Design-Build & Liquid Cooling: Building the Next-Gen Data Center

AI training clusters now ship with chips that draw more than 1,000 watts each, pushing rack densities past 50–80 kW territory where air alone cannot keep equipment within safe operating temperatures. The industry’s response is liquid cooling. But the technology choice is only half the equation. The other half is how that liquid-ready facility actually gets designed and built.

A design-build liquid cooling data center strategy unites engineering and construction under one accountable team, which matters enormously when a project depends on tightly coordinated piping, electrical, and automation work. This article breaks down why liquid cooling has become non-negotiable for high-density computing, why design-build has become the delivery method of choice for these projects specifically, and how owners should evaluate partners, contracts, and risk before breaking ground.

Why Air Cooling Is Hitting a Wall

Air cooling has been the default approach for data centers for decades, and for good reason it’s simple, well-understood, and inexpensive to install. But its physical limits are now showing up in real operational consequences.

The core problem is heat density. Modern AI and HPC servers concentrate far more compute power into far less rack space than the servers air-cooling systems were originally engineered around. A few key dynamics are driving this shift:

• GPU thermal design power (TDP) is rising every generation.Flagship AI accelerators now exceed 700–1,000W per chip, multiplied across racks holding dozens of them.
• Rack densities are climbing past air’s practical ceiling.Many AI clusters now operate at 50–80+ kW per rack, well beyond what hot-aisle/cold-aisle containment and CRAH units can reliably remove.
• Uptime requirements leave no margin for thermal failure.Mission-critical facilities target 99.99% availability; insufficient cooling capacity directly threatens that target.
• Energy costs make inefficient cooling unaffordable at scale.Air-handling systems consume significant power just to move air, on top of the cooling load itself.

When air-cooling infrastructure can’t keep pace, the downstream effects are concrete: throttled chip performance, shortened hardware lifespan, capped rack density, and reduced service capacity for customers. These aren’t future risks they are current constraints for any operator trying to deploy next-generation AI infrastructure today.

Liquid Cooling 101: The Four Main Approaches

Liquid cooling is a data center cooling method that uses liquid instead of (or alongside) air to remove heat from IT equipment, transferring thermal energy far more efficiently than air can. Liquid can move heat away from components dramatically faster than air, which is why it scales to densities air simply cannot reach.

There are four primary architectures in use today, each suited to different density levels, retrofit constraints, and budgets:

• Direct-to-chip cooling:routes fluid through a metal cold plate positioned directly over the highest-heat components, pulling thermal energy into the coolant before it ever reaches the surrounding air. It requires a CDU to separate the closed-loop coolant circuit from facility water, since the small channels in modern cold plates demand high fluid purity.
• Immersion cooling: submerges entire server boards in a dielectric fluid that doesn’t conduct electricity, cooling every component simultaneously rather than targeting specific chips. It tends to deliver the most dramatic density gains but is the hardest method to retrofit into an existing air-cooled facility, since it typically requires entirely different rack and floor infrastructure.
• Rear-door and precision cooling: sit closer to traditional air-cooling retrofits, using liquid-cooled heat exchangers mounted on the back of racks or targeted spray/jet systems aimed at the hottest components. These approaches are generally the easiest and least disruptive to add to an existing facility.
• Hybrid systems: blend liquid and air cooling, letting operators address the hottest racks with liquid while leaving lower-density equipment on conventional air a pragmatic way to phase a transition without ripping out an entire facility at once.

The Business Case: PUE, WUE, and Total Cost of Ownership

Beyond raw cooling capacity, liquid cooling delivers measurable efficiency and sustainability gains that matter directly to the bottom line.

• Power Usage Effectiveness (PUE): the ratio of total facility energy to IT equipment energy improves substantially with liquid cooling. Facilities can achieve PUE values approaching 1.02–1.2, compared to an industry average closer to 1.5–1.6 for conventional air-cooled sites. Some analyses put the overall PUE improvement from switching to liquid cooling at up to 45 percent.
• Water Usage Effectiveness (WUE):also benefits, since many liquid cooling systems run as closed loops rather than relying on evaporative cooling towers. This design can cut water consumption by as much as 90 percent compared to traditional air-cooling approaches.

