Opinion Recent conversations with industry peers highlight growing AI demand juxtaposed with a perceived decline in HPC. This article argues against the notion that AI is overshadowing HPC investment.

My assertion, here on The Reg, is that the idea of AI growth stifling HPC investment is unfounded because AI is a subset within the expansive HPC landscape.

HPC refers to a diverse range of demanding workloads across sectors like finance, pharmaceuticals, and manufacturing, warranting significant investment. AI, gaining visibility and attention, is a critical component of HPC, sharing many attributes and demands.

Both AI (including machine learning and generative models) and traditional HPC are complex, needing robust infrastructures. A laptop could theoretically run AI models, but the performance and relevance would be inadequate. A high-performance setup is essential for both AI and HPC.

There is a relentless drive toward greater accuracy and solving bigger challenges in both AI and HPC domains. This includes augmenting datasets for machine learning or expanding parameters for language models.

Some argue that AI, with its reliance on accelerators like GPUs, differs from traditional HPC applications. However, HPC is defined by performance demands rather than specific technologies.

In the future, AI will seamlessly integrate into applications, enhancing areas from drug discovery to manufacturing. New domains, such as personalized health and optimized agriculture, will also require HPC-like infrastructures.

HPC’s Continuing Relevance

Existing applications will persist, complemented by AI, necessitating high-performance solutions.

The market for HPC and AI solutions will continue to expand, driven by the need for datacenters to adapt to AI demands cost-effectively. The real challenge lies in meeting electrical and cooling capacity needs.

Current estimates suggest racks demand increasing power, especially with AI advancements. Datacenters should evaluate current systems for efficiency improvements, which could yield significant energy savings.

As AI infrastructure demands rise, cooling capacity analysis becomes critical. Transitioning to liquid cooling may offer cost-effective solutions and operational benefits.

Ultimately, embracing IT efficiency and continuously measuring effectiveness are crucial for managing dynamic conditions and ensuring optimal resource use.