Direct Liquid Cooling vs Immersion Cooling: Choosing the Right Solution

For data centers grappling with extreme heat loads, liquid cooling has become an attractive path to higher performance and efficiency. 

Two leading liquid-based approaches are direct liquid cooling – often called direct-to-chip or cold plate cooling – and immersion cooling. 

Each method removes heat in a very different way. Determining whether direct liquid cooling or immersion is the better fit depends on your facility’s requirements, existing infrastructure, and long-term data center cooling strategy. 

We talk about how the two technologies work, highlight their strengths and challenges, and offer guidance on aligning each solution with the right operational needs.

How Direct-to-Chip (Cold Plate) Cooling Works

Direct liquid cooling uses sealed fluid pathways to carry coolant in and out of servers, targeting individual components. A coolant (often water) is pumped through cold plates attached to processors and other hot spots, absorbing heat at the source. 

The heated coolant is then circulated out of the chassis to a cooling distribution unit (CDU), where it transfers the heat to a secondary facility cooling loop via a heat exchanger. Direct-to-chip cooling can be implemented on standard server hardware that has been outfitted with liquid-cooled cold plate assemblies. Importantly, the liquid remains confined within the cooling loop and does not come into contact with any electronics beyond the sealed surfaces of the cold plates.

How Immersion Cooling Works

In an immersion cooling system, entire servers or components are submerged in a bath of specialized non-conductive fluid (such as a dielectric coolant). 

The fluid is in direct contact with all heat-generating surfaces, removing heat through either natural or forced convection. 

In one common design, servers are placed horizontally in a tank filled with a single-phase dielectric fluid. As the servers operate, hot components like CPUs heat the fluid around them, which rises and is cooled through a heat exchanger, then recirculated. 

Other designs use a two-phase approach where the fluid actually boils into vapour on hot surfaces, then condenses on a heat exchanger coil back to liquid form – a very efficient heat transfer method. 

Immersion cooling often eliminates the need for server fans entirely and can enable extremely high power densities, since the fluid directly contacts even small components.

Benefits and Suitability of Direct Liquid Cooling

Direct-to-chip cooling is valued for its efficiency, modularity, and relative ease of integration. It leverages a growing ecosystem – many major server manufacturers now offer liquid-cooled models or kits – and can often be retrofitted into existing data center spaces that have adequate facility water supply lines. Key advantages include:

• Familiar form factor: Liquid-cooled servers look and behave similarly to air-cooled ones, with standard rack mounting and service access. This makes it easier for data center operators to adopt with minimal retraining or changes to maintenance processes.
• Incremental deployment: Operators can start with liquid cooling in a handful of racks where it’s needed most (e.g., an AI cluster) without reworking entire rooms. Both dedicated and hybrid cooling environments (mixing air and liquid-cooled equipment) are possible.
• High efficiency and performance: By lowering component temperatures, direct liquid cooling prevents thermal throttling and cuts fan and chiller energy use, yielding large efficiency gains. It is very well-suited for loads up to and beyond 50 kW per rack while preserving conventional rack layouts and hardware choices.
• Maintenance isolation: Because liquid stays within sealed loops, electrical components remain dry and can be serviced individually by standard IT staff, often while the cooling system continues to run for other servers in the rack.

However, direct liquid cooling introduces new infrastructure requirements, like coolant piping and CDUs, and necessitates careful leak prevention and detection measures. There is also some additional upfront cost and design complexity in equipping servers with cold plates and connectors.

Benefits and Suitability of Immersion Cooling

Immersion cooling’s standout trait is its exceptional heat removal capacity. 

By bathing entire servers in liquid with vastly higher thermal conductivity than air, immersion can handle power densities far beyond even direct-to-chip cooling – on the order of 100 kW+ per rack in some designs. It virtually eliminates dependency on air distribution and server-level fans, leading to quieter operation and potentially lower component failure rates (since fans are a common point of failure). 

Immersion systems can also condense a large amount of compute into a smaller footprint, which is valuable when data center space is at a premium.

That being said, immersion cooling requires specialized equipment and significant changes to operations. Technicians must learn to handle and service submerged hardware, possibly dealing with dripping fluid and re-sealing tanks after maintenance. There may be constraints on the types of hardware that can be immersed (not all server components are tested or guaranteed for use in dielectric fluids). 

Additionally, while immersion cooling shows promise, its ecosystem (available products, standards, and support) is less mature than direct liquid cooling. Some operators also express concerns about how scaling an immersion environment might complicate workflow and hardware replacement.

Key Factors for Choosing the Right Approach

So, when deciding between direct liquid cooling and immersion cooling, as data center leaders, you should consider these factors:

• Target power density: If the primary need is to support extreme “beyond-air” density for a few critical racks (such as in supercomputing or crypto-mining operations), immersion cooling might offer the maximum headroom. 

For moderately high densities (e.g., up to 50–100 kW per rack) paired with broad hardware compatibility, direct-to-chip cooling can often meet requirements with less disruption.

• Integration with existing systems: Facilities that already have or can easily add a water infrastructure (pipes, pump room, etc.) may find direct liquid cooling fairly straightforward to implement. 

By contrast, immersion often calls for a more radical redesign of server layouts and backup systems (since submerged servers don’t use traditional racks).

• Operational and service model: Direct-to-chip is generally closer to “business as usual” in terms of server maintenance – racks remain accessible and individual servers can be removed or replaced without shutting down the whole system. Immersion demands a shift in how servers are handled and possibly longer maintenance cycles (e.g., draining fluids or cleaning components when needing service).
• State of the tech ecosystem: The marketplace for direct liquid cooled equipment is well established, with multiple vendors providing cold plate solutions and CDUs that have been tested in a variety of production environments. 

Immersion is evolving quickly but is still adopted in more limited scenarios, often requiring closer collaboration with vendors to ensure compatibility and support.

Aligning Cooling to Business Needs

Both direct liquid cooling and immersion cooling represent effective ways to break through the thermal limits of air cooling. 

Direct-to-chip cooling tends to be the first choice for many organizations because it offers a balance of high performance, efficient heat removal, and compatibility with existing data center operations. 

Immersion cooling can yield even higher density and efficiency gains, but it asks for a larger shift in how infrastructure is deployed and managed. 

The optimal solution depends on the specific goals: whether the priority is to incrementally boost performance while leveraging familiar architectures (favouring direct-to-chip) or to maximize density and efficiency in a specialized environment (favouring immersion). 

In practice, some operators even combine these approaches – keeping immersion for the absolute highest-density needs and using direct liquid cooling for mainstream high-performance clusters. By carefully evaluating these cooling architectures against operational constraints and future needs, data center teams can choose the approach that best aligns with their performance ambitions and infrastructure strategy.

Check out how CoolIT’s direct liquid cooling products (passive coldplates, rack manifolds, and industry-leading Coolant Distribution Units) scale from a single rack to hyperscale AI clusters, with or without facility water.

Talk to a CoolIT cooling specialist and get a solution matched to your density, footprint, and infrastructure.