For more than a decade, the data center management software landscape remained static. There were a few, clearly defined product categories and use cases, with relatively well-understood functionality, delivered by established vendors. This started to change in 2023, as AI began to make its way into enterprise software.

First came the expansion in the number of products that use AI for dynamic cooling optimization (see Data center management software is evolving — at last). But the conversation has progressed quickly from AI-based cooling controls that often rely on older machine learning technologies.

In the past 12 months, at least five established vendors have briefed Uptime Institute on their intention to add large language models (LLMs) to their management tools to aid in analytics and enable operators to address facility data by making queries in natural language (see AI in facility operations: three applications to watch). A few DCIM developers are already testing this functionality with early customers. Others are working on the first data center applications of agentic AI, where LLMs are equipped with controls to make fully autonomous infrastructure optimization decisions.

Uptime Intelligence surveys have revealed that software vendors, equipment manufacturers and consultants serving the data center sector anticipate quick and near-universal adoption of AI-based infrastructure management tools. However, many of their customers remain doubtful of the potential and wary of the risks.

Fear of missing out

Some vendors are adding new AI-based modules to their existing software suites, a few have launched new products, and at least one company is shutting down its original offering to focus on addressing more use cases for AI in facility operations.

According to the Uptime Institute AI Infrastructure Survey 2026, more than a third (37%) of vendors, engineering firms and consultants now offer AI-based features in their products and services. Another 25% have AI-based features in development.

Survey results confirm that vendors and consultants expect almost every data center to adopt at least some form of AI-based functionality in 2027; 46% of respondents expect AI to be taking part in most operational decisions (see Figure 1). Only a quarter say that their clients will be cautious in terms of AI usage.

Figure 1 Vendors expect broad and rapid AI adoption by clients

Looking ahead, which of the following best describes how you expect your clients will be utilizing AI in the data center a year from now? (n=341)

These opinions appear to clash with the opinions of data center operators, that are more reserved in their attitudes towards AI. Results from the 2025 Uptime Institute data center survey suggest that although the majority of respondents would agree to use AI for analytics, predictive maintenance and document generation, only 35% would allow AI-based systems to control their mechanical and electrical equipment (see Uptime Institute Global Data Center Survey 2025). In addition, only 21% would allow these systems to make staffing decisions, and just 14% would let AI make equipment configuration changes. A total of 6% of respondents would not allow any AI in their facilities (see Lack of trust will hinder adoption of AI-based controls).

The nature of the average data center customer has not changed; technologies that are being embraced in corporate back offices, such as LLMs, are unlikely to see rapid adoption in facility operations. Operators' focus remains firmly on the issues of resiliency and availability, while improving efficiency often takes a secondary role. However, most emerging use cases for AI in the data center focus on efficiency, and their impact on resiliency and availability remains to be tested.

Data center software vendors have two paths open to them:

1. They can prioritize the development of new AI-based capabilities to attract a new breed of customer. There are between 100 and 200 neoclouds operating today, constituting a meaningful customer category. These companies are young, prepared to move fast and are, in many cases, expected to use AI since their core business is supporting enterprise AI projects. For neoclouds that are selling an undifferentiated commodity (access to GPUs and other AI accelerators), efficiency presents a competitive edge — allowing them to either boost their margins or lower prices.
2. Alternatively, vendors can move at the pace set by their more established customers, who are sceptical about new AI-based features. This course would limit AI-based automation to specific systems, use LLMs for analytics, rather than control, and minimize the use of cloud computing. Historically, operators have preferred their management software to remain on-premises.

The bifurcation of the data center management software and services companies presents operators with a choice: align with the first, more adventurous group, or the second, more conservative one. This self-identification will help inform their choice of vendor and consultant partners.