Daikin 60MW Cooling Design Connects NVIDIA AI Racks to the Whole Facility
Daikin Applied introduced its Daikin 60MW cooling design on November 10, connecting NVIDIA AI racks with facility cooling in a coordinated blueprint. Called D60-W, it addresses an integration problem: cooling equipment must work together before computing capacity can operate reliably. The announcement establishes a design framework, while equipment availability and project engineering remain separate milestones.
The reference architecture, an engineering blueprint describing how systems work together, targets facilities built around NVIDIA Vera Rubin NVL72 computing systems. Its significance reaches beyond the size of the proposed installation. Daikin is asking buyers to consider the complete heat-removal system earlier, rather than resolve connections between individually selected products later.
Schneider Electric already offers a Vera Rubin reference design, giving buyers another example of infrastructure suppliers packaging engineering around NVIDIA hardware. The central question is how much coordination these frameworks remove from an actual project. For developers, equipment suppliers, and operators, that shifts attention toward operating conditions, delivery commitments, and responsibility across the facility.
What the Daikin 60MW Cooling Design Actually Announces
D60-W packages a cooling architecture around a specified computing load, giving project teams a shared starting point for engineering.
Daikin’s cooling reference architecture describes support for a 60 MW information technology load across six deployable units. Its U.S. announcement carries a November 10 dateline; Daikin’s global release appeared November 11. Those dates describe the same launch, rather than separate product introductions.
The capacity figure needs careful interpretation. It describes the computing load the proposed cooling arrangement supports. It does not announce electricity generation, identify an operating customer facility, or establish that a construction contract has closed.
That distinction matters because infrastructure announcements often compress several milestones into a headline. Publishing a design, selecting equipment, securing a site, and accepting an operating installation each answer different questions. D60-W belongs initially to the design stage, where engineers establish requirements and evaluate how components will interact.
An AI factory, a data center organized around AI computing workloads, still needs connections between its computing hardware and surrounding building systems. Those connections determine whether an equipment specification fits the conditions elsewhere in the facility. A useful blueprint makes these dependencies visible before procurement decisions become difficult to reverse.
For example, consider a developer evaluating an AI expansion while its mechanical engineering team reviews the cooling plant. A rack supplier describes the computing equipment’s requirements, while another supplier proposes equipment outside the building. Without shared assumptions, each proposal can appear suitable while leaving the connecting system unresolved.
A reference design can give those teams a common basis for discussion. They can identify which temperatures, equipment arrangements, and operating assumptions require adjustment for the proposed site. That is a practical benefit even before a customer commits to the complete architecture.
However, a common starting point does not remove the need for detailed engineering. Site constraints, customer requirements, maintenance access, and equipment substitutions still affect the finished installation. The architecture helps organize those decisions; its publication does not settle them.
The most useful reading of this launch therefore concerns scope. Daikin is presenting cooling as a connected facility system, with equipment choices considered together. Buyers should assess whether that approach clarifies responsibilities and reduces unresolved interfaces in their own projects.
That assessment also explains why the announcement matters beyond Daikin’s immediate customers. It creates another concrete proposal against which engineering teams can compare their current designs. The value comes from the quality of that comparison, rather than the headline capacity alone.
Why Integration Puts Developers and Suppliers Under Pressure
The immediate pressure falls on teams responsible for turning equipment selections into an operating facility.
A cooling component can meet its individual specification while the broader installation still requires substantial coordination. The buyer must reconcile the assumptions behind each selection. When those assumptions remain implicit, the project carries uncertainty into later engineering and commissioning, the process of checking an installed system before operational acceptance.
Daikin’s proposal gives developers a reason to ask more specific questions during supplier selection. Which interfaces does the supplier define? Which operating conditions does the design assume? Who resolves a mismatch when the computing equipment and cooling plant require different arrangements?
These questions affect procurement because responsibility can become fragmented across contracts. A supplier may deliver exactly what its agreement requires while another party remains responsible for connecting that equipment. The buyer then needs a clear account of the work between those boundaries.
A coordinated design has commercial value when it makes that work explicit. It can expose decisions that otherwise emerge during construction or testing. Whether D60-W delivers that benefit on a particular project depends on the documentation, engineering support, and contractual scope available to the customer.
