The AI Boom’s Cooling Problem: How Heat Threatens Tech Progress

The increase in data centers, due to companies outsourcing data storage and using energy-intensive artificial intelligence, is creating challenges in maintaining cool environments.

The increase in data centers, due to companies outsourcing data storage and using energy-intensive artificial intelligence, is creating challenges in maintaining cool environments. A recent outage at CME Group, caused by a cooling issue at a CyrusOne data center near Chicago, highlighted these problems. CyrusOne operates over 55 centers globally and is actively working to fix the cooling system.

Data centers generate significant heat due to high-powered AI and cloud servers that require constant power and cooling. Traditional air cooling systems often struggle to manage this heat, which can lead to server malfunctions. Industry experts explain that server chips must maintain specific temperatures to function properly.

To address cooling challenges, more data centers are considering liquid cooling, which is much more efficient than air cooling. However, liquid cooling presents its own issues, such as leaks and maintenance needs. In response, some companies, like Microsoft, are developing new designs that use no water for cooling by recycling water in a closed loop system.

Outages related to cooling problems are relatively rare due to stringent uptime requirements for data centers, although power-related issues are more common. The demand for data centers has led to significant deal-making in the industry to enhance cooling capabilities. It is estimated that around 40% of a data center’s energy is consumed by cooling systems. Recently, Eaton announced a $9.5 billion purchase of Boyd Corporation’s thermal business to meet rising cooling demands, and Vertiv is pursuing a $1 billion acquisition to expand its liquid cooling services.