AI-Powered Data Centers: Balancing High Performance with Sustainability

Artificial Intelligence (AI) is transforming industries at lightning speed, but this innovation comes with a challenge: power and sustainability. As AI workloads grow more complex, data centers are facing unprecedented demands for energy and cooling. The infrastructure decisions you make today will determine whether your organization can scale AI effectively while meeting environmental and cost-efficiency goals.

We sat down with Raul Pombo, Principal Product Strategy Manager for Intelligent Infrastructure at Panduit, to answer the most pressing questions about AI’s high-power requirements, sustainability challenges, and the solutions shaping the future of data centers.

Question: Why is AI driving such high-power demand in data centers?

Raul: AI workloads, especially Generative AI and Large Language Models, require enormous computing resources. For example, the latest NVIDIA’s GB200 and GB300 NVL72 architectures are pushing rack power over 120kW, with individual GPUs drawing up to 1,400W. Rack power is expected to increase 2–6x by the end of the decade, reshaping data center design.

Question: What trends are shaping AI infrastructure today?

Raul: We’re seeing a shift toward edge and enterprise AI, not just cloud. This means power density is rising everywhere. Expect architectures like DGX SuperPOD to become standard, with racks exceeding 100 kW.

Question: What solutions are emerging for clean power?

Raul: Electrification is key:

• Nuclear & SMRs for zero-carbon baseload
• Wind & Solar nearby or collocated with data centers
• Intelligent microgrids integrating multiple sources for resilience

Raul: Panduit’s High-Density Intelligent PDUs and Universal PDUs deliver 8–57.4 kW per rack with advanced metering, thermal monitoring, and certified network security. Combined with high-density UPS systems, these enable scalability and energy efficiency.

Question: What role does cooling play?

Raul: Cooling can be as high as 40% of data center energy usage. Utilizing Direct-to-chip liquid cooling is not only a must-have for achieving the cooling capacity needed for mitigating the heat generated by AI workloads in the rack, but it also makes the data center 18-27% more energy efficient.

Question: How does connectivity impact sustainability?

Raul: High-speed fiber architecture not only increases bandwidth & reduces latency, which would be required for AI applications, but could also optimize for energy consumption. For example, moving from 100G to 400G, the number of top-of-rack switches & transceivers can substantially be reduced, reducing not only cost but also energy consumption, which impacts sustainability.

Question: Thank you, Raul, for your time. Any closing thoughts?

Raul: AI is transforming industries, but its power appetite is massive. The key is innovative solutions that scale performance without sacrificing sustainability, and at the same time optimizing for cost.

Looking to learn more about solving AI’s high-power and sustainability challenges?

Watch our webinar featuring “Artificial Intelligence, High Power & Sustainability Trends, Challenges & Solutions hosted by Raul, or connect with our team of AI thought leaders. We’ll be happy to collaborate and help you develop a strategy that balances performance, efficiency, and sustainability.