Key takeaways on the business case:

• Liquid cooling can reduce PUE to near-1.0 levels, versus 1.5+ for legacy air systems
• Closed-loop designs cut water usage dramatically, supporting sustainability commitments
• Higher rack density means more compute per square foot of expensive data center real estate
• Some systems support outlet temperatures up to 60°C, enabling waste-heat reuse for district heating or other applications
• Reduced mechanical strain on cooling hardware can lower maintenance costs and downtime over the facility’s life

These gains explain why liquid cooling has moved from “future trend” to active deployment strategy. The remaining question for most owners isn’t whether to adopt liquid cooling it’s how to build it without blowing the schedule or budget.

Design-Build vs. Design-Bid-Build vs. CM-at-Risk

The delivery method an owner chooses determines how design and construction teams are structured, who holds risk, and how fast a project can move. Three models dominate data center construction today.

Design-Bid-Build (DBB)

• Contract system: adopting a split independent contract model, the owner signs a separate contract with the design unit and the construction unit, and the rights and responsibilities of the two parties are independent of each other and are not subordinate to each other.
• Construction schedule: The overall construction period is the longest, and the project follows the linear progression process of design, bidding, and construction, with each stage carried out in turn, and cannot be interspersed with parallel operations.
• Owner’s design control authority: the owner has the highest design control authority, which can dominate the whole process of program preparation, detail adjustment and drawing validation, and control the project design standard and landing effect in an all-round way.
• Risk of engineering change: The risk of project change is high, the design and construction phases are completely separated, and there are barriers to information connection, which is very likely to cause process conflicts, program inconsistencies and other problems.
• Attribution of responsibility: project authority and responsibility are scattered and fragmented, and the responsibilities of design and construction units are split up, so it is easy to have ambiguous authority and responsibility and shift the responsibility to each other when there are quality and progress problems.
• Applicable scenarios: Applicable to owners who have stringent requirements on design accuracy and program specifications, and need to firmly control the design details and engineering quality throughout the project.

Construction Manager at Risk (CMAR)

• Contract system: the contracts of all parties remain independent, and the synergistic linkage of all parties is stronger in the pre-project stage, which can be intervened and docked in advance, and effectively improve the problem of fragmentation in the traditional mode.
• Construction progress: the construction speed is moderate, and design and construction processes can be interspersed with each other, which can effectively compress the construction period compared with the traditional linear construction mode.
• Owner’s design control authority: The owner has high design control authority, and can deeply participate in the optimization of the program, drawing validation and other core aspects to control the project design quality.
• Engineering change risk: the project change risk is controllable and the overall risk is moderate. Relying on the mechanism of multi-party collaboration in the early stage, the hidden dangers can be detected in advance, which can greatly reduce the deviation of the interface between design and construction.
• Responsibility: The multi-party responsibility sharing mechanism is implemented, whereby the construction, design and construction parties perform their duties in collaboration, and jointly bear the responsibilities and risks during the construction of the project.
• Applicable Scenario: Applicable to owners who wish to lock the project cost in advance, take into account the construction progress and project risk control, and pursue the balance of comprehensive benefits.

Design-Build (DB)

• Contract system: adopting a single general contract model, the owner only docking the general contracting unit, unified contracting, unified responsibility, simple and clear contract structure, avoiding multi-party docking chaos.
• Construction progress: the overall construction efficiency is optimal, breaking the linear construction restrictions, realizing simultaneous interspersing of design and construction, and advancing in parallel, significantly compressing the total project construction period.
• Owner’s design control authority: the owner’s direct design control authority is relatively limited, and the general contractor coordinates and dominates the full set of design, optimization and landing work, and advances the whole process in an integrated manner.
• Engineering change risk: The overall change risk of the project is the lowest, relying on the integrated design and construction coordination to avoid professional disconnection, process conflict, connection deviation and other common problems from the source.
• Attribution of responsibility: project rights and responsibilities are highly centralized, clear and unique, with the general contractor assuming the main responsibility for the whole process, completely eliminating problems such as multiple shirking of responsibilities and blurring of rights and responsibilities.
• Applicable Scenario: Applicable to owners who prioritize schedule and pursue efficient project implementation, and at the same time hope to transfer construction risks centrally and simplify the project control process.

