Retrofitting Data Centers for AI: How Adaptive Reuse and New Design Tackle Power and Space Challenges

JLL: Retrofitting is Essential for AI Data Center Growth

The rise of AI is changing data center design in fundamental ways. As GPU costs soar to $30 million per megawatt, operators are shifting focus to smaller, more power-dense facilities rather than sprawling campuses.

AI Drives New Data Center Design Needs

Traditional hyperscale data centers, built to support massive compute demands, are different from AI-only operations. The high cost of AI hardware and its unique physical characteristics mean that building million-square-foot AI facilities is often not financially viable except for the largest players.

AI servers resemble heavy, jet-engine-like machines rather than the manageable traditional servers. This requires rethinking building structures to support heavier floor loads and redesigning cooling systems to handle increased heat output.

Cooling and Structural Changes Are Key

AI data centers are moving towards liquid cooling to manage heat more efficiently. However, air cooling remains necessary for network equipment and spaces occupied by personnel. This hybrid cooling approach complicates HVAC design and calls for innovative facility solutions.

Retrofits Provide Near-Term Solutions

With colocation vacancy rates dropping to just 2.6% and rents rising over 11% nationwide, finding available space is a growing challenge. Adaptive reuse—converting industrial or commercial buildings into AI-capable data centers—is gaining traction as a faster and more cost-effective alternative to new builds, especially in urban areas.

Many existing data centers that migrated workloads to the cloud now have unused power capacity. These spaces can support smaller AI workloads of 1–3 megawatts, ideal for development and testing labs.

Challenges and Opportunities in AI Data Center Development

• Power and land shortages are significant hurdles for new construction.
• Smaller, power-dense buildings help reduce capital expenses linked to expensive AI hardware.
• Liquid cooling integration allows retrofits to extend the life and capability of existing facilities.

Data center providers are adapting quickly, evolving design and operational strategies to meet the unique demands of AI infrastructure. This flexibility is critical for supporting the growing AI ecosystem within real estate and construction sectors.

For professionals interested in staying informed on AI infrastructure trends and technical skills, exploring specialized AI courses can be valuable. Resources are available at Complete AI Training.