It’s well understood across the data center industry and beyond that AI has completely transformed the digital landscape. From the way facilities are designed and built to the next-generation technology and infrastructure that supports them, setups and processes that were commonplace just a couple of decades ago are now almost unrecognizable – and continuously being upgraded to meet what comes next.

But this sweeping transformation demands closer examination. Unpacking the complexity of the AI-driven landscape to identify the direction of design and operations, and to determine which trends are already shaping how operators build and operate today, and how they will evolve in the future is the focus of the Vertiv Frontiers report.

Building on the company’s previous annual Data Center Trends predictions, the report presents Vertiv’s analysis of the macro forces and technology trends driving both current and future data center innovation.

The result is a clear and actionable framework through which the nuanced and far-reaching developments can be better understood and managed. Scott Armul, chief product and technology officer at Vertiv, defines the report’s core message:

“The evolution of AI depends on highly integrated digital infrastructure deployed at unprecedented speed and scale. The data center industry must keep pushing the frontiers of innovation, driving advances from grid to chip, chip to heat reuse, while preparing for the next wave of breakthroughs.”

Macro forces

Modern data centers are characterized by more of everything: higher densities, larger campuses, greater volumes of equipment, and increased pressure to deliver always-on operations while consistently upgrading to keep pace.

Part of this densified reality means that everything is increasingly interconnected. Vertiv has identified four macro forces fueled by the rise of AI and accelerated compute that are driving innovation in tandem: extreme densification, gigawatt scaling at speed, the data center as a unit of compute, and silicon diversification.

In response to these macro forces, the Frontiers report also outlines five key technology shifts shaping the landscape: power architecture evolution, distributed AI, energy autonomy, digital twin-driven design, and adaptive liquid cooling.

“At the heart of this transformation is extreme densification,” states Armul. “It’s the defining macro force whose effects are felt across the entire data center and technology landscape. The additional macro forces align to this shift, representing broad changes that extend from the chip level to system-level integration and full facility design.”

Rack power densities are now being pushed well beyond 25kW and increasingly into triple digits. Systems that once occupied a full row or even an entire hall are now slotted into tightly integrated rack-scale deployments. This concentration of compute power inevitably introduces complexities across power delivery, thermal management, and physical infrastructure design.

Emerging alongside this densification is gigawatt scaling at speed. Data centers are now being designed and deployed at unprecedented scale, with new facilities moving more rapidly than ever from concept to construction as operators race to meet surging demand.

Simultaneously, the modern data center is increasingly being designed and operated as a single system rather than a collection of independent components. Power delivery, thermal management, and computing platforms must work in tandem as tightly integrated architectural components, pushing the industry toward deeper integration across the full stack of digital infrastructure.

Finally, comes silicon diversification. The chips powering AI are rapidly evolving beyond traditional GPU-centric architectures to include custom silicon and a growing variety of accelerators. This diversification introduces new performance profiles, plus thermal and power requirements that the infrastructure must accommodate.

“Silicon diversification is probably the most underestimated,” explains Armul. “Extreme densification gets most of the attention, and rightly so. But the chips powering AI are already diversifying to include custom silicon and other form factors. Future data center infrastructure must be designed and optimized to support the full spectrum of chips and compute.”

1) Rethinking power architectures

Among the trends identified in the Frontiers report, the evolution of data center power architecture is perhaps the most immediate concern for operators planning AI-ready facilities.

Traditional AC-based architectures have served well for decades, but the rapidly increasing power demands of AI workloads are pushing their limits.

As rack densities move beyond 100kW and toward 300kW in high-performance environments, the amount of copper required for traditional AC distribution increases dramatically, along with thermal losses and spatial constraints.

Hybrid AC/DC systems remain the norm today and will likely continue to do so in the near term. However, the trajectory is becoming increasingly clear as operators explore higher-voltage DC distribution.

“AC power will be used more selectively, including at the grid interface, which may be relegated to backup in some scenarios, with DC becoming an important approach for internal power distribution,” says Armul.

The shift toward DC architectures is also closely linked to the growing role of onsite energy generation and storage. Technologies such as solar power, fuel cells, and battery energy storage systems produce native DC power, making DC distribution inside the facility increasingly attractive as a highly efficient solution.

“Operators and decision-makers can take immediate steps toward planning future-ready AI factories that will support these coming generations of AI servers,” adds Kyle Keeper, senior vice president of power management at Vertiv.

Higher-voltage systems introduce new safety considerations, plus the need for specialized installation expertise, and may involve greater upfront costs during the early stages of adoption – meaning this transition won’t occur overnight. But for operators planning for sustained growth in high-density compute, exploring these architectures today may prove critical for ensuring long-term scalability.

2) The rise of distributed AI

While hyperscale AI training clusters dominate headlines, the rise of distributed AI and on-premise inference is quietly reshaping the infrastructure landscape.

Highly regulated industries are increasingly deploying AI workloads closer to their data sources, often within private or hybrid infrastructure environments. Financial institutions, defense organizations, and healthcare providers frequently face strict data residency and latency requirements that make reliance on remote hyperscale infrastructure impractical for certain applications.

“As data gravity and regulatory forces intensify, organizations will face pivotal choices about whether to build this foundation through trusted service providers or invest in their own next-generation data center infrastructure,” says Martin Olsen, vice president of segment strategy and deployment at Vertiv.