For owners, design-build’s single-point accountability typically means fewer change orders, faster dispute resolution, and greater cost and schedule certainty particularly valuable on fast-track or hyperscale projects. The trade-off is that owners give up some late-stage design flexibility and must define performance expectations clearly upfront, since one firm now controls both halves of the project.

Why Delivery Method Becomes Critical for Liquid Cooling Projects Specifically

Liquid cooling raises the coordination stakes well beyond a standard data center build, which is exactly where delivery method stops being a procurement detail and starts being a risk-management decision.

A liquid-cooled facility whether direct-to-chip, immersion, or hybrid requires tight integration across disciplines that traditionally operate in silos:

• Specialized process piping design: Optimize the piping specifications and laying direction for the coolant distribution system, effectively reduce the pressure loss of media transmission, and minimize the leakage potential points.
• Optimization of CDU layout and selection: Combined with the water and cooling needs of the site, weighed the centralized and rack-distributed deployment options, scientifically selected CDU equipment, and finalized the optimal layout.
• Cooperative layout for electrical safety: Coordinate the planning of electrical equipment layout, strictly guarantee the physical isolation of power distribution units, UPS systems and high-risk areas for fluid leakage, avoiding the electrical safety risks caused by liquid leakage.
• Intelligent self-control system configuration: Equipped with an automated monitoring and control system, the system can collect and monitor the dynamic refrigeration load, pipeline leakage status, coolant quality and other core operational data in real time, realizing intelligent operation and maintenance control.
• Structural and Floor Adaptation: Targeted optimization of the building structure and floor layout, adapting to the installation and layout needs of the converging pipeline, leak-proof tray, liquid storage tank immersion cooling system, and guaranteeing the adaptability of the equipment to the ground.

When these systems are designed by one team and built by another under separate contracts, misalignment between electrical load planning and cooling system design is a well-documented source of costly late-stage redesigns. A coolant distribution unit sized without close coordination with the electrical team, or piping routed without input from the construction trades that will actually install it, generates change orders precisely the kind of project these facilities cannot afford to absorb given uptime requirements and capital intensity.

Design-build closes that gap. Because the same entity controls both engineering and execution, cooling, electrical, and automation decisions get made together rather than handed off turning a liquid cooling concept into a built, commissioned system without the friction of multiple competing contracts.

Greenfield vs. Retrofit: Different Delivery Strategies

Not every liquid cooling project starts the same way, and delivery strategy should reflect that.

For Greenfield Projects:

1. In the planning stage of the project, before the layout of the site is finalized, the design and construction team will work together to ensure that the overall scheme is scientifically unified and smooth on the ground.
2. The overall planning is centered on the liquid cooling system, integrating the layout of the server room, structural load and piping direction, and abandoning the passive design thinking of traditional air-cooled post-reconstruction.
3. Based on the rack load development planning of the whole life cycle of the server room, it is not limited to the initial installation demand, and reasonably selects the cooling architecture programs such as direct chip cooling, submerged cooling, and hybrid cooling.
4. The automated monitoring and intelligent monitoring system is integrated into the design stage, realizing integrated supporting construction and avoiding the problems of retrofitting and inadequate adaptation at a later stage.

