state of Wyoming, the University of Wyoming, the University Corporation for Atmospheric Research (UCAR), the National Center for Atmospheric Research (NCAR), the National Science Foundation (NSF), the White House Office of Science and Technology Policy (OSTP), and HPE. Its capabilities will enable scientists and researchers to better model and understand earth science processes to develop environmental intelligence in support of key societal risk management areas against natural events such as hurricanes, wildfires, and floods.

In an announcement made in early August 2023, IBM announced the release of their joint project with NASA, watsonx.ai geospatial foundation model, as open-source on Hugging Face. The model, which is the largest of its kind available in the open-source AI development environment, is also described in the press release as the first open-source AI foundation model built in collaboration with NASA. As part of the Space Act Agreement with NASA, IBM began construction on the model earlier this year to address obstacles the aerospace association faces in analyzing tremendous and growing amounts of climate and geospatial data.

This Hyperion Research study presents the latest five-year forecast (2022-2027) for HPC on-premises technical servers. Worldwide revenue for the HPC technical server market in 2022 was $15.4 billion, representing a modest growth (3.9%) over 2021 revenues. Hyperion Research now predicts that the HPC technical server market will grow at a 7.7% CAGR between 2022 and 2027 to reach $22.3 billion in 2027. Compared to the previous version of the forecast, there have been some slight downgrades to the revenues for 2022 through 2025, primarily due to rescheduling of some large exascale systems.

The QView market tracking database includes profiling on all of the HPC server sales for the past 20 years. This report focuses on historical data and trends for the HPC on-premises technical servers by region over the last five years (2017-2022). The QView has recently expanded to incorporate China as a major region. Now the six regions represented in the QView are: North America, EMEA, Japan, China, APAC (without Japan and China), and the rest of the world. As seen below in Figure 1, North America contributed less than half (44.7%) of the worldwide HPC server revenue in 2022, with EMEA contributing just under a third (28.0%).    

The newly-formed Ultra Ethernet Consortium (UEC) seeks to establish an open, interoperable, high-performance architecture leveraging the broad Ethernet ecosystem. With founding members including AMD, Arista, Broadcom, Cisco, Eviden, HPE, Intel, Meta, and Microsoft, the UEC envisions a system interconnect solution that will capitalize on the ubiquity and flexibility of Ethernet to address demanding heterogenous HPC and AI performance, functionality, and scalability requirements.

In a press release from August 4, 2023, Qualcomm Technologies, Inc. announced the formation of a company aimed at the acceleration and future of products based on the open-standard RISC-V instruction set architecture (ISA). Formed with Robert Bosch GmbH, Infineon Technologies AG, Nordic Semiconductor, and NXP Semiconductors, this as of yet unnamed company will also provide reference architectures and help establish industry-wide solutions. The initial focus of the effort will be in the automotive sector, with expectations for expansion to mobile and IoT.

The global on-premises HPC broader market closed 2022 at $30.8B. This represents a 3.6% growth rate over 2021 and a 5.1% CAGR from 2017-2022. The modest growth over 2021 is largely a result of 2021 being such a high growth year, along with global economic uncertainties in 2022. The overall market from 2017-2022 experienced some major inconsistencies, such as supply chain issues caused by the covid-19 pandemic on the downside and the acceptance of what are now the two systems atop the current Top500 list, Frontier and Fugaku.

Recent study results indicate that a reasonable percentage of respondents' overall annual HPC budget goes toward HPC-related third-party professional services. Overall, about half of all respondents spend somewhere between 10% and 50% of their overall HPC budgets on HPC-related professional services.

• Industry was much more committed to professional services than the other major sectors, with 27% of industry survey respondents reporting commitment of 10% to less than 20% of their budgets to HPC-related professional services and about one-quarter of industry sites committing between 20% and 30%.
• More than one-third of government respondents report spending between 10% to less than 20% for HPC professional services.
• The academia sector was the least dependent on professional services, with 40% reporting less than 5% of their annual budgets going to HPC-related professional services.

The Barcelona Supercomputing Center-Centro Nacional de Supercomputación (BSC-CNS)1 has long been recognized as one of the leading scientific research and computational science centers in Europe, if not the world. With its mission to "…research, implement, manage, and transfer technology and knowledge in the area of HPC with the aim of facilitating progress in a variety of scientific fields…", BSC-CNS has made significant research and scientific contributions to its primary focus areas of computer, life, earth, and engineering sciences.

