Inside the data centres powering AI: gold racks, demented noise, and 40 diesel generators
What changed
A detailed tour inside Digital Realty’s AI test data centre in Ashburn, Virginia reveals what powers AI at the ground level. It’s far from a sleek software story. Instead, there are towering racks of gold-plated server components, relentless noise from whirring fans, intense heat, and a massive demand for electricity met by 40 diesel generators. The facility design reflects the raw physical infrastructure required to run AI workloads at scale.
Why builders should care
AI innovation depends heavily on data centre density, cooling systems, and uninterrupted power supply. The presence of dozens of heavy-duty diesel generators signals the critical need for resiliency against grid failures, especially during peak loads. For AI operators, this means infrastructure costs and operational risks are higher than typical cloud compute setups. It also highlights a growing tension between AI growth and energy sustainability, which affects how and where AI services can reliably run.
The practical takeaway
Running or relying on AI infrastructure is not just about software or chips; it demands careful attention to physical building dynamics. Power costs, backup strategies, and heat management significantly impact operating margins and service reliability. High-density AI workloads may force data centres closer to expensive, fossil fuel–powered backup systems until clean energy and storage catch up. Businesses scaling AI products must factor infrastructure reality into their cost projections and risk assessments.
What to watch next
Track how data centre operators evolve their power and cooling approaches under AI demand pressure, particularly regarding clean energy adoption. Changes in regulatory and environmental pressures might restrict diesel backup usage and raise operating costs. Also monitor if AI hardware manufacturers optimize for lower energy intensity or if edge data centre solutions gain traction to reduce reliance on massive central facilities. How AI infrastructure scales will influence the pace and geography of AI deployment.
AI Quick Briefs Editorial Desk