For Retrofit Projects

1. Comprehensively assessing the current situation and structural conditions of the site, we accurately screen the areas suitable for liquid-cooling retrofit, minimize structural changes, and compress the retrofit costs and construction difficulties.
2. In view of the complex pipelines and restricted space conditions, we flexibly adopt the rear door plate heat exchanger, hot and cold hybrid cooling and other adaptable solutions to fit the actual construction conditions of the site.
3. It adopts a phased iterative upgrade strategy, prioritizing the implementation of liquid-cooled transformation for high-density heat generating racks, and retaining the air-cooled operation mode for low-density equipment, so as to achieve a smooth transition and gradient upgrading.
4. Focusing on controlling the risk of pipeline leakage, most of the original pipelines and pipe fittings are not designed to meet the working condition tolerance of the liquid cooling system, so it is necessary to verify and optimize the transformation in advance, so as to avoid leakage potential hazards.

In both cases, a design-build partner with in-house MEP engineering, automation, and construction capabilities can move directly from concept to commissioned system without the handoff delays that multi-contract models introduce.

Risk Allocation and Contract Considerations

Choosing design-build doesn’t eliminate risk it concentrates and reallocates it, which means contract structure deserves careful attention before signing.

Key contract considerations for liquid cooling design-build projects:

• Performance specifications upfront.Because the design-builder controls both design and construction, owners should define clear performance expectations early particularly for cooling system capacity, redundancy standards, and commissioning requirements since there’s less opportunity for late-stage owner-driven design changes.
• Price adjustment clauses.Given procurement volatility in specialized cooling equipment (CDUs, dielectric fluids, custom piping), contracts should account for material pricing indices or tariff-driven cost increases.
• Incentive and liquidated damages structures.Pairing cost-efficiency incentives with liquidated damages for schedule or performance shortfalls helps align the design-builder’s interests with the owner’s.
• Transparency and audit rights.Owners should retain rights to review procurement assumptions and pricing, particularly on long-lead electrical and cooling equipment.
• Leak and commissioning protocols.Given that liquid cooling unlike air cooling can be difficult to service without planned downtime, contracts should specify leak detection, containment, and commissioning standards explicitly rather than relying on general construction-quality language.

Neither design-build nor any other delivery model eliminates the inherent risks of a liquid cooling project. What design-build does is centralize accountability, which can deliver greater price and schedule certainty provided the contract clearly assigns performance obligations from the outset.

Choosing the Right Partner Team

Liquid cooling design-build projects demand a broader skill set than a typical data center build, and most traditional data center contractors don’t have it all in-house.

A strong design-build partner for liquid cooling should offer:

• In-house MEP engineering and process design, including specialized piping for cooling liquids not subcontracted as an afterthought
• Mechanical expertise in fluid dynamics and heat transfer, beyond standard HVAC and air-system experience
• Automation and controls capabilitiesto manage dynamic cooling loads and integrate real-time monitoring
• Strategic sourcing and procurementrelationships for long-lead electrical and cooling equipment
• Self-perform construction laborfor critical trades, reducing dependency on third-party subcontractors for schedule-critical work
• A track record in mission-critical, high-density facilitiesnot just general industrial construction

Many traditional data center construction firms have deep HVAC and air-system experience but lack the specialized mechanical expertise that liquid cooling deployment demands. Vetting for that gap before selecting a partner is one of the most consequential decisions an owner will make on a liquid cooling project.

Conclusion

Liquid cooling has moved from emerging technology to operational necessity for any data center supporting AI or high-performance computing workloads. But the technology alone doesn’t determine project success delivery method does. Design-build’s single-point accountability directly addresses the coordination failures that most often derail liquid cooling projects: misaligned electrical and cooling design, late-stage change orders, and disjointed handoffs between engineering and construction teams.

For owners planning a greenfield liquid-ready facility or retrofitting an existing site, the strategic question isn’t simply “which cooling technology?” It’s “which delivery model and which partner can execute that technology without sacrificing speed, cost certainty, or reliability?” Owners evaluating their next data center project should start by vetting design-build partners specifically for liquid cooling mechanical expertise, not just general construction experience.