Artificial intelligence is no longer just another workload inside the data center—it’s reshaping the data center itself. That’s the central message of Vertiv’s new Frontiers report, which examines how macro forces tied to AI are driving fundamental changes in how facilities are designed, powered, cooled, and operated.

Drawing on expertise from across Vertiv’s engineering and technology teams, the report expands on the company’s annual data center trends outlook, offering a deeper look at the forces pushing the industry toward what Vertiv increasingly describes as “AI factories.” These facilities are denser, faster to deploy, and more tightly integrated than anything that came before them.

“The data center industry is continuing to rapidly evolve how it designs, builds, operates and services data centers, in response to the density and speed of deployment demands of AI factories,” said Scott Armul, Vertiv’s chief product and technology officer. According to Armul, cross-technology pressures—especially extreme densification—are accelerating shifts toward higher-voltage DC power, advanced liquid cooling, on-site energy generation, and digital twins.

Together, these changes point to an industry in the middle of a structural reset.

The macro forces reshaping data centers

At the heart of the Frontiers report is Vertiv’s identification of four macro forces that are redefining data center innovation.

First is extreme densification, driven primarily by AI and high-performance computing workloads. GPU-rich racks are pushing power densities far beyond what legacy facilities were designed to handle, stressing everything from power distribution to cooling systems.

Second is gigawatt scaling at speed. AI demand is forcing operators to deploy capacity faster—and at larger scale—than ever before. Hyperscale-style growth is no longer limited to cloud giants; enterprises, governments, and AI-native companies are now planning facilities measured in hundreds of megawatts or more.

Third is the idea of the data center as a unit of compute. In the AI era, facilities can no longer be treated as collections of loosely coupled systems. Power, cooling, IT, and software must be designed and operated as a single, tightly integrated platform.

Finally, silicon diversification is complicating infrastructure planning. Data centers must now support a growing mix of CPUs, GPUs, accelerators, and custom silicon, each with different power, cooling, and operational profiles.

These forces set the stage for five technology trends that Vertiv believes will define the next phase of data center evolution.

Powering up for AI

Power architecture sits at the center of the AI data center challenge. Most facilities today still rely on hybrid AC/DC power distribution, with multiple conversion stages between the grid and the IT rack. While proven, this approach introduces inefficiencies that become increasingly problematic as rack densities climb.

AI workloads are exposing those limits. According to Vertiv, the industry is moving toward higher-voltage DC power architectures, which reduce current, shrink conductor size, and eliminate some conversion stages by centralizing power conversion at the room level.

Hybrid AC/DC systems remain common, but as standards mature and equipment ecosystems develop, full DC architectures are expected to gain traction—especially in high-density environments. The shift is further reinforced by on-site generation and microgrids, which naturally align with DC-based distribution.

In practical terms, power design is becoming less about incremental efficiency gains and more about enabling scale. Without rethinking power delivery, gigawatt-class AI deployments simply won’t be feasible.

Distributed AI changes where compute lives

The first wave of AI investment focused heavily on centralized hyperscale data centers built to train and run large language models. But Vertiv’s report suggests the next phase will be more distributed.

As AI becomes mission-critical, organizations will make more nuanced decisions about where inference workloads run. Factors such as latency, data residency, security, and regulatory compliance are pushing some industries toward on-premises or hybrid AI environments.

Highly regulated sectors—including finance, defense, and healthcare—are prime examples. For these organizations, sending sensitive data to public clouds may not be an option, even if cloud-based AI services are readily available.

Supporting distributed AI requires flexible, scalable infrastructure, particularly high-density power and liquid cooling systems that can be deployed in new builds or retrofitted into existing facilities. This trend blurs the traditional line between hyperscale and enterprise data centers, bringing AI-class infrastructure closer to the edge.

Energy autonomy accelerates

Resiliency has always required on-site power generation, but AI is changing the equation. Power availability—not just reliability—is becoming a limiting factor for new data center projects in many regions.

Vertiv notes that extended energy autonomy is emerging as a strategic priority, especially for AI-focused facilities. Investments in natural gas turbines and other on-site generation technologies are increasingly driven by grid constraints rather than backup requirements alone.

This shift is giving rise to strategies like “Bring Your Own Power (and Cooling)”, where operators design facilities around self-generated energy and tightly integrated thermal systems. While capital-intensive, these approaches offer more predictable scaling and faster time to deployment in power-constrained markets.

Energy autonomy also intersects with sustainability goals, forcing operators to balance capacity expansion with emissions considerations and long-term regulatory risk.

Digital twins move from planning tool to necessity

As AI infrastructure grows more complex, traditional design and deployment processes are struggling to keep up. Vertiv’s report highlights digital twin technology as a critical enabler for speed and scale.

By using AI-driven digital twins, operators can virtually model entire data centers—including IT, power, and cooling systems—before anything is built. These virtual environments allow teams to validate designs, optimize layouts, and integrate prefabricated modular components.

The payoff is speed. Vertiv estimates that digital twin-driven approaches can reduce time-to-token by up to 50%, a metric that matters deeply in competitive AI markets. Faster deployment means faster access to compute, which can translate directly into business advantage.

Digital twins also support the concept of the data center as a unit of compute, reinforcing tighter integration between physical infrastructure and AI workloads.

Adaptive, resilient liquid cooling

Liquid cooling has rapidly moved from niche to necessity as AI workloads push beyond the limits of air cooling. But Vertiv argues that cooling innovation isn’t stopping at adoption—it’s becoming smarter.

AI itself is now being applied to optimize liquid cooling systems, using advanced monitoring and control to predict failures, manage fluid dynamics, and improve overall resilience. In high-value AI environments, where downtime can be extraordinarily expensive, predictive cooling intelligence could significantly boost uptime and hardware longevity.

As liquid cooling becomes mission-critical, adaptive systems that learn and respond in real time may become a standard expectation rather than a premium feature.

Why this matters beyond data centers

While the Frontiers report is focused on infrastructure, its implications ripple outward into cloud strategy, AI economics, and even MarTech and AdTech ecosystems. AI-driven services—from personalization engines to real-time analytics—ultimately depend on the scalability and reliability of the underlying compute layer.

If AI factories struggle with power, cooling, or deployment speed, innovation at the application layer slows as well. Conversely, breakthroughs in infrastructure efficiency can lower costs and expand access to AI capabilities across industries.

Vertiv’s framing also underscores a broader industry truth: AI transformation isn’t just about models and software. It’s equally about electrons, heat, and physical space.

A data center industry in transition

Vertiv operates in more than 130 countries, spanning power management, thermal management, and IT infrastructure from the cloud to the edge. That breadth gives the company a wide-angle view of how infrastructure demands are changing—and how quickly legacy assumptions are being challenged.

The Frontiers report makes it clear that incremental upgrades won’t be enough. AI is forcing a rethinking of foundational design choices, from power architecture to cooling strategy to how facilities are conceptualized and delivered.

As Armul puts it, gigawatt-scale AI innovation depends on embracing these shifts. The data center, once a supporting actor, is now a central character in the AI story—and its evolution may determine how fast the next wave of AI progress arrives.

Get in touch with our MarTech Experts.