Artificial intelligence (AI) is reshaping the data center industry at a pace unseen before. GPU-driven AI workloads now push rack densities beyond the limits of traditional air cooling. With their increased demands, the heat output per rack far outpaces the cooling efficiency of air-based systems. As a result, the industry is accelerating towards hybrid and fully liquid-cooled facilities.
Among liquid cooling methods, direct-to-chip (D2C) stands out as the most widely adopted, delivering coolant directly to processors via cold plates. This method enables operators to extract heat precisely where it’s generated while retaining the flexibility to maintain some air-cooled equipment. The transition doesn’t always require a greenfield build; retrofitting existing facilities is increasingly the preferred path for operators seeking faster deployment and reduced capital expenditure.
The case for retrofitting
Retrofitting offers clear financial, operational, and sustainability benefits. In many cases, a retrofit can cut time-to-market dramatically compared to new construction, allowing operators to meet AI demand today rather than waiting years for new builds. It can also reduce capital expenditure by around 20 to 40 percent compared to new builds. That combination of lower cost and faster delivery is a compelling reason for many developers under pressure from customers demanding immediate GPU capacity.
Environmental impact is another driver. New builds carry significant embodied carbon from concrete, steel, and large-scale mechanical systems. Retrofitting, by contrast, can cut embodied carbon by up to 50 percent in some cases. For operators under ESG pressure, retrofit strategies provide a faster, greener route to meeting AI requirements.
Grid availability adds another layer of complexity. In many global markets, grid connections are increasingly capped or delayed, making it harder to power new greenfield facilities. Retrofitting existing sites avoids lengthy grid approval cycles and helps operators get to market faster with the capacity they already have. The ability to leverage an existing grid connection is becoming just as important as cooling technology in shaping retrofit strategies.
The challenges for retrofitting
Retrofitting is not without obstacles. Structural limitations of legacy structural ceilings, raised floor congestion, and risks of downtime all present challenges. Some facilities were never designed to support the static loads of manifolds or the weight of additional cabling for GPU clusters. In older halls, underfloor space may already be congested with decades of legacy cabling and air-handling infrastructure. Live environments bring further risk: integrating liquid pathways can introduce leak concerns if not managed with care.
In some cases, operators have found it more cost-effective to demolish and rebuild rather than navigate these constraints. However, these risks can be mitigated through factory-tested manifold solutions, phased installation strategies, and integrated leak detection systems that reduce disruption.
Strategies for retrofitting liquid cooling
At the heart of any liquid cooling retrofit is the manifold. Playing the key role in the technology cooling system, it acts as the distribution hub between the cooling distribution unit (CDU) and the racks. Delivered as pairs, for supply and return, Manifolds ensure coolant is delivered precisely where it’s needed and returned efficiently for re-cooling. In retrofit projects, the design and placement of the manifold can make or break the success of the installation.
Manifolds must integrate seamlessly into existing white space, whether mounted overhead, underfloor, or on independent frames. They also carry additional static weight once filled with fluid, making structural support a key consideration during retrofits.
For facilities early in their retrofit journey, or those undergoing partial upgrades, structural ceiling capacity is the foundation for long-term liquid cooling success. Modern manifolds, when filled with fluid, impose substantial static loads, and this is compounded by added cabling from GPU clusters. Legacy structural ceiling grids are rarely sufficient. Upgrading to stronger structural ceiling solutions ensures that manifolds and containment systems can be safely supported while also simplifying the routing of both air and liquid pathways. For contractors and installers, structural ceiling-mounted systems also free up valuable floor space, reducing complexity during retrofit projects.
Where structural ceiling upgrades are impractical, floor-mounted goal post frames provide a reliable alternative. These structures are modular, relocatable, and ideal for facilities rolling out liquid cooling in phases. By decoupling manifold loads from structural ceiling structures, operators can retrofit without costly structural overhauls. This approach also allows data centers to deploy GPU clusters strategically, targeting high-demand zones first while retaining flexibility for future repositioning or expansion.
In facilities with raised floors, underfloor manifolds offer another approach. This method minimizes overhead congestion and simplifies installation in live environments. It requires careful planning to balance airflow and static pressure, but when implemented effectively, underfloor routing provides a clean, integrated solution. Operators considering underfloor manifolds should carefully evaluate airflow trade-offs, ensuring that the introduction of liquid pathways does not compromise cooling efficiency elsewhere in the hall.
Beyond Cooling
Retrofitting is no longer just about keeping pace with technology; it is also about maintaining competitiveness. Operators who retrofit can bring new AI-ready capacity online significantly faster than those pursuing greenfield construction. In a market where demand is growing exponentially, speed to market can often make the difference between winning and losing major contracts.
At the same time, embodied carbon is now a board-level concern. Investors, regulators, and customers are increasingly scrutinising Scope 3 emissions, and new builds come with heavy upfront carbon costs. Retrofitting allows operators to extend the lifespan of existing assets while avoiding the carbon footprint of demolition and new construction.
Grid connection constraints only heighten the urgency. In regions where new connections may take years to approve, retrofitting allows operators to extract maximum value from existing infrastructure. In effect, retrofit strategies are no longer a stopgap; they are the most practical way to scale AI-ready capacity under current market conditions.
Retrofitting liquid cooling isn’t a compromise; it’s an opportunity to maximize the value of existing facilities while preparing for the AI-driven future. Operators gain speed to market, reduce capital costs, and cut embodied carbon, all while overcoming grid constraints. The most successful operators will be those who adopt a modular, forward-thinking approach that balances immediate needs with long-term flexibility.
