Why air is under pressure

AI servers concentrate more heat in fewer racks. Traditional air cooling can work at moderate densities, but the fans, room airflow and heat exchangers become harder to scale as accelerators draw more power. Liquid carries heat more efficiently than air, which is why direct-to-chip and immersion systems are moving from specialty deployments into mainstream data-center planning.

The choice is not simply “which liquid is better.” It is a decision about server design, rack density, facility plumbing, service procedures, vendor support, operating temperature, capital cost and the ability to mix old and new equipment.

Direct-to-chip basics

Direct-to-chip cooling places a cold plate on the heat-generating component, commonly the GPU and CPU. A liquid loop moves heat from the cold plate to a coolant distribution unit, or CDU, which separates the technology loop from the facility water loop. The rest of the server may still use air.

This partial approach is why direct-to-chip can be attractive for a transition. It can remove most of the heat from the highest-power components while leaving storage, memory, power supplies and some networking on air. Vertiv describes direct-to-chip systems as using single-phase cold plates or two-phase evaporation units mounted on the board’s heat-generating components.

Direct-to-chip is not a single standard. Some designs use warm-water loops, some use chilled water, and some use refrigerants or two-phase fluids. The operating temperature determines how much free cooling the facility can use and whether a chiller is needed.

Immersion basics

Immersion cooling submerges servers or selected components in a dielectric fluid that does not conduct electricity. Single-phase systems keep the fluid liquid as it absorbs heat. Two-phase systems boil the fluid at a controlled temperature and condense the vapor back into liquid.

The biggest advantage is coverage. Immersion can cool more of the system and support high rack densities without forcing every heat source to have a separate cold plate. It can also reduce reliance on server fans and room airflow.

The tradeoff is that immersion changes the service environment. Operators need tanks, fluid management, filtration, lifting procedures, compatible materials, and a way to remove and service hardware. Some components, seals, cables, drives and warranty programs are not designed for submersion without qualification.

Density and retrofit

Direct-to-chip is often easier to insert into an existing facility because the data hall can keep some air infrastructure and use CDUs at the row or rack level. That does not make it plug-and-play. The floor, piping, leak detection, water treatment, power distribution and maintenance process still need to be engineered.

Immersion can be compelling in a new hall built around a stable hardware platform. A facility designed for tanks can optimize space, pump systems and heat rejection from the beginning. Retrofitting an air-cooled room for immersion can require more demolition and a new operating model.

The decision should start with the rack, not the cooling brochure. Ask what the rack will draw today, what it may draw in two years, which components are cooled by liquid and what happens when a customer uses a different server design.

Temperature and heat rejection

Liquid cooling can raise the temperature of the facility loop enough to increase the hours when dry coolers or economizers can reject heat without mechanical chilling. That can reduce compressor energy. It does not mean every liquid-cooled system is waterless or low-energy.

A CDU may transfer heat into a facility-water loop that still reaches a cooling tower. A dry cooler may eliminate evaporative water use but require more fan power on hot days. A two-phase refrigerant system may reduce water use while introducing fluid management and environmental considerations.

Vertiv and Intel reported a pumped two-phase direct-to-chip demonstration capable of dissipating up to 170 kilowatts of IT heat load while maintaining a maximum case temperature of 56.4 degrees Celsius at a flow rate of 0.48 liters per minute per kilowatt. That is a technology demonstration, not a universal performance guarantee, but it illustrates how the system should be measured: heat removed, flow, temperature and operating conditions.

What to measure

A serious comparison needs more than PUE. Track:

• Rack power density in kilowatts.
• Percentage of IT heat removed by liquid.
• Coolant supply and return temperatures.
• CDU efficiency and pump power.
• Water use effectiveness, or WUE.
• Mechanical cooling hours.
• Dry-cooler performance at peak ambient temperature.
• Leak incidents and mean time to repair.
• Hardware service time.
• Fluid replacement and disposal cost.

The business decision

Direct-to-chip usually wins when a customer wants high-density compute quickly, wants to preserve some standard service practices and expects the server platform to evolve. Immersion may win when density is extreme, the workload is stable, the facility is new and the operator can support fluid-based maintenance.

Neither is a universal answer. Mixed cooling can be rational: air for general-purpose racks, direct-to-chip for AI rows and immersion for the highest-density or edge cases.

BOSSTOX takeaway

Cooling should be evaluated as a chain from silicon to room to atmosphere. The right question is not whether direct-to-chip or immersion sounds more advanced. It is which system delivers the required thermal margin, serviceability, water profile and cost at a dated rack density.

