Keynotes

When: Wednesday, April 22, 2026 | 8:30 am to 9:45 am
Location: Ballrooms A & B, 3rd Level

Artificial intelligence is rapidly transforming data centers from traditional IT environments into highly specialized AI factories built for extreme compute density, massive energy throughput, and global scale. As AI workloads accelerate, operators must rethink how infrastructure is designed, powered, cooled, and deployed.

In this keynote, engineering leaders from Oracle, NVIDIA, and Google will share firsthand insights into how hyperscale organizations are building and operating the next generation of AI infrastructure. The discussion will explore the real-world engineering and business decisions behind modern AI data center design.

From securing power in grid-constrained markets to supporting GPU-dense environments and deploying infrastructure faster than ever before, this session will provide a practical look at how leading organizations are scaling the facilities that power the AI economy.

Attendees will learn:

• How AI is redefining the data center and transforming facilities into high-performance compute factories.

• Strategies for securing and managing power as AI infrastructure pushes energy demand to new levels.

• How operators are supporting extreme rack densities and preparing for next-generation GPU environments.

• Where liquid and hybrid cooling strategies fit in the future of AI infrastructure.

• How hyperscalers are accelerating deployment timelines through modular design and supply chain innovation.

• What infrastructure leaders must do today to prepare for the next decade of AI-driven demand.

When: Tuesday, April 21, 2026 | 8:30 am to 9:45 am
Location: Ballrooms A & B, 3rd Level

AI facilities are no longer scaling like conventional data centers. As rack densities move from tens of kilowatts toward hundreds, operators are managing a new class of infrastructure challenge: power, thermal, controls, white space, and services now behave as one interdependent system. In this session, Scott Armul will present a practical operator framework for planning and scaling AI infrastructure as a physical AI engine, where the facility itself becomes part of the compute equation.  

The session will address four realities shaping AI deployments today: extreme densification, compressed time-to-capacity (time to token), rapid campus-scale expansion, and the operational risk created by disparate systems. Scott will outline how operators can reduce these risks by shifting from component-by-component decisions to a converged infrastructure approach that integrates the power train, thermal chain, and controls/services layer as a coordinated system.  

Attendees will also see how repeatable AI building blocks—such as 12.5 MW units scaled into larger campus architectures—can improve deployment predictability, reduce onsite complexity, and preserve flexibility across future compute generations. The emphasis is on operational outcomes: faster deployment, better utilization, and lower integration friction at scale.  

Attendees will leave with three takeaways:     

1. A clear framework for treating AI facilities as integrated physical systems, not isolated infrastructure domains.           

2. A practical scaling model for using repeatable building blocks to accelerate deployment while maintaining flexibility.

3. An operator-focused approach to converged power, thermal, and controls integration that improves reliability, efficiency, and time-to-capacity.

When: Wednesday, April 22, 2026 | 8:30 am to 9:45 am
Location: Ballrooms A & B, 3rd Level

As data centers and high-performance computing environments scale to unprecedented levels, traditional power distribution architectures face critical limitations. This presentation explores the next frontier in energy delivery: the 1 MW rack and the evolution beyond sidecar designs through advanced 800 VDC strategies. We will examine how VDC optimizes efficiency, reduces losses, and supports the growing demands of AI workloads and hyperscale infrastructure.

Attendees will gain insights into design considerations, safety standards, and integration approaches that enable sustainable, high-density power distribution for future-ready facilities. Join us to understand why 800VDC is not just an alternative—it’s the foundation for powering the next generation of compute.

Key Takeaways:

• High-voltage direct current (HVDC) architectures are essential for meeting the massive power demands of AI workloads and hyperscale data centers. Moving beyond traditional sidecar designs unlocks the ability to support 1 MW racks efficiently.

• HVDC reduces conversion losses, improves energy efficiency, and minimizes infrastructure footprint—critical for sustainable, future-ready facilities.

• Learn how optimized HVDC strategies lower operational costs while supporting green initiatives. Implementing HVDC at scale requires careful attention to design considerations, safety standards, and integration approaches.

These best practices ensure reliable, secure, and scalable power distribution for next-generation compute environments.

When: Thursday, April 23, 2026 | 8:45 am to 10:00 am
Location: Ballrooms A & B, 3rd Level

As AI data centers outgrow conventional commercial design, they must evolve into true industrial infrastructure. This session introduces MHI’s system level approach, applying proven industrial energy, power, cooling, and control technologies to AI scale facilities.

We illustrate the transition from grid dependent power to hybrid on site generation, from low voltage to medium voltage and HVDC distribution, and from air based to highly efficient water based cooling. We also show how plant grade monitoring and control enable global optimization across efficiency, reliability, and deployment speed.

Attendees will gain a practical framework for building decarbonized AI infrastructure that scales responsibly and resiliently.

3 Takeaways:

1.Why AI data centers must evolve from commercial facilities to industrial infrastructure, using hybrid grid and on-site generation to overcome energy, reliability, and carbon constraints.

2.How integrated industrial-grade optimization—across power generation, HVDC distribution, cooling, heat recovery and plant-level control—outperforms siloed design approaches.

3.How modular systems accelerate megaproject deployment while improving efficiency, scalability, and overall performance.