The expansion of data center infrastructure over the last decade has been driven by an increasing demand for distributed services and coordinated multi-site deployments aimed at eliminating downtime and providing failover contingency. However, with the rise of AI, data centers are set for a transformative shift in both the technology they deploy and how they are structured. This transformation is poised to be both vast and significant, affecting everything from the design of data centers to their location, which will be heavily influenced by environmental factors.

So, what does the future hold for data centers, and how will they differ from the hyper-scale model we are familiar with today?

1. Massive Growth Ahead

Even without AI, data center growth and power consumption were already expected to surge in 2024. In 2022, data centers consumed 460 terawatt-hours (TWh) of electricity, accounting for 2% of global electricity usage. By 2023, this grew by 55%, reaching 7.4 GW. With the rise of hyperscale, AI, and crypto data centers, power demand is expected to increase by 160% by 2030, potentially making up nearly 4% of the world’s total electricity consumption.

While AI isn’t the only factor driving this growth, it is certainly accelerating the scale and number of data centers needed. AI data centers differ from conventional hyper-scale environments primarily in their power demands. While storage needs are similar, the compute power required for AI is significantly higher. As a result, AI data centers will need careful planning around power access, heat management, and local environmental impact.

In the U.S., which has ample space to build data centers away from urban centers, the AI revolution may be more manageable. However, without a unified power strategy like that found in Europe, reliable power supply remains a challenge. As a result, issues surrounding location, power, and thermal management could slow the speed of this transformation. The key to success will lie in sites with access to renewable energy sources such as solar, wind, and hydropower, as well as the right infrastructure for mitigating heat.

2. The Emergence of New Data Center Types

Traditional cloud data centers often mixed storage and localized processing to serve data to client-facing platforms. These facilities were generally homogeneous in terms of services offered and scalability, often adding unused racks or larger drives to increase capacity.

However, AI data centers are set to be much more specialized and complex. Depending on the model they host—whether generative or predictive networks—optimization for specific workloads will be key. These data centers will likely have two distinct functions: development environments, where AI models are created and refined, and deployment hubs, where these models are put to work.

Development AI data centers can be located anywhere in the world and may even be disconnected from the internet for security reasons. On the other hand, deployment AI data centers need to be closer to the point of service, particularly when latency is a factor. For example, a real-time AI model used to manage traffic would need to process data from vehicles quickly, making a local deployment necessary to avoid excessive latency.

While not all AI deployments are time-sensitive, the need for low-latency, localized data centers will drive demand for more specific, smaller installations.

3. Superior Thermal Management

The electricity consumed by hyper-scale data centers is becoming a growing concern for climate scientists, and AI is expected to further intensify this issue by converting more energy into heat. AI could help manage data center power use by adjusting operations based on local weather conditions, reducing reliance on air conditioning in hot summer months.

AI-driven data centers will also need to be built with thermal management in mind. This could involve using natural heat sinks like solid rock or taking advantage of cooler northern (or southern) latitudes. In some cases, excess heat could even be repurposed to benefit nearby communities, much like how geothermal energy is harnessed in Iceland.

Gone will be the days of air-cooled data centers. As AI brings increased rack densities, new facilities are likely to use liquid cooling to handle the added thermal load. Every data center will need a detailed thermal emissions plan to avoid making poor environmental decisions, such as venting hot air into already hot regions.

4. Minimal Headroom Operations

AI data centers are incredibly capital-intensive, so it’s crucial to maximize their efficiency to get the full return on investment. Much like the just-in-time manufacturing model, the goal is to minimize waste and optimize performance. In the past, unused capacity in a data center was considered a selling point; with AI, however, the opposite is true.

Operating at maximum capacity for extended periods will lead to higher temperatures and voltages, so cooling and electrical systems will need regular assessments and enhancements. Additionally, with AI’s increased need for power and water, facilities will need significant investment to meet these demands.

The growing complexity of maintaining such high-efficiency operations will require a local engineering workforce, available around the clock. The days of unmanned data centers, managed remotely by engineers from far away, are numbered. Like the rigorous maintenance schedules in the airline industry, AI data centers will require constant attention to ensure optimal long-term operation.

5. AI Running AI

The final irony of AI data centers is that while engineers today are designing these advanced facilities, future data centers will likely be designed by AI itself. AI will adapt and optimize the data center infrastructure based on workload demands, managing power, heat, and service delivery with minimal human intervention.

AI will also play a crucial role in security, continuously monitoring for physical or network intrusions and adjusting defense systems in real-time. Unlike conventional antivirus software, which relies on pattern-matching, AI will use data from various sensors, network traffic, and video surveillance to maintain an adaptive, always-on defensive posture.

While this may seem like a step toward a dystopian future, the role of AI in running data centers will ultimately depend on how much control human engineers retain. The evolution of data centers driven by AI will redefine the way these vital systems operate and serve the world.