According to recent data from the independent public registry Energy Incident Register, the global data center infrastructure is facing an unprecedented surge in technological failures. Analysis shows that starting from 2024, the number of serious fires and emergency power outages in data centers worldwide has increased more than twofold compared to previous periods.
Scale of the Problem and Regional Statistics
While official statistics recorded eight major incidents in 2024, this figure sharply jumped to 20 and 19 cases in 2025 and 2026, respectively. The monitoring covers open data collected from 2010 to 2026 across 26 countries. Out of 93 documented incidents during this period, nearly half (47 episodes) occurred in 2024 or later. The absolute leader in reported accidents is the United States with 33 incidents, followed by the United Kingdom with 14 similar occurrences.
Main Causes and Vulnerable Infrastructure Components
A detailed analysis of the 93 incidents demonstrates that 52 of them are directly related to fires and so-called thermal events, with 18 classified as major accidents. Uninterruptible power supply (UPS) systems and battery rooms remain the primary risk zones, accounting for 22 incidents. Backup diesel generators rank second, causing 15 events. Notably, a high degree of uncertainty remains: in 53 cases, open sources did not allow exact identification of the specific equipment component that triggered the failure.
Contradictory Data
A deeper examination of the registry revealed certain discrepancies in versions and reporting volumes. Since the initial publication of the report, the total number of registered incidents in the database has risen to 174, though the period from 2024 onwards now accounts for 50 incidents, differing slightly from the initial sample. Furthermore, the updated version includes data on two accidents at Russian data centers belonging to Yandex (in Sasovo and Vladimir), while similar incidents at facilities in Kaluga and Sber's infrastructure were not yet integrated into the official registry at the time of reporting.
Expert Assessments and Industry Implications
Data from independent registries contrasts with recent reports from industry analysts such as Uptime Institute, which reported in the spring that overall data center fault tolerance had increased over the past five years. Nevertheless, experts agree that even with improved baseline reliability, mitigating the consequences of accidents is becoming increasingly expensive for operators, and the time required to restore system functionality is steadily rising. Human error and vulnerabilities in guaranteed power supply systems remain the main drivers of ongoing failures.