For the largest HPC system at the surveyed sites, the primary processor type for future system purchases was x86, at 85.2% of the sites, led by Intel (68.0%) and AMD (14.4%). The next largest identified primary processor was the IBM Power or OpenPower chips at 3.9%. Despite its growing visibility in the overall HPC sector, Arm-based components were only a small percentage of the overall processor space at 3.9%. This data is from an annual study that is part of the eighth edition of Hyperion Research's HPC end-user-based tracking of the HPC marketplace. It included 181 HPC end-user sites with 3,830 HPC systems.

In June, at the first ever RISC-V Summit Europe, IBM and the Barcelona Supercomputing Center (BSC) signed a partnership agreement devoted to enhancing and developing advanced technology under their "Future of Computing" initiative. While the initiative is primarily aimed at advancing collaboration between IBM and BSC, it will also contribute to both regional technology capabilities and the overall advanced computing missions of the EU.

At their annual HPE Discover conference, HPE announced its entry into the public AI cloud market with their HPE GreenLake for LLMs (large language models) supercomputer as a service offering. HPE GreenLake for LLMs aims to:

• • Make HPE's leadership-class AI-native supercomputer infrastructure more accessible to users developing and leveraging LLMs for their business operations.
  • Deliver turnkey LLMs supporting industry and domain-specific AI applications.
  • Strengthen global sustainability and carbon-reduction initiatives partnering with colocation facilities running on nearly 100% renewable energy.

The HPC User Forum was established in 1999 to promote the health of the global HPC industry and address issues of common concern to users. In April 2023, the 82nd HPC User Forum took place in Princeton, New Jersey. This update summarizes a presentation from that conference given by Molly Peeples, tenured associate astronomer at the Space Telescope Science Institute and astrophysics theorist. Also, a research scientist at Johns Hopkins University, Peeples leads the Figuring Out Gas & Galaxies In Enzo (FOGGIE) collaboration. Her discussion at the April 2023 meeting covered some of the recent laboratory activities as well as milestones and her vision for the future. Additional topics included recent activities in the FOGGIE collaboration, simulation domain factors and their relationship to compute demands, as well as the technological future of the field of galaxy simulation.

A part of the annual Hyperion Research HPC Multi-Client study (MCS) tracks the various application frameworks respondents use on their HPC systems to discover trends in user preference for the way applications are built and deployed, as well as trends in the functionalities users desire in their solutions and products. This year's iteration of the MCS found the most popular AI/ML/DL and data intensive frameworks were TensorFlow, PyTorch, Jupyter Notebook, SQL, CUDA, Hadoop, and Spark. This data is from an annual study that is part of the eighth edition of Hyperion Research's HPC end-user-based tracking of the HPC marketplace. It included 181 HPC end-user sites with 3,830 HPC systems.

HPE recently announced at their Discover 2023 conference keynote an outpouring of GreenLake cloud-centered offerings, including a partnership with AWS that promises a more integrated hybrid cloud experience. The goal of the HPE GreenLake/AWS partnership is simplicity for users, provided through a seamless dashboard that offers control of both cloud and on-premises resources. The specifics include features such as increased security and expanded container deployment options for Kubernetes in hybrid environments. HPE emphasized that their partnership will support various users (enterprise and HPC alike), from consulting with new adopters to managing diverse applications and workloads.    

Much attention is paid to the compute elements of HPC architectures, and rightly so. However, if those elements are not appropriately connected in a balanced and performant manner to other servers and to storage, then the optimizations provided by the server technologies will be unrealized as new bottlenecks appear. While there are multiple approaches for compute and storage interconnect architectures, recent survey data suggests the prevailing architecture is expected to shift from a preference of independent node-node and node-storage networks to a preference of converged networks with respondents' next HPC procurement. This data is from an annual study which is part of the eighth edition of Hyperion Research's high-performance computing (HPC) end-user-based tracking of the HPC marketplace. It included 181 HPC end-user sites with 3,830 HPC systems.

