Darwin, Dinosaurs And Data Centers: How The AI Revolution Is Forcing A Structural Evolution Of The Industry
Kevin Brown is SVP and CMO, Secure Power & Data Centers at Schneider Electric.
gettyDinosaurs never went extinct. They are around us every day—we call them birds.
I was familiar with this concept, but I didn’t fully grasp its implications until I read Steve Brusatte’s The Rise and Fall of the Dinosaurs.
Sixty-six million years ago, the impact from a catastrophic asteroid triggered a global ecological bottleneck. For the dinosaurs that survived the initial blast, the world became an unforgiving wilderness of starvation and severe climate shifts. Survival favored animals that could cope with the collapse of food webs and rapidly changing environments.
Small avian dinosaurs had already developed many of the traits associated with modern flight: lighter body frames, beaks instead of heavy teeth in some lineages and feathered forelimbs. When the catastrophe wiped out their massive non-avian relatives, the bird lineages that survived were positioned to take advantage of newly vacant ecological niches. When the dust settled, birds were among the major groups to diversify in the changed landscape.
The data center industry faces a similar evolutionary bottleneck. The rapid deployment of artificial intelligence is placing unprecedented stress on digital infrastructure, both technologically and politically. If the biological analogy holds, we are on the cusp of an intense era of architectural innovation.
For decades, data centers evolved at a relatively predictable, incremental pace. Facilities grew outward, spreading across vast acreage to house rows of standard x86 servers. These architectures relied primarily on air cooling and operated at relatively low power densities of 5 to 15 kilowatts (kW) per rack. Over the past 15 years, the push for energy efficiency, measured in part by Power Usage Effectiveness, or PUE, led the industry to adopt more efficient cooling systems, including evaporative cooling, which uses water to reject heat and can save electricity.
Modern AI workloads have shattered this long-standing equilibrium. Training large language models requires dense clusters of specialized GPUs that can push rack power demands past 200 kW, with 1-megawatt (MW) racks on the horizon. At these densities, conventional air cooling becomes increasingly difficult and energy-intensive; liquid cooling can remove heat much more efficiently because of its greater heat capacity and heat-transfer characteristics. Furthermore, delivering that volume of power into a single rack is driving the industry to explore higher-voltage power delivery, including 800 VDC architectures, as an alternative to today’s traditional 400/480 VAC systems.
To avoid obsolescence, data centers must remodel their internal and external physical designs. Supply chains, hardware design and deployment speed need to scale at a pace so fast it’s never been experienced.
This shift is no longer only an engineering challenge; it is a community issue. Communities are voicing concerns over power consumption, noise and resource usage—specifically water—in light of the growing footprint of these facilities.
Historically, the industry relied on evaporative cooling, essentially sweating away millions of gallons of water to reject heat into the atmosphere. The industry traded water consumption for lower electricity usage.
Closed-loop liquid cooling systems reverse this trade-off. These systems operate much like an automobile engine: Liquid circulates through a closed circuit to transfer heat from the processor to a heat exchanger. Just as a car rarely requires additional coolant, closed-loop server cooling consumes no ongoing water. Water consumption is not an inherent trait of data centers—it is an architectural choice.
Furthermore, modern AI chips run safely at higher baseline operating temperatures. This higher temperature differential reduces the energy penalty of liquid-to-air heat rejection. It is now entirely practical to build high-density data centers that consume virtually no water.
Technology can mitigate noise by using acoustic baffling, bigger fans and applying variable frequency drives.
However, technology alone will not secure the industry’s survival; cultural evolution is equally critical.
The legacy data center strategy relied on secrecy—unmarked, fortress-like facilities hidden behind high security fences with minimal community outreach. For mission-critical infrastructure, keeping a low profile was standard operating procedure.
Today, data center developers must move toward more transparency and symbiotic design. It’s possible for data centers to be built as architectural assets rather than “industrial blights.” It’s possible for data centers to have dedicated energy generation to protect local power grids from cost spikes. Many developers already are engaging more transparently during planning phases. I don’t mean to imply that any of this is easy, but it can be done.
Just as evolutionary stressors helped reshape theropods along the path to modern birds, the AI revolution is forcing a structural evolution of the data center industry.
To paraphrase a quote often attributed to Charles Darwin that was originally coined by business professor Leon C. Megginson: “It is not the strongest of the species that survives, nor the most intelligent that survives. It is the one that is most adaptable to change.”
Under the severe pressure of the AI deployment wave, data center technology will evolve, the culture will adapt and the infrastructure will survive.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
