Hyperion: HPC Community’s Interest in LLMs Has ‘Exploded,’ with Complexity, Cost Concerns

Print Friendly, PDF & Email

HPC-AI industry analyst firm Hyperion Research said its new study on Large Language Models in the HPC community shows that interest in LLMs has exploded in the last six months driven by unique capabilities of the technology to answer queries, generate concise summaries, and even produce unique works of fiction.

The recently completed study, Large Language Models: Finding Their Place in the HPC Ecosystem, was conducted last July and August and collected insights from 100 respondents who indicated that their organization was currently involved in or planning to use LLMs within the next 12 to 18 months to support current or planned HPC-based workloads. Respondents came from a mix of sectors: industry (63 percent), academic (23 percent) and government (14 percent).

But LLM’s are only one part, albeit one of the most recent and attention grabbing, of the wide ranging field of AI, which depends heavily on advanced AI-centric computer systems. Hyperion Research forecasts that the purchase of systems dedicated for AI use will grow at over 30 percent a year at least out to 2027.

“Within the HPC community, many organizations are eager to explore the opportunities that LLMs can bring to the advanced computing space, offering tangible gains in computational capability for key workloads,” Hyperion said in its announcement. “LLMs are already considered to be an important emerging technology for both current and planned HPC-related activities.”

Study highlights include:

  • LLMs are viewed as already having widespread benefits to organizations.
  • Numerous LLM applications are currently being considered, and many organizations are looking at multiple options.
  • Organizations seeking to leverage LLM capabilities do, however, see challenges with the complexity of integrating LLMs into existing HPC-based workloads as well as concerns with the cost of LLM-specific hardware or software.
  • The majority of surveyed organizations are willing to increase their computing budget to support LLM inclusion.
  • Survey respondents are looking for a broad range of LLM expertise to support the various stages of LLM development.

“Generative AI and LLMs in particular show promise in being able to bring a spate of new capabilities to the overall AI space in key areas including content creation, language translation, and even code development,” said Bob Sorensen, Hyperion SVP of Research. “For their part, HPC end users are interested in LLM’s to help accelerate some of their most demanding computational problems in fields as diverse as finance, bioscience, and advanced manufacturing. Exploration of LLM capability has already begun in earnest across a wide swath of HPC sites. Keeping apace of the rapid development in this field will be complicated but necessary for any organization seeking to make the optimal decisions about the hows, whens and whys of including LLM’s as part of their computational arsenal.”

Hyperion Research has been tracking and studying the AI and HPDA markets for over 10 years and has now expanded its AI/HPDA subscription service to include new and emerging AI technologies and use cases. These use cases represent a broad range of application areas, from the life sciences, manufacturing, finance, government research and even within the legal profession.