As utilization of HPC resources in the cloud continues to grow, users are turning to multiple CSPs to satisfy their HPC cloud resource requirements. Sites surveyed in this study who indicated they utilized HPC resources in the cloud employed an average of approximately two CSPs for their HPC workloads. Reasons for employing a multi-cloud strategy include:

• Gaining access to specialized hardware only available on one CSP platform, like the TPU from Google or the Trainium processor from AWS
• Accessing specific tools or services
• Leveraging a dataset hosted on a specific cloud platform
• Collaborating with other researchers who utilize different platforms

When respondents were asked to identify specific system attributes for which they would be willing to pay a 10-15% premium on top of system price, the most desirable attributes were faster or higher performance processors (48%), and larger or faster memory (39%). More than one third of respondents were willing to pay a premium for higher performance external I/O and storage interconnects between nodes (35%) and better density, power, or cooling attributes (32%). In contrast, only a small percentage (9%) expressed a willingness to pay a premium for a specific vendor, and about one in ten indicated that they would not be willing to pay a 10-15% premium for any specific attribute/feature. This data is from an annual study that is part of the eighth edition of Hyperion Research's HPC end-user-based tracking of the HPC marketplace. It included 181 HPC end-user sites with 3,830 HPC systems.

The HPC User Forum was established in 1999 to promote the health of the global HPC industry and address issues of common concern to users. In April 2023, the 82nd HPC User Forum took place in Princeton, New Jersey. This update summarizes a presentation from that conference given by William Dorland, Princeton Plasma Physics Laboratory's (PPPL) first associate laboratory director for computational science. A renowned computational physicist, Dorland leads the effort to fulfill PPPL's expanded mission to develop computational science as a core capability that will provide high-performance computing support to understand and predict fusion plasma physics, design fusion facilities, and simulate complex plasma phenomena. His discussion at the April 2023 meeting covered some of the recent laboratory activities as well as milestones and his vision for the future.

According to recent study results, a quarter of HPC users (28.2%) either currently employ edge computing or expect to within two years. The top motivators driving edge computing growth include improving real-time data collection and processing, accelerating HPC applications, access to IoT devices for data collection, and a wider range of sensor data. Top deterrents dampening edge computing growth include complex and varied IoT formats, inadequate edge vendor support, lack of in-house expertise for integration, and the cost of integrating into existing infrastructure. This data is from an annual study that is part of the eighth edition of Hyperion Research's HPC end-user-based tracking of the HPC marketplace. It included 181 HPC end-user sites with 3,830 HPC systems.

In a recent study, respondents reported that only about half of all their HPC applications (46.8%) run on multiple nodes, leaving half of the applications running on a single node or less. Looking at the top two applications per site, two-thirds (68.9%) were run on multiple nodes. This suggests that the typical HPC site has at least several important, large-scale applications, and a very large set of single node jobs. This data is from an annual study that is part of the eighth edition of Hyperion Research's HPC end-user-based tracking of the HPC marketplace. It included 181 HPC end-user sites with 3,830 HPC systems.

While innovations associated with HDD capacities, SSD performance, latency, bandwidth, and throughput advancements dominate HPC storage-related investments and headlines, tape-based storage solutions remain a critical part of data center storage architectures. Tape continues to be utilized at roughly half of the sites surveyed, with average tape capacity at those sites being approximately 2.5x greater than average disk (both HDD and SSD) capacity across all surveyed sites. This data is from an annual study which is part of the eighth edition of Hyperion Research's high-performance computing (HPC) end-user-based tracking of the HPC marketplace. It included 181 HPC end-user sites with 3,830 HPC systems.

Hyperion Research recently undertook an effort to gather information on currently available mechanisms for existing and potential quantum computing (QC) end users to connect to a quantum system through a cloud access model. The effort considered the widest range of currently available options that spanned direct QC access through a specific vendor, such as D-Wave, IBM, and Rigetti, QC cloud access capabilities offered through cloud service providers (CSPs) such as Amazon Braket, Azure Quantum, and Google Cloud, as well as curated access offered by third-party players such as QC Ware, Strangeworks, and Zapata.

Sustainability is the societal goal of existing on this planet in a way that meets the needs of today without compromising the needs of future generations. Sustainability in terms of the HPC market can be described as meeting the needs of present users without compromising the needs of future users, with a focus on the extraordinary power requirements of HPC systems. Over the past few years, Hyperion Research has been tracking the growing influence the goal of sustainability has had on HPC procurement decision-making at various HPC sites around the world. HPC site managers/directors have had to consider the current global energy crisis and geopolitical climate alongside the HPC market, and begin planning and implementing ways to make HPC operations at their sites more sustainable.

