Getty Images/iStockphoto
The road to automation: Exploring lights-out data centers
The lights-out data center has been a hot idea for years, but sites still don't have them. How far can we realistically go with data center automation?
The concept of a "lights-out" data center, which is fully automated and operates without human assistance, lacks consensus among data center managers.
To the max, a lights-out data center is 100% automated in every way. It operates 24/7 year-round without human assistance. The work is done with full automation; the lights are turned off to save energy; the staff is not on-site; and the site saves money. However, some managers view lights-out as running unattended during certain periods, such as weekends or overnight, while staff are still present during peak times. Some lights-out data centers rely solely on IT for remote monitoring and occasional equipment repairs.
This article examines the potential and challenges of lights-out data centers, highlighting the trend toward automation while acknowledging obstacles such as skill shortages, strategic uncertainties and operational risks.
Reality of a lights-out data center
Shortages of data center skills and the increasing use of automation and rule-based decision-making are driving data centers toward operations that are approaching lights-out capabilities.
Microsoft demonstrated the feasibility of lights-out data center automation several years ago with an underwater data center on the seabed off the coast of Scotland. The data center housed 12 racks of servers that operated independently and without human intervention for 25 months, yielding strong results. However, Microsoft abandoned this effort in 2025 because the data center had a fixed capacity, making upgrades and servicing difficult. Equipment also had to be brought back to the surface for repairs.
Despite the reality check, this still doesn't mean that full data center automation can't work in less challenging circumstances. Many data center configurations and hardware, software and network tweaks can now be handled virtually, along with security. Virtualization reduces reliance on physical servers and storage, and zero-trust networks can automatically detect, alert and intervene when unknown activity or network assets are discovered at edge locations.
What's stopping full data center automation?
From the CIO's vantage point, several strategic issues remain unresolved for lights-out data centers, including inexperience, risk, rearchitecting, facility management and budget constraints.
Inexperience
Even the most seasoned data center personnel have limited experience with lights-out data centers. The inevitable operational anomalies will come up, and staff will need to learn as they go.
Risk
Executive management, boards of directors and the insurance companies that provide protective policies for data centers are unlikely to be comfortable with lights-out data centers. What if something goes wrong? What happens if the data center automation triggers a shutdown with no human present to confirm the decision?
Rearchitecting
Full data center automation means that everything in the data center must be autonomously run. Today, data centers are distributed across central IT, remote edge locations, field offices, employee home offices and the cloud. There are no longer monolithic data center glasshouses. Consequently, lights-out automation must follow these same hybrid data center trajectories if it is to carry out everything that a data center is charged to do. This most likely means there needs to be a whole infrastructure overhaul.
Facilities management
Data centers aren't just about IT. They also encompass facility assets such as HVAC systems, microgrids and physical security. Facilities skillsets are in short supply, and many of these assets still require human intervention and support.
Budget constraints
AI and data center automation aren't cheap, despite arguments that reduced data center headcounts can recoup these investments. Cost-justification arguments also fall short because most enterprises and their liability insurers still want real humans working in the data center.
Lights-out technologies and more AI in data center operations will continue to expand, if for no other reason than that data center vendors are embedding these capabilities in their products. However, enterprises are still not on board with autonomous data centers. While AI-driven automation is advancing, human intervention will remain necessary in many operational areas for the foreseeable future.
Mary E. Shacklett is president of Transworld Data, a technology analytics, market research and consulting firm.