NVIDIA’s AI factory platform provides the broader context. The company describes DSX as bringing design, simulation, and operations together across computing and facility infrastructure. That approach encourages suppliers to explain how their products participate in a larger operating system.
NVIDIA’s description is a platform strategy, not independent evidence that every associated project achieves faster deployment. Its relevance here is organizational. Computing requirements increasingly shape discussions that involve mechanical engineers, electrical engineers, software teams, equipment manufacturers, and facility operators.
Developers must translate those discussions into a schedule they can defend. An equipment delivery date has limited usefulness if supporting systems remain unspecified. Equally, a completed building cannot provide the intended service until its installed computing equipment has suitable operating conditions.
Suppliers face a related challenge. They must explain how their products fit into the project’s assumptions, including behavior during maintenance and changing operating conditions. A strong individual specification remains useful, but buyers also need evidence about the connections surrounding that product.
For enterprise customers, these details influence how confidently they can plan access to new computing capacity. They should distinguish proposed capacity from capacity available under an operating service agreement. A cooling announcement adds information about infrastructure planning without establishing when a customer can use the resulting compute.
The Daikin 60MW cooling design makes this coordination visible through a named architecture. Its practical influence will depend on whether project teams adopt its assumptions and retain them through engineering. That creates a clearer basis for scrutiny than a general promise of simpler deployment.
How D60-W Coordinates Different Cooling Requirements
The design’s central mechanism is coordination between cooling paths with different temperature requirements.
Liquid cooling transfers heat through a circulating fluid, while air cooling transfers heat through moving air. Daikin assigns different roles to these paths within D60-W. The engineering challenge is to connect them to a plant that responds appropriately as conditions change.
Daikin’s D60-W design conditions allocate 92% of the load to liquid cooling and 8% to air cooling. The design uses separate temperature loops for these requirements. Its published reference weather conditions use Ashburn, Virginia, making geography an explicit part of the engineering basis.
A coolant distribution unit, or CDU, manages the connection between facility cooling and the liquid circuit serving computing equipment. Its role makes it an important interface rather than an isolated capacity figure. Engineers must evaluate what happens on both sides of that connection.
A chiller uses mechanical refrigeration to cool fluid. A dry cooler transfers heat from circulating fluid to outdoor air. These devices perform different jobs, so their operating relationship matters as much as their individual ratings.
Daikin describes operating modes that vary the contribution of outdoor cooling and mechanical refrigeration. The general engineering logic is straightforward: outside conditions affect how much cooling work each part of the plant must perform. The finished system must manage those changes while maintaining suitable conditions for the computing equipment.
This is why temperature assumptions deserve close attention. A component selected for one operating condition may require a different configuration elsewhere. Reviewing the full path helps engineers identify where equipment choices depend on assumptions made by another team.
The smaller air-cooling allocation also deserves scrutiny. Its size does not make its assigned heat optional. A buyer evaluating the architecture needs to understand how that cooling path operates, how it receives support, and how maintenance affects it.
Separate temperature loops introduce another coordination requirement. The facility must serve different needs without treating every load as if it requires identical conditions. Buyers should assess how the proposed controls manage those relationships during normal operation and during transitions.
Controls are the instructions and monitoring systems that direct equipment behavior. They determine when equipment changes operating mode and how the plant responds to measurements. Their implementation belongs in the engineering discussion alongside physical equipment.
The supporting keyword “Daikin D60-W explained” points toward this mechanism, rather than a single new cooling component. The design proposes an arrangement of equipment and operating rules. Its significance depends on how clearly those rules translate into an installation.
For a project team, the useful question concerns traceability. Can engineers follow each requirement from the computing rack through the facility equipment to outdoor heat removal? A design that supports that review helps buyers identify the assumptions they must preserve, modify, or test.
Schneider Electric Shows Why Capacity Is an Incomplete Comparison
Competing reference designs should be compared by their boundaries and assumptions before buyers compare headline megawatts.
Schneider Electric’s Vera Rubin design provides a relevant comparison. Its Reference Design 113 describes a 10.4–11.4 MW data hall using Vera Rubin NVL72 racks. Published September 1, 2026, it covers facility power, cooling, computing space, and lifecycle software.