Key findings from a recent Hyperion Research study showed that accelerated applications are growing as percent of workload, especially in the industry sector. In addition, many HPC users reported that they have or plan to migrate away from CentOS in response to the changing support model. Insights into the critical factors driving these and other trends are detailed in the 2022 iteration of Hyperion Research's annual MCS end users' study, 2022 HPC Multi-Client Study: Vertical/Application Workload Areas and HPC System Software and Middleware. Key findings from the report are summarized in this document.

Building upon the success of the initial phase of its Centres of Excellence (CoE) program, the EuroHPC JU (European HPC Joint Undertaking) recently announced the launch of ten new Centres of Excellence to support research and innovation targeted to develop and adapt HPC applications for the exascale and post-exascale era. The domain-area CoEs will work collaboratively with EU National Competency Centers focused on country-specific resources and research initiatives.

In early February 2023, pure-play quantum computing (QC) firm Rigetti Computing Inc announced the layoff of 50 employees, 28% of its staff, fueled in part by an effort to stave off a delisting of its stock from the Nasdaq exchange. To avoid that, Rigetti must maintain a per share price of above $1 for ten consecutive days before a July 24th deadline. The announcement, coupled with a reshuffling of both CFO and CTO positions, also highlighted a revised technology roadmap for Rigetti that concentrates on the delivery of its Ankaa 84 qubit system in the first quarter of 2023, with increased 2-qubit gate fidelity, capable of demonstrating quantum advantage on a practical, operationally relevant problem over counterpart classical solutions.

While barriers to cloud adoption persist in some areas, many have been overcome, and users are moving beyond surge and experimentation use cases to an increasing number of production workloads. The cloud market for HPC continues to grow at a strong pace as cloud service providers (CSPs) address the needs and concerns of HPC users as well as the on-going education of users on how to optimize HPC workloads for the cloud. Insights into the critical factors driving these and other trends are detailed in the 2022 iteration of Hyperion Research's annual MCS end users' study, 2022 HPC Multi-Client Study: Use of Public/External Clouds for HPC Workloads, Trends, and Drivers. Key findings from the report are summarized in this document.

For 2023 and beyond, Hyperion Research makes these predictions for the worldwide HPC market:

1. Strong growth in the leadership-class segment will support modest growth across the global on-premises HPC market.

2. The advanced computing sector and its associated supply chain will become increasingly driven by national and regional government policies that stress domestic capabilities.

3. Sustainability and energy efficiency considerations will become a dominant factor in many procurements.

4. Cloud utilization will shift from predominantly experimentation to predominantly production workloads as users gain familiarity and confidence with cost, performance, and workflow integration expectations. This will include the initial erosion of on-premises spending in the low end of the HPC market.

5. 2023 will be the year of AI regulation.

Key findings from a recent Hyperion Research study indicate that HPC storage solutions, and associated storage and system interconnects, continue to be critical for HPC infrastructure to deliver optimal capabilities and provide the fastest time to results for the systems' users. Data-intensive workloads driven by new AI/ML/DL workloads, increasing scale of traditional HPC modelling and simulation, emerging edge computing, and emerging composable systems are placing greater demands and requirements on HPC storage systems. Insights into the critical factors driving these and other trends are detailed in the 2022 iteration of Hyperion Research's annual MCS end users' study, 2022 HPC Multi-Client Study: Trends and Forecasts in HPC Storage and Interconnects. Key findings from the report are summarized in this document.

Key findings from a recent Hyperion Research study revealed that HPC users see continued dominance of the x86 processor in HPC, led by Intel but with a growing interest in AMD, the nearly universal adoption of coprocessor or accelerators in HPC servers, albeit at low overall systems counts, and modest expectations for HPC budgets growth for the bulk of the sector. Insights into the critical factors driving these and other trends are detailed in the 2022 iteration of Hyperion Research's annual MCS end users' study, 2022 HPC Multi-Client Study: Processors, Coprocessors/Accelerators, and HPC Budgets. Key findings from the report are summarized in this document.

