From incident to insight: Fire risk in modern data centers

Modern data centers are becoming increasingly vital infrastructure, yet several recent high-profile fire incidents have exposed persistent vulnerabilities. As artificial intelligence (AI) technologies continue to advance, these risks will only intensify. Contributing causes of such fires include electrical faults, battery failures, cooling system malfunctions, and human error. This perspective paper synthesizes key information from recently reported incidents and discusses practical fire safety strategies for both prevention (i.e., AI-driven fault detection and fire-safe battery storage) and suppression (i.e., clean agents and liquid nitrogen system). Emerging technologies are highlighted as potential fire safety enhancements, and their development and implementation in modern data centers are recommended. Two relevant methods for fire risk assessment are explored, specifically non-scenario-based consideration of common fire causes and scenario-based examination of recent incidents. These assessment methods should be utilized while considering engineering design practices, operational feasibility, and regulatory alignment to enhance resilience and promote adoption in modern data centers. This work intends to offer a perspective on data center fire risk assessment by examining past incidents, presenting insights into current knowledge gaps, and proposing future research and stakeholder efforts for the improvement of data center fire safety. • Fire incident risk across modern data centers is increasing rapidly. • Contributing causes of these fires are analyzed based on representative case studies. • Practical fire safety strategies for both prevention and suppression are discussed. • Emerging technologies are highlighted as potential enhancements to fire safety.