Sustainability vs. Performance? Data Center Cooling Shouldn’t Have To Choose

AI’s advances have pushed the data center industry to a seeming fork in the road.

To the left: performance. Last week, Jensen Huang’s keynote at NVIDIA GTC DC 2025 indicated TDPs and rack densities are set to skyrocket—a sure sign that cooling needs to drastically improve to satisfy next-gen GPUs’ thermal demands.

To the right: sustainability. The recent boom in data center construction has thrown a national energy crisis into stark relief, driving community-led pushback on greenfield builds. With $64 billion in U.S. data center projects blocked or delayed by “a growing wave of local bipartisan opposition,” it’s clear that efforts must be made to boost efficiency and curb energy consumption if the industry seeks to continue its current growth.

Several popular cooling methods like single-phase direct-to-chip (1P D2C) have to forever navigate a fraught path between these two ideals. Fully committing to peak performance, for example, translates to a massive spike in energy costs; meanwhile, an emphasis on sustainability risks your data center losing its competitive edge. This balancing act therefore requires constant analysis, tuning, and maintenance—and a constant headache for data center operators.

Fortunately, there’s a cooling option that wholly sidesteps this dilemma. The basic nature of two-phase direct-to-chip (2P D2C) means it doesn’t need to treat “sustainability vs. performance” as an “either/or” proposition; instead, it’s capable of maintaining the industry standard for direct-to-chip cooling (at 4500W per socket) while still delivering up to 32% energy savings vs. single-phase. In short—it’s here to deliver sustainable cooling without compromise.

The single-phase balancing act

There’s a saying in winter sports: Skiing is easy to learn, but harder to master. Snowboarding is harder to learn, but effortless once you’ve mastered it.

Single-phase D2C is like skiing. The familiarity of its water-based cooling at first made it easy to adopt. Now, however? Operators have become increasingly confronted by the sheer vigilance and maintenance that single-phase demands to keep it at its best.

That’s because 1P D2C is perpetually trapped in a liquid state—a liquid whose temperature remains an erratic variable throughout its cooling cycle. After this liquid (a blend of water and PG25) enters the cold plate, it absorbs heat and exits warmer. Much warmer, in fact: typically, there’s a 10℃ rise between inlet and outlet temperatures. Of course, that’s just a typical scenario. The “spiky” behavior of AI workloads further complicates the matter: at a moment’s notice, AI can compel cooling demands to surge. It’s difficult to 100% predict when surges will occur, so operators tend to assume—and prepare—for the worst to occur.

How do they do this? It’s simple: colder facility water (FW), which maintains GPU performance despite the spikes—but also inflates facility energy costs and subsequently sacrifices sustainability. While a warmer FW could save energy on chillers and improve PUE, it carries the implicit risk of GPU throttling if workloads soar past the usual limits.

Simply put, with single-phase, you’re forced to pursue one of two paths: hotter water for greater energy savings (and greater risk of throttling); colder water for less risk (and less sustainable cooling). Operators who seek to “thread the needle” need to contend with a cooling medium that’s constantly changing—leading to a maddening, neverending chase for equilibrium.

Two-phase: Performance and sustainability in sync

The reason single-phase is forever tied to this tradeoff is the same reason two-phase can eliminate it entirely; namely, its cooling medium. 

Two-phase’s dielectric refrigerant undergoes phase change (i.e., boiling from a liquid to a vapor) to capture and transfer heat away from critical infrastructure. The boiling process occurs at a consistent temperature, which means it likewise keeps the chip at a consistent temperature as vapor exits the cold plate. Meanwhile, as vapor leaves, liquid dependably fills the void it leaves behind, ensuring a stable flow of cooling throughout your product’s lifecycle. 

Stability, consistency, dependability—these terms connote the “set-it-and-forget-it” nature of 2P D2C, which maintains cooling equilibrium without constant intervention. Beyond that, it means operators won’t need to set a two-phase cooling system at “worst-case” settings, eliminating sustainability for the sake of eliminating risk. In fact, boiling allows for a 6-8℃ headroom buffer, allowing your system to handle power spikes without throttling. Additional energy savings thanks to 2P D2C’s lower CDU pumping power and associated maintenance costs can also deliver up to a 35% reduction in annual OpEx vs. single-phase, according to a study conducted with Jacobs Engineering. 

Put to scale, these benefits aren’t just theoretical; they deliver a tangible, real-world impact. Based on ASHRAE analysis, 2P D2C’s 6-8℃ increase in FW temperature could enable free cooling for Singapore data centers 97% of the year, compared to 24% of the year for 1P D2C. That staggering difference in free cooling frequency translates to an enormous win for energy and financial savings alike—while still delivering needed power to next-gen compute.

Let’s return to our sports analogy. If 2P D2C is like snowboarding, that means it’s harder to learn but effortless to master. Fortunately, that’s where this analogy falls apart. The difficulty in 2PD2C lies in its design and engineering, not its operation. That means we’ve handled the hard part for you—you can simply enjoy the ride.

No more tradeoffs

The breakneck pace of AI’s innovation and power demands means we can no longer consider an “either/or” future. Instead, it’s imperative we can implement solutions that deliver both stellar thermal performance and a needed boost in efficiency.

2P D2C can deliver all of the above. While newer, we’ve mastered this technology’s learning curve, overcoming its frontloaded complexity with a streamlined design. All that’s left is for you to experience its runaway gains in sustainability and performance right out of the box.