Just as mass production created industrial winners in the last century, gigawatt-scale infrastructure is driving true innovation in the AI-era.
When Henry Ford perfected the moving assembly line in the 1910s, he didn't just revolutionize car manufacturing. He created an entirely new industrial category. Gigawatt-scale AI factories represent a similar inflection point. Not just an evolution of the data center, but the emergence of a new infrastructure class. As NVIDIA founder and CEO Jensen Huang declared at GTC Washington recently: "The next industrial revolution is here, and with it a new kind of factory."
NVIDIA recently unveiled its Omniverse DSX Blueprint and AI Factory Research Center in Virginia, establishing the foundational architecture for multi-gigawatt AI infrastructure. This effort includes working closely with infrastructure partners like Vertiv to translate this blueprint into deployable systems, "We are truly witnessing the birth of a new era," said Giordano Albertazzi, Vertiv's CEO. "We are in the age of the AI factory."
At NVIDIA GTC in Washington, D.C., Vertiv leaders outlined how the company is enabling the next generation of AI factories. CEO Gio Albertazzi, EVP Scott Armul, and Americas President Anand Sanghi explain how the new gigawatt-scale reference architectures for the NVIDIA Omniverse DSX Blueprint support faster, repeatable deployment of fully integrated power and cooling systems.
Infrastructure determines outcomes
Every major technology shift has been constrained or accelerated by specific infrastructure. The internet needed fiber optics. Cloud computing needed hyperscale data centers. AI at true scale needs something different: facilities where power, cooling, and compute are designed as one system, not assembled from separate parts.
The Stargate Project exemplifies this shift. Announced in January 2025 by OpenAI, SoftBank, and Oracle, the plan includes $500 billion over four years in AI infrastructure, starting with $100 billion immediately. The flagship campus in Abilene, Texas is operational, with Oracle delivering NVIDIA GB200 racks in June. Buildings, power systems, and cooling are now co-designed with compute, not added later.
Traditional data centers can take months to years to build. AI has the potential to advance faster. Stargate alone plans over 7 gigawatts of capacity running 2 million chips across multiple sites. Achieving this goal demands faster deployment with built-in flexibility.
That flexibility matters because chip generations evolve rapidly. NVIDIA's Blackwell platform, announced in March 2024, delivers up to 30x better performance than H100 GPUs for large language model inference while cutting costs and energy use by up to 25x. Just 18 months later, in September 2025, NVIDIA and OpenAI announced plans to deploy the next-generation Vera Rubin platform starting in the second half of 2026. Facilities designed today must support multiple hardware generations while protecting billion-dollar investments that need to perform for decades
Intelligent infrastructure
The NVIDIA DSX Blueprint changes how AI factories operate: the facility itself becomes intelligent. Digital twin technology allows the entire infrastructure to be simulated, optimized, and monitored in real-time before and after construction.
The workflow follows a clear progression: engineers design and optimize layouts in the digital twin, simulate thermals and electricals with precision, then deploy prefabricated modules that arrive factory-built and tested. This approach shrinks build time significantly, achieving faster time to revenue.
Once operational, the digital twin acts as an operating system. AI agents trained in the virtual environment optimize power consumption and reduce strain on both the facility and the grid in real-time. Infrastructure is no longer a fixed asset. It's a responsive system that adapts continuously.
Industry transformation at scale
Gigawatt AI infrastructure will reshape sectors. Healthcare gains diagnostic capabilities currently constrained by compute. Financial institutions deploy models that process market dynamics at new speeds and scales. Manufacturing integrates AI into production systems that optimize across global supply chains. Federal agencies build intelligence capabilities that enhance security.
These transformations depend on infrastructure that supports AI at production scale, not lab scale. The ecosystem enabling this shift (i.e., digital twin platforms, prefabricated modules, grid integration technologies) represents one of the largest coordinated infrastructure buildouts in modern history.
A new industrial era
Ford's 1913 assembly line collapsed production time from 12 hours to 93 minutes, not by working faster, but by redesigning the factory itself. Moving conveyor belts replaced stationary workstations. Electricity powered coordinated workflows. The entire system moved as one.
AI at scale demands the same rethinking. Facilities where compute, cooling, and power operate as one integrated system and not separate parts retrofitted together. The organizations building gigawatt infrastructure today aren't just scaling up. They're collapsing deployment timelines and reimagining production for a new era.
