Artificial intelligence is often described as weightless. Software. Models. Clouds.
But AI has a physical constraint.
Every new model, every training cycle, every query answered draws power. Global data center power demands are projected to nearly triple between 2024 and 2030, with AI workloads accounting for about 70 percent of this growth.
And not in smooth, predictable patterns. AI workloads can spike, surge and stress infrastructure in ways traditional data centers didn’t.
Running an AI data center isn’t just a question of powering today’s demand.
It’s about planning for growth that doesn’t slow down.
Across the industry, this is reshaping how facilities are designed, built, and operated — and how technology providers like ABB are innovating alongside the ecosystem.
Speed
AI demand isn’t arriving gradually. It’s arriving all at once.
Across markets, the same questions keep surfacing:
How do you plan for demand when forecasts are obsolete within a year?
How do you bring new capacity online fast enough to stay competitive?
In the U.S., Applied Digital is tackling scale with speed —developing AI-ready facilities that can be delivered faster, expanded modularly, and operated reliably under extreme loads. ABB supports this approach with standardized, pre-engineered electrical solutions that reduce deployment time while maintaining reliability and introducing an innovation element: a medium-voltage power architecture, including medium-voltage UPS.
Scale
AI is pushing data centers beyond familiar limits.
Megawatts quickly become gigawatts. Facilities evolve into campuses.
And expansion is no longer a future phase—it’s a constant condition.
Instead of isolated components, operators are moving toward end-to-end architectures—from grid connection to rack—designed to scale in stages, adapt to changing loads, and be deployed globally. It’s a shift already visible across the industry: today, one in four data centers worldwide operates on ABB technology, spanning power, digitalization, service, and cooling.
In the Philippines, the country’s first AI-ready hyperscale data center reflects a broader trend: emerging markets leapfrogging legacy designs to meet hyperscale demand from day one. The focus is on scalable, energy-efficient infrastructure that can expand rapidly without sacrificing operational control.
Performance
AI is driving a massive surge in data center power density.
As density increases, performance also becomes a matter of space.
In the UK, Ark Data Centres needed to support AI-driven density growth within tight space and sustainability boundaries. The response was a first-of-its-kind, medium-voltage power architecture, enabling higher capacity per site while reducing footprint and improving efficiency—an approach pioneered and led by ABB.
Designing for the gigawatt era
We're entering an inflection point where Alternate Current (AC) and Direct Current (DC) work together to unlock new possibilities.
AI workloads are demanding fundamentally different power delivery models, and the industry needs innovation that bridges proven AC infrastructure with emerging DC technologies.
ABB’s collaboration with NVIDIA reflects a broader industry shift: aligning power infrastructure and AI compute earlier in the design process. As AI factories emerge, DC opens doors to higher efficiency for AI workloads where electrical architecture, energy management, and digital control are no longer downstream considerations—they are part of the core system.
From megawatt to gigawatt-scale capabilities, ABB’s role is increasingly about orchestration: connecting power, automation, digital intelligence, and service into systems that can scale globally, while being engineered locally.
AI may be written in code.
But it is built—physically—through infrastructure that can grow as fast as ideas.
And that is now the defining challenge of the data center industry.
