Takeaways from Supercomputing Conference 2025

There was no shortage of insight into the future of AI and the infrastructure needed to support it at Supercomputing Conference 2025 (SC25). Here's what stood out to some of our experts who attended the international event for high-performance computing (HPC), networking, storage and analysis.

Takeaways from Earl Dodd

Earl Dodd leads WWT's Global HPC Business practice

A major theme of SC25 was that HPC and supercomputing are evolving into innovation engines for IT. The conference showed how they serve as research and development labs for tomorrow's enterprise architectures, patterns and solutions.

Many foundational technologies (e.g., cluster computing, high-speed fabrics, GPUs, parallel filesystems and advanced workflow orchestration) were proven at extreme HPC scale 5–20 years before becoming indispensable in enterprise IT, especially in the era of big data and AI.

The conference also emphasized how the line between HPC and the enterprise is fading. Data-heavy workloads, AI, real-time analytics and simulation are closing the old divide. Supercomputing environments now preview architectures, cooling, memory and scheduling strategies that often become enterprise standards within three to eight years, down from the historical 10- to 20-year lag.

Additionally, SC25 showed how HPC is the technology "time machine" and "Formula 1" for IT. For example, technologies stress-tested in national labs are commoditized via cloud hyperscalers, accelerating their adoption and shortening the enterprise onboarding cycle. Relatedly, batch orchestrators (e.g., SLURM/PBS), cluster approaches, and accelerators migrate into managed cloud, SaaS and composable enterprise platforms when market forces align.

Outside of this major theme of IT innovation, the event provided key enterprise-ready HPC trends. Expect to see advances in the following areas in the coming years.

• Direct liquid cooling (DLC): Chip power densities demand liquid cooling in enterprise data centers — a decade-old HPC norm— especially for AI-centric hardware.
• Disaggregated/composable memory (CXL): Memory pools shared across servers will solve the virtualization and AI "memory wall" with CXL-based platforms from Dell, HPE and cloud hyperscalers arriving within three to five years.
• AI-centric scheduling and workflows or "Cluster as the Computing Unit:" AI and analytics workflows require orchestrators with HPC's scale-awareness—now native in platforms like Ray, Databricks and next-gen Kubernetes.
• Scaling accelerators: Next-gen GPU/AI/ML silicon, hybrid on-premises and SaaS acceleration, ultra-low-latency fabrics and quantum simulators are already enterprise-available or imminent.
• Partial convergence: Exascale and quantum remain lab/boutique specialties. Most enterprises will consume these via cloud platforms, not through direct operation.

Takeaways from Jeff Fonke

Jeff Fonke is the director of WWT's High-performance Architecture practice. 

The SC25 conference in St. Louis was a remarkable event for the global HPC, AI, networking and analytics community, and WWT made a significant impact. Building off SC24, innovation, research, education and AI were at the forefront of the show. 

Keynote speaker Thomas Koulopoulos captivated the audience with his talk, "Gigatrends: The Exponential Forces Shaping our Digital Future," which explored the forces driving change in the 21st century. The talk's focus on "agility over perfection" resonated with me as it's why we built the AI Proving Ground. It's critical for organizations to iterate, learn and adapt to the massive technology wave we're seeing specific to AI.

Our booth featured a range of interactive demonstrations, including WWT's generative-AI chatbot, Atom.  In addition, we hosted a variety of flash-talk sessions that explored real-world workflows, including discussions about the transformation of hybrid and multi-cloud environments through orchestrated pipelines.

Beyond the exhibit floor, executive sessions provided a platform for discussing the future of AI and HPC, while presentations by WWT teams offered insights into the intersection of AI, HPC and workflow orchestration. We also held a memorable restaurant takeover at Buddy's Wine Bar, where attendees could network and unwind. 

Overall, WWT's participation in SC25 underscored our commitment to innovation and collaboration in HPC and AI solutions. With such a packed event, we'll be sharing more insights in an upcoming episode of the AI Proving Ground Podcast.

Takeaways from Liz Gattra

Liz Gattra is a senior manager with WWT's AI practice

SC25 was a solid event, and I was glad to be part of it. Engagement was stronger than I expected for my sessions, and the follow-up conversations were genuinely useful. Several partners and clients stopped me afterward to dig deeper into prompting, which reinforced that the interest is real even in a heavily HPC-focused environment. 

It was also a good chance to reinforce what WWT can actually do. There are still people who think of us mainly as a reseller or hardware shop, so being able to demonstrate our depth in AI and how far our capabilities go beyond the infrastructure layer mattered. The variety of presentations helped make that point more clearly than a single session could.

Networking throughout the event was steady. I reconnected with a few WWT colleagues I hadn't seen in a while and met others for the first time. The technical sessions were useful for context, especially around the approaches that are likely to shape our work over the next few years. And given the weather, the umbrellas ended up being a practical giveaway rather than just swag, which helped break the ice in a low-key way.

Takeaways from Ina Johnson

Ina Johnson is a TSA and data scientist with WWT's AI practice

From the viewpoint of a data scientist, SC25 was impressive. The progress made highlighted a stronger link between HPC and data science workflows. For individuals like me, this promises more opportunities to create AI solutions that utilize HPC-scale computing power and data infrastructure.

At the booth, I had valuable conversations with HPC engineers, researchers from academia and various partners. I especially enjoyed my exchanges with the academic professionals, where we discussed topics like WWT, our activities and what it's like to be part of this company.