The AI boom is driving unprecedented demand for computing power, leading to a race to build larger and more efficient data centers.
Legacy infrastructure is struggling to handle the thermal load from AI, creating a bottleneck in the industry’s growth.
Ian Parker of facilities-level cooling company Healixa joined us to discuss his company’s energy-efficient cooling solutions for data centers.
The technology reduces stress on power grids, conserves water in stressed global markets, and can increase AI computing capacity.
The AI rush is creating an insatiable demand for computing power that makes the demands of previous tech cycles look quaint. While the emergence of cloud was largely dominated by U.S.-based companies, AI is complicated by geopolitical tensions and sovereign strategies led by the world’s most powerful nation-states. The shift is driving a fundamental rewiring of the global economy, driven by a push for capacity that is exploding at an unprecedented scale. Every major player is racing to build bigger, faster, and more powerful data centers to stake their claim.
But digital ambition is running headfirst into a very physical wall (emphasis on the physics in physical!). The silent bottleneck threatening to stall the entire AI revolution isn’t the software, the algorithms, or even the supply of chips. It’s the buildings themselves—the legacy infrastructure buckling under a thermal load it was never designed to handle.
We spoke with Ian Parker, Chairman of the Board at Healixa, to discuss his company’s role in driving AI progress by helping to cool hyperscalers at the facilities level. Healixa is responsible for developing a closed-loop thermal system called FrostPulse that provides energy efficient cooling applications for data centers. While the world is mesmerized by what AI can do, Parker and his team have been obsessed with a more fundamental problem: what to do with all the waste heat.
Under-built infrastructure: In a recent LinkedIn post, Parker warned, “AI isn’t just pushing compute. It’s stressing infrastructure.” He pointed to a recent partnership between NVIDIA and Schneider Electric to develop next-gen water cooled data centers. He noted that cooling is no longer background infrastructure as the hard limits of legacy systems struggle to keep up with the pace of innovation. “The demand for AI is driving an endless push for more capacity,” Parker said. “And that push is not going away.” Data center rack densities are set to skyrocket from a legacy 20 kilowatts to “hundreds of kilowatts” in the coming months and years. The core problem, according to Parker, is that the industry is trying to run this revolution on an outdated physical footprint. “The problem becomes you have a structure that’s not built for that.”
No free lunch for hyperscalers: “There’s no free lunch in physics,” Parker joked. “More work equals more heat.” This reality is creating a cascading crisis for hyperscalers, who Parker notes are on a “very clear path” to building massive infrastructure and will inevitably face a huge thermodynamics problem. The consequences are twofold: a water crisis, as traditional evaporative cooling consumes millions of gallons annually in water-stressed regions like the US Southwest and India, and a power crisis, with cooling accounting for an unsustainable 30-40% of a data center’s entire operational cost.
Healixa’s strategy is one of pragmatic discipline. They are not trying to compete in the AI race itself but are instead positioning themselves as an indispensable supplier to all the competitors.
The picks and shovels play: “There are the people that are going mine the gold out of the AI Boom, and then there are the people that are going to sell them picks and shovels,” Parker said. “We are proud to be the later.” This means deliberately avoiding the crowded “componentry game” and instead providing the critical off-ramp for heat at the facility level. It’s a defensible business model built on patents and a realistic forward-looking view of the market.
The efficiency gains are not just about saving money or energy. Parker presented a fascinating choice: a facility can either use the savings to reduce stress on the grid, or it can use its existing power allocation to dramatically increase its AI computing capacity.
Unlocking capacity and doing good: “One argument is that our technology takes stress off the grid. The other is that maybe the demand stays the same, but we build out our computing capacity much faster because of it,” he said. “Either one of those is a good thing.”
A global view: This vision extends globally, particularly to places like India, where the technology isn’t just a business solution, but a humanitarian one. By eliminating the need for water-intensive cooling, Healixa can help alleviate stress on the country’s already-taxed water infrastructure, ensuring more of that critical resource is diverted to the population. It’s a mission that brings the company’s entire value proposition to its ultimate conclusion, and a rarity in the high-stakes world of AI dominance.
