
How to Optimize Data Center Power Usage to Increase AI Compute Capacity
January 28, 2026
Airsys Announces First European Manufacturing Facility Expanding Global Production and Regional Customer Support
February 4, 2026Article by Paul Quigley
PTC has a way of clearing the fog.
When you put operators, investors, utilities, technologists, and policy people in the same rooms, the real takeaways don’t come from the slides. They come from the patterns that repeat across conversations, panels, and the questions people ask afterward in the hallway.
This year, those patterns were unusually consistent.
Yes, the AI infrastructure boom is real. Capital is still pouring in. Campuses are getting bigger, racks are getting denser, and power numbers are getting louder. But what became clear at PTC is that the industry is no longer debating whether demand exists. The conversation has shifted to where risk is building, who is carrying it, and how long the current approach holds together.
That shift showed up everywhere, including in the panel I participated in.
Demand Isn’t the Problem. Structure Is.
Across the conference, the same underlying truth kept surfacing: demand is overwhelming, but it is also highly concentrated and structurally uneven.
A small number of hyperscale players now shape construction timelines, power allocation, and pricing norms across entire regions. That scale creates efficiency, but it also introduces fragility. When demand gets this big and this centralized, infrastructure decisions stop being incremental. They become binary.
One dynamic that kept coming up, sometimes explicitly and other times with a knowing smile, was the industry’s obsession with announcing ever-larger power numbers. After listening to presenter after presenter talk about massive gigawatt pursuits, I started referring to this era as one of “bragawatts.” Everyone is racing to talk about how much power they plan to consume.
But megawatts alone don’t tell you whether that power can be delivered efficiently, economically, or responsibly. And they don’t tell you who’s holding the bag when assumptions change.
Power Has Replaced Real Estate as the Constraint
If there was one theme that cut across nearly every session, it was power.
Not just how much of it, but:
- where it comes from, before or after the meter
- how fast it can be delivered
- how stable the pricing is
- and how utilities and communities respond to it
The industry has crossed a line. Power is no longer just an input underneath the building. It’s the gating factor that determines whether a project happens at all.
That reality came through clearly during our panel discussion. Regardless of perspective — operator, utility, or investor — there was surprisingly little discussion about how to ensure the future isn’t just about chasing more power. There wasn’t enough time spent on how to use the power we already have more intelligently.
That shift in mindset feels inevitable, but it’s not yet top-of-mind.
Data Centers Are Being Viewed as Energy Systems
Another thing that stood out at PTC is how much the underwriting lens is changing.
Data centers are no longer being viewed purely as real estate. Investors may not have fully landed on a rallying framework yet, but they’re moving closer to valuing data centers — and their cooling methodologies — as energy conversion systems.
The questions I heard repeatedly sounded like this:
- How exposed is the facility to power price volatility?
- How adaptable is the design as AI workloads evolve?
- What happens if today’s density assumptions don’t age well?
One question in particular is beginning to replace an older one: not just how efficiently does this facility run, but how effectively does it turn power into compute. That line of thinking is at the heart of Power Compute Effectiveness (PCE).
At the same time, the concept of Return on Invested Power (ROIP) resonated strongly with analysts and investors. It reframes power as a capital asset, not just an operating cost, and ties productivity, risk, and return together in a way that shows up clearly on balance sheets.
Different lenses, same destination.
Challenging Assumptions on Cooling and Temperature
One of the more pointed moments at PTC came during discussions around operating temperatures for AI infrastructure.
Alongside Joe Capes, CEO of LiquidStack, we pushed back on the idea that 45°C entering temperature represents some kind of new ceiling. That framing simply doesn’t reflect what we already know to be possible.
We have lab-proven results showing that 60°C entering temperatures are achievable today. Moving in that direction isn’t about extremes — it’s about practicality.
Higher entering temperatures allow compressors to be removed. That reduces energy use, lowers noise, and directly addresses NIMBY concerns. It also aligns with growing utility pressure to stabilize grids and reduce unnecessary load.
This isn’t theoretical. It’s about matching infrastructure design to physics, grid realities, and community expectations.
A Moment That Really Stuck With Me
One of the most rewarding moments of the week didn’t happen on stage.
We were staying at a large beachfront resort campus. During the panel, I challenged people to do something simple when they got back to their rooms: step out onto the balcony and look around at the other hotel buildings. Look at the roofs. Look at the scale.
Then I asked them to think about what’s happening inside those buildings every day — laundry running constantly, kitchens operating from morning to night, showers, pools, spas, cleaning crews. All of it driven by hot water, most of it produced by gas or electric boilers.
I wasn’t making a new claim. Heat reuse is something we talk about often. It’s something we’ve been pushing in our marketing and many of my presentations for years.
But placing that idea in a setting everyone could see and feel made it click.
More than once, people I’d never met stopped me in the hallway or while I was walking with my wife and said some version of the same thing:
“I always thought heat reuse was a nice idea, but only practical in cold climates or very specific cases. Looking out over my balcony and thinking about what you said made it feel real.”
That mattered to me. Not because the idea was new, but because it finally felt touchable.
Why Liquid Cooling and LiquidRack Resonated
Against all of this, it wasn’t surprising that liquid cooling — and LiquidRack specifically — was received so positively.
What resonated wasn’t performance for performance’s sake. It was versatility and feasibility, combined with a fresh way of thinking about so-called stranded data centers.
Liquid cooling came up again and again as a way to:
- extend the life of powered-and-permitted assets
- unlock constrained or stranded power
- support higher densities without tearing down entire facilities
People stopped thinking strictly in terms of greenfield versus brownfield and started thinking about how to get more value out of what already exists.
Powered and Permitted Hit Home
That’s also why the “Powered and Permitted” concept landed so well.
In a world constrained by grid access, permitting timelines, and community pushback, assets that already sit inside the fence line are becoming strategically important again. The idea that progress might come from freeing those assets, not bypassing them, resonated across the room.
It reframes innovation away from endless expansion and toward smarter use.
The Bigger Lesson from PTC
The biggest thing I took away from PTC is that the industry is quietly moving from acceleration to stewardship.
Whether driven by regulation, balance-sheet pressure, power scarcity, community reaction, or simply the need for better returns, the change in focus is coming — and it will be more openly discussed by the time PTC 2027 rolls around.
Growth isn’t stopping. But the definition of success is changing.
Those who simply consume power will face increasing friction.
Those who control power will gain leverage.
Most importantly, those who optimize power early will build resilience.
Cooling, efficiency, and thermal design are no longer supporting details. They are central to how risk is managed and value is preserved in a power-constrained world.
PTC made one thing clear: the future of data centers won’t be defined by who announces the most bragawatts, but by who builds systems that hold together as conditions change.
That’s the phase we’ve entered. And it’s where the real work begins.
