Building AI infrastructure for what comes next
AI is pushing data center infrastructure beyond the assumptions, architectures, and delivery models that defined the cloud era. Meeting the demands of AI at scale requires a more coordinated approach to infrastructure design, construction, and deployment. We’re applying product thinking to the data center to make infrastructure more scalable, efficient, resilient, and predictable.
The future of AI won’t be defined by compute alone. It will be defined by our ability to deliver power, capacity, and infrastructure with greater certainty.
— Revathi Advaithi, CEO
Featured speakers
From Dirt to Tokens: Building AI Infrastructure as an Integrated System
Chris Butler | President of Embedded and Critical Power
Rob Campbell | President of Communications, Enterprise and Cloud
What the Automotive Industry Can Teach Us About 800 VDC Data Centers
Adam Morton | CTO, Data Centers
High-Speed Open Silicon Photonics Reference Designs
Janos Veres | Sr. Director, Assembly & Test Technologies
Scaling Digital Infrastructure in Hardware Manufacturing
Brandon Weilbacher | Sr. Director, Technology Strategy
The New Focus on Energy Storage
Shrikanth Venkateshappa | Director of Accounts
The Compliance Gap: ESS Safety in the Age of AI Infrastructure
Shrikanth Venkateshappa | Director of Accounts
Related resources
ARCHITECTURE
System-level modular infrastructure reference design for NVIDIA Vera Rubin NVL72
Design a scalable modular data hall with a factory-integrated approach to power, hybrid cooling, controls, testing, commissioning, and lifecycle support from Axiom. This reference design establishes a standard electrical and modular infrastructure baseline for the NVIDIA Vera Rubin NVL72 platform, with detailed architectural schematics, design parameters, and technical specifications.
ARCHITECTURE
Rethinking 5 assumptions about data center power
AI is forcing a fundamental shift in grid-to-chip power distribution. Optimizing individual technologies is no longer enough; the infrastructure must perform as an integrated system. As rack densities race beyond 1 MW, long-held assumptions about voltage, cooling, construction, and supply chains are colliding with physical limitations. Here are five that deserve a closer look.
EXECUTION
800 VDC data center power: Accelerate adoption with lessons from EV automakers
EV manufacturers spent more than a decade refining the technologies, safety mechanisms, and operational practices required to manage 800 VDC systems. Explore which lessons transfer to data centers and where the demands of continuous operation call for a different approach in this white paper by Flex and UL Solutions.
EXECUTION
Scaling AI starts with repeatable infrastructure
The technologies powering AI infrastructure don’t evolve independently. As dependencies multiply, the engineering challenge shifts from optimizing individual systems to managing the interactions between them. Replacing project-by-project engineering with a common platform simplifies manufacturing, reduces commissioning effort, and provides a stable foundation for repeatable deployment across multiple sites.
OUTLOOK
What’s really constraining AI infrastructure — and what to do about it
AI infrastructure is exposing constraints beyond GPU production. Energy availability, cooling resilience, inter-site connectivity, and permitting are defining where, when, and how fast new capacity comes online. As AI scales, overcoming these constraints will require infrastructure designed for greater integration, resilience and deployment certainty.
OUTLOOK
Cooling custom silicon at scale
Custom cooling cannot be designed around the processor alone. It must support the thermal needs of the chip, connect cleanly into the tray, rack, CDU, and facility water system, and be ready for the next generation without forcing an infrastructure redesign. Find out how Flex is balancing silicon-level optimization with infrastructure-level standardization at scale.
OUTLOOK
Built different: The rise of data center productization
AI infrastructure won’t scale through better construction. It will scale through products designed for industrialization. Moving from a traditional construction model to a productized model shifts complexity from the field to the factory for faster, repeatable deployment that scales at the speed of demand.