Midwest Research Computing and Data Consortium
Written by: Esen Gokpinar Shelton
Last month, our team at the Pervasive Technology Institute organized the Next Gen Data Center Workshop at the Indiana University Indianapolis Campus Center. Led by the Next Gen Data Center Affinity Group within the Midwest Research Computing and Data Consortium, this event was made possible by the incredible leadership of planning committee members Todd Shechter, Winona Snapp-Childs, Dave Hancock, Brock Palen, and Graham Allan—big shoutout to them! Over 1.5 days, top minds in research computing, data center architecture, and infrastructure management from academia, industry, and vendors exchanged innovative ideas and collaborated on the future of the field.
Setting the Stage: Kicking Off with Big Ideas
The event kicked off with powerful opening remarks from Winona Snapp-Childs, PI of the Midwest Research Computing and Data Consortium, and Rob Lowden, Vice President for Information Technology and CIO at Indiana University. Their message was clear: collaboration and forward-thinking strategies are key to navigating the rapidly evolving world of data center operations.
We dove straight into the heart of the event with Dell Technologies leading the way. Steve Nowak, Data Center Director at Dell, delivered an engaging talk on integrated rack scalable systems, challenging our preconceived notions about data centers. It wasn’t just about rows of servers, but it was a deep dive into AI-driven performance boosts and innovative liquid cooling solutions. And that jaw-dropping stat about global AI infrastructure spending reaching $422.55 billion by 2029? It certainly had the room buzzing.
Big Names, Big Ideas: A Day Full of Innovation
As the day progressed, we were treated to presentations from some of the biggest names in the industry. Michael Simon from Intel shared the latest advancements in data center solutions tailored for Generative AI and High-Performance Computing. Keith Kirkendall from AMD tackled the challenges of power, cooling, and scalability in data center infrastructure. And then came Marc West from NVIDIA, who gave us a glimpse into the future of AI workloads and GPU-accelerated computing. It felt like we were peering into tomorrow’s technology, and the future looked incredibly promising.
After such a packed day of tech talk, the evening offered a well-deserved change of pace. The networking dinner at Rathskeller gave us the perfect opportunity to unwind and chat, where conversations shifted from high-density rack optimization to favorite sci-fi movies. Let’s be honest, discussing cutting-edge tech over German chocolate cake made it all the more memorable.
Diving Deeper: Day Two Highlights
Day two began with a deep dive into cooling and power trends in the IT industry, led by Andrew Pack from Dell Technologies. His session on direct-to-chip liquid cooling had everyone taking notes. It’s clear this is a technology that’s about to change the game.
Then, we split into two breakout sessions:
After a well-deserved lunch break, we reconvened for three panel discussions and a solo session that explored some of the most pressing issues facing the industry today:
Facility Operations & Capacity Planning: Dave Hancock from IU with experts from the University of Illinois and HEAPY Engineering shared valuable lessons learned from data center migrations and large-scale infrastructure projects. Panelists highlighted challenges when retrofitting spaces, often due to insufficient prior knowledge of facility transitions. They emphasized the importance of long-term planning to prevent inefficient ad hoc changes, with a focus on system lifecycle management to avoid outdated equipment taking up valuable space. Key design considerations for new builds included choosing between raised floors versus slab-on-grade construction, efficient water cooling, and well-planned loading docks to prevent delays. Scalability and adaptability were seen as vital for future-proofing, with the suggestion to allow for flexible, modular upgrades. Coordination between project managers and operators was also stressed to avoid downtime risks during maintenance. A recapitalization strategy for aging infrastructure was highlighted as critical, with financial plans for hardware upgrades being essential for long-term sustainability.
Sustainability and Considerations for Power Dense: Brock Palen from the University of Michigan led a fascinating panel on the future of power-dense data centers and AI-driven sustainability strategies. Concerns were raised about the shift to DC power systems, which offer energy efficiency but introduce safety risks. The panel explained that wider racks are becoming standard to improve airflow and manage high-density setups, while arc flash safety needs more attention. On sustainability, large data centers are exploring nuclear microreactors and microgrids, while smaller ones are considering solar power and HVO as cleaner alternatives to diesel. They even touched on Scandinavian countries that are repurposing waste heat for nearby buildings. However, space and bureaucratic challenges prevent universities from hosting data centers on campuses, despite the potential for district heating system integration.
RFI Projects & Vendor Relationships: The panel, featuring UW-Madison experts, Tadd Smejkal and Nathan Royko Maurer, discussed their 15-year Data Center Initiative, aimed at supporting research computing and consolidating data centers. Key challenges include increasing computational demands, limited capacity for high-density racks, and the need for liquid cooling. Factors for selecting a provider include location, power costs, and regulatory compliance. With limited local options, UW-Madison is weighing trade-offs between proximity and connectivity. The panel also touched on staffing issues, the cost of physical machines, and securing fiber routes. The emphasis was on balancing efficiency, cost, and scalability to avoid potential roadblocks.
The final solo session was from Thomas White at the University of Washington, where he discussed how their data center, built for lower power demands, is struggling to meet the growing AI infrastructure needs. Limited space, power, and cooling, especially for AI workloads, are key challenges. The university is planning a phased retrofitting with a $10 million investment in hyper-compute clusters to support AI. The challenge lies in fitting new infrastructure into existing space while managing union labor and resource constraints. Efforts include exploring liquid cooling, optimizing underutilized resources, and considering off-site expansion to meet rising AI demands. Each discussion was filled with expert advice, engaging questions, and a few spirited debates. Because what’s a good panel without some healthy disagreement?
Final Thoughts: A Truly Meaningful Experience
Reflecting on the Next Gen Data Center Workshop, it wasn’t just about the presentations, but it was about sharing knowledge, learning from each other, and looking ahead to the future of research computing and data center technologies. A huge thank you to everyone who attended, presented, and contributed to making this workshop a success. We can’t wait to see what the next workshop has in store!
Subscribe to get the latest posts sent to your email.