In a signed agreement published in December of 2022, the EuroHPC Joint Undertaking (EuroHPC JU) called for proposals to fund a €270M effort to develop a European HPC infrastructure built on the RISC-V open ecosystem. The effort also includes fostering an overall open-source RISC-V community, code porting capabilities, and the development of chiplets based on the RISC-V open instruction set architecture (ISA). RISC-V stands apart from other ISAs as an open-standard that is free to license and has no IP requirements. RISC-V currently has a few offerings in the base product but has clear potential in meeting flexibility and customizability needs over proprietary chip counterparts.

Data collected in the most recent iteration of the Hyperion Research Global End-User Multi-Client Study indicates a strikingly low usage of AI-specific software licenses across all sectors, especially academia and government. This research revealed that only 3.5% of academic and government respondents, combined, were currently using licensed AI software, while the entire survey group exhibited a usage rate of 13.8%. Respondents representing academic sites were the least likely to answer positively in paying for AI software licenses, with a rate of 1.9%, and industry were the most likely at 21.1%. This data indicates an overwhelming preference for open source and in-house solutions for AI software, including training analytics and library interface tools, over proprietary AI[1]specific software offerings currently on the market.

US quantum computing (QC) supplier IonQ recently announced plans to build what it is calling the first US-based QC manufacturing plant. The 65,000 square foot facility, scheduled for completion in the first half of 2024, will be located outside Seattle, Washington and will most likely be targeted for the firm's next-generation 32-qubit Aria quantum computer. The facility is also slated to become IonQ's second internal data center, but there were no announced details whether this new center will either supplant or augment IonQ' s existing cloud service provider relationships with Azure Quantum, Amazon Bracket, and Google Cloud, that currently support access to IonQ systems.

With over 11,000 on-site attendees approaching pre-pandemic levels, and only 700 virtual participants, SC22 in Dallas, TX far exceeded the high expectations of the broad HPC community. There was no shortage of avenues for participants to obtain the latest knowledge relative to market developments and technology innovations occurring across the industry. The Hyperion Research team of analysts has compiled its primary takeaways and perspectives from the event.

During the recent SC22 event in Dallas, Texas, RISC-V Foundation Chairman Professor Krste Asanović from the EECS Department at UC, Berkeley gave a presentation detailing the technological and market position of RISC-V and its potential to become a leading ISA, supplanting the slate of proprietary ISAs that currently dominate the sector. According to Asanović, it is only a matter of time before the RISC-V instruction set architecture claims its rightful place in the HPC stack among other industry standards like Ethernet, Posix, or SQL.

During the recent Wells Fargo 2022 TMT Summit, Mark Papermaster, CTO of AMD, reported that the Instinct MI300 accelerated processing unit (APU) is, as of early December 2022, up and running. Currently confined to their in-house lab, the Instinct MI300 will be used in the exascale supercomputer currently in development at Lawrence Livermore National Laboratory, El Capitan, scheduled to be delivered in 2024. Described by Papermaster as a "true datacenter APU," AMD expects general availability of the processor in 2023. This multi-chiplet processor makes use of both AMD's Zen 4 (x86) CPU architecture and CDNA 3, AMD's GPU architecture, designed specifically with exascale computing in mind, and will be produced by Taiwan's TSCM at the 5nm process node. Papermaster sees this development as an important way to continue introducing greater density and optimization into components now that the sector is no longer in the era of the "old Moore's Law."

The US National Oceanic and Atmospheric Administration (NOAA) and Microsoft have entered into a Cooperative Research and Development Agreement (CRADA), formalizing NOAA's commitment to using Microsoft Azure cloud computing resources in the pursuit of NOAA's mission to build a Climate[1]Ready Nation by 2030. Several initiatives are envisioned whereby NOAA scientists and engineers will work with Microsoft experts to leverage Azure's machine learning and HPC capabilities: ▪ Fast-tracking innovative contributions to NOAA Earth Prediction Innovation Center (EPIC) earth systems modeling and research ▪ Applying machine learning capabilities to improve models supporting air quality, smoke, and particulate pollution forecasts, as well as relevant NOAA climate models ▪ Accelerating NOAA Fisheries' survey and observations data collection and management ▪ Creating new ocean observations cataloging efforts ▪ Designing resilient and accessible weather modeling and forecasting that can incorporate external data sources with NOAA enterprise data

Due to a shift in HPC sites' perception of HPC cloud resources available and in future roadmaps from CSPs, the HPC cloud market continues to be one of the strongest growth markets in today's HPC ecosystem. While the HPC on-premises market is not projected to dramatically change from the steady 6%-8% five-year CAGR seen in years past, the cloud market is currently showing a projected 17.6% five-year CAGR for the next five years.

