HPCs Are the Tip of the Iceberg
Originally published in The New Stack Update.
HPE is by far the leading manufacturer associated with the world’s fastest, commercially available computing systems in the most recent TOP500 list. After them, four of the next six companies are Chinese (Lenovo, Sugon, Inspur, Huawei). So what? Supercomputers can be super boring, but the market is relevant because it shows where hardware technology companies are investing their R&D budgets. Since R&D leads to long-term corporate success, it is not surprising that Sugon and Inspur have recently jumped onto IDC’s server market leader boards. But servers is a dying market, right? It is shrinking, but it’s also strategically important for 1) economic growth and 2) its connection to the fast-growing demand for artificial intelligence.
Machine learning (ML) and artificial intelligence (AI) are now common use cases for high-performance computing (HPC). What was once the domain of academia, operating scientific workloads is moving beyond a niche ability. Of course, many ML and AI workloads can be handled at the edge, or at least don’t need HPC. However, the model in which these workloads can be off-loaded to a third-party cloud service is something we hear about often. Intel, Nvidia and ARM are focused on optimizing hardware for AI and ML. Along with the big cloud providers, The New Stack expects the supercomputer manufacturers to be among the leading buyers of a new wave of CPU innovation.