It’s getting increasingly obvious that AI isn’t just changing software – it’s reshaping the physical world of data centers too, and that’s the core message coming out of Vertiv’s new Frontiers report.
The company explains that macro forces like extreme densification, gigawatt-scale deployments, and the idea of treating the data center as a single “unit of compute” are now defining how facilities are designed and operated, while silicon diversification is adding another layer of complexity that infrastructure must adapt to.
One of the biggest shifts highlighted is around power, as most data centers today still rely on hybrid AC/DC power with multiple conversion stages, which is increasingly inefficient as rack densities climb due to AI workloads, so Vertiv points toward higher voltage DC architectures as a way to reduce losses, simplify distribution, and better support future high-density environments, especially as on-site generation and microgrids become more common.
At the same time, the report suggests that AI itself won’t be centralized forever. While hyperscale AI data centers have absorbed billions in investment to support large language models, Vertiv believes more organizations will move toward distributed AI, particularly in regulated sectors like finance, healthcare, and defense, where on-premise or hybrid deployments are often necessary for data residency, security, and latency reasons, which in turn drives demand for flexible, scalable power and liquid cooling in both new builds and retrofits.
Energy autonomy is another area gaining urgency. Traditionally, on-site generation existed mainly for backup, but power availability constraints are now pushing operators toward longer-term autonomy strategies, including concepts like Bring Your Own Power (and Cooling), with natural gas turbines and other technologies increasingly seen not just as resilience tools but as strategic enablers for AI-scale growth.
Digital twin also comes in clutch at this point in time due to the flexibility the technology provides, allowing operators to virtually design, test, and integrate entire data center environments before deployment, often using prefabricated modular approaches, helping reduce deployment timelines and potentially cutting time-to-token by up to 50%, a major advantage when the industry is racing toward gigawatt-scale AI buildouts.
All of this is backed by mission-critical cooling, with liquid-based solutions rapidly moved from niche to necessity for high-density AI infrastructure, and Vertiv believes the next evolution will be AI-enhanced liquid cooling systems that use deeper monitoring and predictive intelligence to anticipate failures, optimize performance, and improve overall resilience. The irony, and opportunity, is that AI workloads are not only driving the need for advanced cooling but could also be the tool that makes these systems smarter and more reliable over time.
In general, the industry is rapidly evolving in response to AI factories, with higher voltage DC power architectures, advanced liquid cooling, on-site energy generation, and digital twins all becoming critical enablers to support the scale and speed AI now demands.
