The DICE South 2025 conference took place in the heart of Dallas at the Fairmont Hotel on Wednesday, August 6. This year, the attendance doubled to over 600 professionals spanning the digital infrastructure ecosystem. At the same time,  the event remained intimate enough for direct conversations among industry leaders. While some participants traveled from out of state, the focus stayed on the great state of Texas and its important role in the data center ecosystem. As this event continues to grow, DICE Global Director, Adam Knobloch, noted that next year’s event will be a day and a half long, as there was so much to discuss and it will only increase in the months to come.

The State of the Southern Data Center Market.  

Jerry Blair of DataBank, and Gordon Kellerman of Skybox Datacenters, shared insights into developments across Texas. Infrastructure constraints and legislative factors emerged as key themes for the future direction. With shuffling and uncertainty from the customer side to what the design of the data center should be, we need to approach things differently.  

“We need to interface with utilities differently; it’s no longer a real estate function. The grid infrastructure is antiquated, and it is incredibly difficult to solve the future needs, ” said Nelson Abramson, Verrus.

Data Center Cooling Innovations: Preparing for the Next Wave of the AI Era

Speakers discussed the need to support high-density chip workloads and the transition from air to liquid cooling. Steve Altizer of Compu Dynamics, Jim Roche of CyrusOne, and Craig Cook of Flexential outlined how AI demands are reshaping data center design and equipment. With the next generation of NVIDIA chips, the risk is significant because if there’s a point of failure, the chips can begin to melt in seconds.

“AI is not just for the big guys – enterprises, universities, HPC labs – they’re all in,” said Jim Roche, CyrusOne.

Data Center Financing and Capital Markets: Navigating a Dynamic Landscape

This session focused on investment models and build-versus-buy decisions. Fentress Boyse, Partners Group, Vardahn Chaudhry, PowerHouse Data Centers, and Bryan Marsh, StratCap Data Centers, emphasized that it’s a gold rush, so we need cautious investment strategies in a fast-changing market.  The average cost for a 1GW data center is $1B,, and there aren’t many public stocks available on the data center build side.   

“You kiss a lot of frogs in this business because you have to have a lot of deals in the funnel,” said Fentress Boyse, Partners Group. “Diversification between markets for investors is still recommended.”

Best Practices for Maximizing Performance and Minimizing the Risk in Direct to Chip Applications

Speaker Pete Elliott of ChemTreat discussed operational risks and technical specifics for liquid cooling systems.  He shared insights on water chemistry and the three big effects of pH, biocells, and corrosions on data center liquid cooling. He emphasized a low pH balance, glycol mix, and going beyond the stringent ASHRAE water standards.

Next-Gen Data Center Operations: Managing Uptime, Automation, and Human Capital in the AI Era

Automation tools and predictive analytics are being adopted to reduce manual oversight. Andrew Jimenez, Sr. Director of Data Center Solutions at Wesco, noted that less than 50 percent of data centers are currently automated. The challenge, he explained, is not just in extracting data, but in making sense of it and figuring out how to orchestrate that information to drive greater efficiency.

“Through our partnership with Schneider Electronics, we’ve launched a new program that uses predictive analytics to reduce human touchpoints by 40%. This directly cuts downtime and operational expenditures. What’s equally important is how we’re now able to pass data anomalies upstream to inform better design decisions and create a tighter feedback loop between operations and engineering,” said Michael Murphy, VP, Commissioning & Operations Engineering, Compass Datacenters.

According to Nabeel Mahmood, Co-Founder of the Nomad Futurist Foundation, the industry is facing a serious workforce gap, with an estimated 300,000 unfilled jobs projected in the U.S. by 2030. David McCall, QTS, shared efforts to train workers and integrate AI. “You still need people for data centers and you need to season them, train them, give them AI assistants,” said David McCall, QTS. 

Modern Data Center Development: Navigating Speed, Scale, and Sustainability in the Field

Developers addressed challenges around planning, permitting, and utility constraints. Luke Peters, VP of Data Center Construction at Prologis, discussed how digital twins are accelerating development by enabling reality capture, real-time model updates, and precise tracking of placements. While technology is advancing, he noted that robotics still hasn’t reached full applicability in construction.

Tyler La Susa, VP of Business Operations at TRC Companies, emphasized that the commissioning phase is now just as critical as design. He noted that while modular manufacturing offers advantages, such as offsite construction for electrical and battery rooms, it comes with tradeoffs compared to traditional stick-built data centers. Modular designs require planning up to two years in advance, limiting agility and leaving little room for last-minute equipment changes. For AI workloads, cabinet constraints in the white space can be a challenge, but modular remains a strong option for supporting infrastructure.

Data Center Security From the Outside In

Altaf Bora, Metro One LPSG, discussed new layers of protection, and John Riotte, Global Physical Security for Data Centers at Meta, highlighted key security considerations when integrating SMR nuclear on-site. Tighter physical and cybersecurity measures are required, including armed guards, background checks to exclude felons, enhanced training, and thorough vehicle searches. He recommended placing the SMR adjacent to the data center to reduce the risk of protocol violations and potential fines.

Optimizing Supply Chain, Quality, and Capacity Planning Amid Unprecedented Growth

Panelists discussed material shortages, pricing swings, and sustainability pressures. Marc Bhuyan, Google, pointed to predictive analytics and scoring models for suppliers.

“We’ve moved from reactive to predictive. That’s how we stay ahead,” said Marc Bhuyan, Google.

Powering the Future: Onsite Energy Strategies, Utilization Efficiency and Sustainable Design

This session looked at sustainability, alternative fuels, mobile turbines, and DC infrastructure. Chris Handwerk, Chief Development Officer at Amp Americas, explained that renewable natural gas—sourced from decaying dairy manure—is a cleaner alternative to diesel. It addresses both sustainability regulations and power reliability. He added that although natural gas makes up 43 percent of total electric consumption in the U.S., only half of it is currently used for power generation. In Texas, natural gas is abundant across most regions, except in the central part of the state.

Knoell Coombs, VP of Life Cycle Power, discussed onsite power generation using turbine-based natural gas systems. He noted that for any significant load, the timeline is over two years out. However, mobile turbine units, such as a 16.5W generator within a 100-foot by 10-foot footprint, can be deployed to roll in 100MW of power and be commissioned within days. While large-scale turbines typically rely on diesel for backup, natural gas turbines can serve as a continuous primary power source. With hyperscale workloads fluctuating by as much as 9MW within seconds, he stressed the importance of pairing turbines with battery backup to ensure stability.

JP Buzzell, VP and Chief Architect at Eaton, emphasized the need to rethink current AC infrastructure and shift toward DC systems to improve efficiency and reduce energy loss. He pointed out that by transitioning to direct current, operators can unlock cost savings and optimize power delivery, especially for high-density applications.

DICE South in Summary

Although power remains a central topic in every session, speakers spent significant time discussing water conservation and the cooling requirements that come with AI infrastructure. Most agreed that the scale of liquid cooling is accelerating fast, with deployments shifting from less than 20 percent adoption to more than 80 percent due to the demands of Nvidia chip power densities.

Another consistent theme was workforce development. AI is seen as a tool to empower workers rather than replace them. Sustainability was also present in every discussion. Whether talking about cooling fluids, energy strategies, or build materials, every panelist acknowledged the importance of meeting both company and customer sustainability goals.

There was a strong sense of collaboration at DICE South 2025. Industry leaders showed a real commitment to working together, not only to solve technical and operational challenges but to build a stronger, more sustainable data center future.  We look forward to continuing the conversations at DataCloud USA in Austin, Texas September 16-17th.