There is also a failure-domain question. Direct-to-chip systems keep the coolant in a controlled circuit and can isolate a row or CDU, but a leak, pump failure or clogged cold plate can affect a large set of servers. Immersion reduces some air-flow and fan failure modes, yet a tank, fluid-quality problem or service mistake can affect every system in that tank. Operators should ask how failures are detected, isolated and repaired before comparing steady-state efficiency.

The customer’s refresh cycle matters too. A facility expecting to replace accelerators every twelve to eighteen months may value a cooling architecture that preserves standard server handling. A stable scientific-computing environment with long-lived hardware may accept immersion’s deeper integration. Cooling is therefore a deployment and operations decision, not only a heat-transfer decision.

Cooling is the quiet part of AI infrastructure until it fails. Every watt used by a processor eventually becomes heat. In a conventional server room, airflow and air-conditioning may be enough. In a dense AI rack, the same approach can become inefficient, expensive, or physically inadequate.

Density is the driver

The basic measure is power density: how much electrical power is installed in a rack or a room. Older enterprise racks might operate at a few kilowatts. Modern accelerated systems can require far more. NVIDIA’s rack-scale Blackwell systems illustrate the direction of travel: dozens of accelerators, high-speed networking, and a liquid-cooled design packaged as one tightly integrated unit.

Air versus liquid

Air cooling works by moving heat from components into the room and then into a cooling system. Fans push air through heat sinks, and computer-room air handlers or in-row systems remove heat from the space. It is simple, familiar, and relatively easy to maintain. It also becomes harder as racks get denser. More air requires larger fans, more ducting, tighter containment, and greater control over pressure and temperature.

Liquid cooling moves heat with a fluid instead of relying on air alone. A cold plate sits on a processor, and a liquid loop carries heat away to a heat exchanger. The fluid can remove more heat in a smaller space because liquids generally carry heat more efficiently than air. Direct-to-chip cooling is especially useful for high-power GPUs and CPUs.

There is no single liquid-cooling design. In a direct-to-chip system, cold plates contact selected components while fans still cool the rest. In a rear-door heat exchanger, a radiator-like door removes heat from the exhaust air before it enters the room. In immersion cooling, components or complete servers sit in a nonconductive fluid. Each approach changes maintenance, monitoring, cost, and equipment compatibility.

Matching the building to the workload

The most important engineering question is not air or liquid. It is whether the cooling system matches the workload and the building. A facility may have plenty of electrical capacity but lack the piping, heat exchangers, pumps, or floor layout needed for liquid cooling. Retrofitting can be expensive because the cooling loop must be reliable, leak-monitored, serviceable, and integrated with the building’s mechanical systems.

Water, reliability and software

Water use also matters. Evaporative cooling can be energy efficient in the right climate, but it consumes water. Closed-loop liquid systems can reduce direct water use, although the heat still has to be rejected somewhere. Operators should report the source of water, the seasonal profile, and the difference between water withdrawn and water consumed.

Reliability is the reason cooling deserves operational attention. A brief temperature excursion can throttle a server, reduce performance, or trigger a shutdown. A leak can damage equipment and create a safety event. Pumps, valves, sensors, heat exchangers, and control software all become part of the availability plan.

The software layer is important as well. Workload schedulers can spread heat across a cluster. Telemetry can identify fans or pumps that are operating outside expected ranges. Operators can use power caps to keep a rack within a thermal envelope. Model-serving systems can route requests to the most efficient available hardware rather than treating every accelerator as interchangeable.

Where cooling shapes siting

Cooling also affects siting. A data center in a cool climate may have more hours when outside air can assist cooling. A facility in a hot or water-stressed region may need a different design. Heat reuse can be attractive in some locations, but it requires a nearby customer such as a district heating system, greenhouse, or industrial process.

What to ask a provider

For smaller businesses, these issues are mostly hidden inside the cloud service. They still matter when evaluating providers. Ask where the workload runs, whether the provider publishes power or water metrics, and how it handles hardware generations with different thermal requirements. A low-cost compute option may use older hardware with a different efficiency profile.

AI cooling is not just a mechanical problem. It is a coordination problem involving chips, racks, software, buildings, utilities, and operations staff. The best design removes heat at the point where it is created, uses power and water responsibly, and gives operators enough visibility to prevent small failures from becoming outages.

What BOSSTOX is watching

• Retrofit costs for adding liquid loops to existing halls.
• Disclosure of water withdrawn versus water consumed.
• Heat-reuse projects that find a real, nearby customer.