The European High-Performance Computing Joint Undertaking (EuroHPC JU) recently announced the selection of six sites to host what could be the first major round of EU government-sponsored quantum computer (QC) procurements. The new QC systems will be integrated into leading EU-based HPC sites including the IT4Innovations National Supercomputing Centre in Czechia, the Barcelona Supercomputer Center in Spain, and Cineca in Italy, as well as key HPC sites in France, Germany, and Poland. The QC systems will be networked to support EU-wide access by academic, commercial and government research facilities. Half of the 2023 €100 million procurement budget will come from the EU and the remainder from the 17 countries participating in the EuroHPC JU. The QC hardware and software for this effort will draw exclusively on EU technology developed under EU-funded quantum initiatives, related national programs, and private investments.

According to a recent survey of over 200 HPC sites worldwide, the on-premises and cloud HPC markets are expected to grow over the next five years, although at slightly lower rates than previously anticipated earlier in 2022. Future cloud budgets for HPC workloads are expected to grow at a faster rate than the on-premises market for industry and academic user sites, while government sites are anticipating higher on-premises budget growth.

IBM recently announced that it is combining the Red Hat storage and associated teams with the IBM Storage business unit. The combination aims to bring consistent data storage solutions across on-premises and cloud infrastructures to deliver a unified storage experience regardless of file type.

Catalog and Seagate Technology recently announced a collaboration to advance DNA-based technology towards becoming commercially viable storage and computing solution. Catalog brings their molecule designs for storing data in DNA and performing computation across a library of molecules to the partnership, while Seagate Technology will be contributing its silicon-based lab-on-a[1]chip technology to reduce the volume of chemistry required for DNA-based storage and computation.

During an inaugural ministerial meeting of the U.S.-Japan Economic Policy Consultative Committee (EPCC) in July, a joint statement was presented detailing a renewed and explicit commitment to regional economic stability, fairness, and hardiness. The statement, which includes an action plan, enumerates four main goals: realizing peace and prosperity through rules-based economic order, countering economic coercion and unfair opaque lending practices, promoting and securing critical and emerging technologies and critical infrastructure, and strengthening supply chain resilience. While renewed and steady efforts to maintain regional welfare are an end within themselves, this joint statement takes on an additional layer of complexity and purpose when considered in light of the recent U.S. CHIPS Act, a semiconductor promotion policy whose U.S.-only tone has the potential to cause regional turbulence and heighten international trade tensions.

In mid-June of 2022, Cerebras Systems announced a new feature that allows users to train some of the largest AI models in the world within a single CS-2 machine using a simplified software support scheme. The announcement highlights multiple capabilities that Cerebras sees as their competitive advantages over other companies. Notable examples cited include the ability to accommodate an entire training model within the memory, through Cerebras' Weight Streaming software on the Wafer Scale Engine (WSE), instead of splitting it across processors, as well as the ability for users to manipulate a few inputs within the software scheme and GUI to choose the scale of model desired for training (i.e., GPT-3 13B, GPT-3XL 1.3B). Cerebras claims that this advancement can cut down the setup of large model training runs from months to minutes, with the Cerebras software managing much of the initial setup.

The development of AI and HPC hardware is currently undergoing a fundamental shift, with many countries recognizing the negative ramifications that can occur when supply chains and technology dependence is based on international relations. A growing urgency for indigenously developed technology has become a staple in many ten-year technology roadmaps worldwide. At the recent Hot Chips 34 conference, Chinese firm Biren Technology announced a new GPU product line, the BR100, one of the first such devices developed in China and targeted for widespread HPC and AI markets. The Chinese advanced computing sector has been an increasingly sensitive geopolitical and economic topic over the past decade, with some domestic IT system suppliers experiencing heavy restrictions from the United States government on access to advanced computing technology from US sources. The announcement of the Biren GPU is a step toward China improving their indigenous HPC and AI capabilities from a hardware perspective, a progressive move targeted to reduce China's dependence on US-based providers including AMD, Intel, and NVIDIA.