This trend is expected to grow as enterprises develop more specialized AI capabilities tailored to their own data sets and operational contexts. Instead of relying solely on generalized models hosted by large cloud providers, organizations are beginning to build domain-specific AI systems designed around their own intellectual property and regulatory constraints.

Distributed AI doesn’t necessarily replace centralized hyperscale infrastructure. Rather, it expands the overall architecture of AI computing to create a more diverse ecosystem of training hubs, enterprise facilities, and Edge inference deployments.

In practice, this means planning for infrastructure that can support both large, centralized clusters and smaller, high-performance environments closer to end users or regulated data sets.

3) Data centers as energy ecosystems

Power availability has quickly emerged as one of the most significant constraints on AI infrastructure expansion. The increased energy requirements of AI workloads are placing unprecedented pressure on existing electrical grids, particularly in regions where data center development is most intense.

The scale of this challenge is already abundantly evident. As outlined in the Frontiers report, data centers consumed approximately 1.9 percent of total electricity production in the US in 2018. That figure has climbed to around 4.5 percent and could approach six percent in 2026 as deployments continue to expand at pace.

This trajectory has driven the adoption of more diverse power generation solutions. Investment in onsite generation and microgrid technologies continues to rise in order to supplement grid capacity and maintain power availability.

“Operators may not aspire to become energy providers, but the rapid growth of AI and its massive power demands is driving those decisions,” says Peter Panfil, engineer and vice president of technical business development at Vertiv. “In many cases, that means embracing self-generation as a strategic bridge as grid capacity continues to evolve.”

These hybrid energy strategies typically combine multiple technologies – including solar generation, natural gas turbines, fuel cells, battery energy storage systems, and traditional generators. The goal is to deliver both resilience and greater operational autonomy.

With time and the right conditions, some data centers could evolve into sophisticated energy ecosystems capable of interacting dynamically with the grid. Beyond simply drawing power, these facilities may also store energy, provide demand response capabilities, and integrate renewable resources in ways that support energy stability well beyond the data center walls.

4) Digital-driven design

As facilities scale from hundreds of megawatts to gigawatt-scale campuses, traditional design and engineering approaches simply cannot support these complex realities.

As a result, digital twin technology is playing a transformative role. By creating highly detailed virtual models of data center infrastructure, operators can simulate performance, test operational scenarios, and validate design decisions before beginning physical construction.

“Digital twins can be used to create physics-based, photorealistic models of reference architecture,” the Frontiers report explains. “This enables real-time collaboration with architects, engineers, and operators, allowing rapid design iteration and validation before deployment.”

In practice, this capability can significantly reduce risk during large-scale infrastructure development. Engineers can model power flows, thermal behavior, and system interactions under a wide range of operating conditions – identifying potential issues long before they physically occur.

Steve Blackwell, vice president of engineering at Vertiv, emphasizes that digital twins also extend beyond the initial design phase: “Digital twin technology allows the entire infrastructure to be simulated, monitored, and optimized in real time before and after construction. Infrastructure no longer operates independently of computing platforms; a data center must operate as a system that continuously adapts as computing loads change.”

Maintaining accurate digital models during live operations requires continuous data integration from facility management systems and sensors throughout the infrastructure. As these feedback loops become more sophisticated, digital twins have the potential to evolve into powerful operational tools capable of optimizing efficiency, predicting failures, and guiding future expansion strategies.

5) Towards adaptive liquid cooling

Cooling infrastructure is undergoing an equally dramatic transformation as AI workloads drive unprecedented heat densities inside modern facilities.

Liquid cooling technologies have already emerged as a central solution for managing high-performance compute environments. However, the Frontiers report suggests that the next phase of cooling innovation will go beyond simple efficiency gains.

The concept of adaptive liquid cooling introduces a new level of intelligence into thermal management systems. Instead of operating as static mechanical infrastructure, cooling systems are evolving into data-driven platforms capable of monitoring conditions in real time and adjusting performance dynamically.

This shift depends on the integration of advanced sensors within cooling loops to monitor temperature, pressure, coolant quality, and flow rates. Combined with specialized data center management software and digital services, these sensors enable predictive maintenance and real-time optimization of cooling performance.

“Liquid cooling is evolving into a self-optimizing system, using AI for predictive maintenance to maintain peak performance, resiliency, and efficiency through advanced management and intelligent control,” says Nigel Gore, vice president of high-density and liquid cooling at Vertiv.

And these adaptive capabilities could extend even further. Future systems may incorporate advanced materials or automated maintenance mechanisms that allow cooling infrastructure to respond dynamically to changing operating conditions – improving reliability and performance across high-density environments.

Preparing for the next phase of AI infrastructure

Taken together, the trends, and macro forces outlined in the Vertiv Frontiers report paint a picture of an industry entering a new phase of technological transformation.

Beyond dramatically increasing demand for data center capacity – AI is fundamentally reshaping the architecture of digital infrastructure, pushing operators to rethink everything from power distribution and energy sourcing to cooling systems and facility design.

The scale and speed of this transformation mean that planning decisions made today will shape infrastructure capabilities for years to come. And as a result, operators must explore new architectures, adopt advanced modeling tools, and prepare for increasing integration across power, cooling, and compute platforms proactively, rather than reactively.

For more on Vertiv Frontiers visit The technology trends shaping the future of the data center