That documented scope differs from a cooling architecture organized around a larger total computing load. The numerical difference alone does not establish which supplier offers the better solution. Buyers first need to determine what each number includes and what project configuration it represents.
A comparison becomes misleading when one figure describes an entire arrangement and another describes a component or smaller deployment block. Both figures can be accurate while answering different questions. Treating them as directly interchangeable hides the engineering detail buyers need.
For the same reason, a CDU rating cannot be read as equivalent to supported facility capacity. A component participates in a wider system whose operating limits require separate review. Adding component ratings together does not automatically establish the capacity available under every operating condition.
A useful Daikin versus Schneider comparison therefore begins with the buyer’s project. Does the proposal address the required computing configuration? Does it specify the relevant facility interfaces? Does its operating basis match the site and the customer’s planned expansion?
Schneider’s inclusion of electrical and software areas also illustrates the importance of scope. A buyer may need a broad facility framework, detailed cooling engineering, or a combination of supplier documents. The right comparison depends on which decisions remain unresolved.
Neither approach eliminates the buyer’s responsibility to examine service and support. Equipment maintenance requires access, procedures, spare parts, and a clear division of work. Those arrangements affect whether a proposed architecture fits an organization’s operating capabilities.
Supplier concentration deserves similar attention. A coordinated offering can simplify communication when responsibility follows the same boundaries as the supplied system. Buyers should still examine what happens when equipment changes, a component becomes unavailable, or another supplier joins the project.
This is an assessment of procurement structure, rather than a claim that either company imposes restrictive terms. The public design descriptions do not establish the terms of an individual customer contract. Those details require direct commercial and engineering review.
The competitive significance is therefore broader than a contest over capacity. Suppliers are giving buyers increasingly specific starting points for facility design. Buyers gain the most when they use those documents to expose differences in assumptions, rather than select a winner from the headline figures.
Availability, Water Use, and Reliability Need Separate Evidence
D60-W’s engineering proposal is clearer than the evidence currently available about a completed customer deployment.
Daikin states that its CDU and data center dry cooler remain under development, with availability targeted for the second half of 2027. That disclosure is material for projects entering procurement now. A planned product requires a different scheduling assessment from equipment with confirmed delivery commitments.
The timeline does not invalidate the architecture. Developers often evaluate designs before every proposed component reaches commercial availability. It does mean buyers must separate the design review from decisions that require firm product specifications, delivery dates, and support arrangements.
A near-term project should ask what equipment configuration the supplier proposes for its schedule. Any substitution needs review against the wider design assumptions. Buyers should avoid assuming that a different component preserves every operating condition simply because its nominal capacity looks similar.
Water claims require another boundary check. Daikin says dry coolers replace evaporative cooling towers in this architecture, eliminating cooling-tower make-up water. That is a specific statement about a cooling process, rather than a measurement of every water use associated with a facility.
A water-cooled system can circulate water without continuously consuming it through evaporation. However, readers should distinguish circulating fluid from water consumption. They should also distinguish a design feature from measured consumption during an actual operating period.
Efficiency claims need comparable discipline. A proposed operating mode describes how the equipment is intended to work under specified conditions. It does not establish annual electricity consumption at every location or under every workload.
Weather, computing demand, and operating choices all belong in that assessment. A buyer needs assumptions that match its intended facility before drawing conclusions about annual performance. Without that context, a general efficiency statement offers limited help in comparing projects.
Reliability is equally difficult to establish from architecture alone. The presence of backup equipment does not describe every failure scenario. Buyers need to understand which functions remain available when equipment, controls, power, or communication paths experience problems.
Those questions should lead to documented acceptance tests. An acceptance test checks whether the delivered installation meets agreed requirements under defined conditions. Its value depends on the scenarios, measurements, and pass criteria established before the test begins.
Testing should also connect equipment behavior to computing requirements. A cooling plant can recover while the computing service experiences a different recovery sequence. The customer needs to know what the complete installation must deliver, rather than rely on an isolated equipment response.
None of these questions establishes that D60-W will fall short. They identify the evidence required to evaluate its promises fairly. The distinction protects buyers from premature conclusions while giving the supplier clear opportunities to substantiate the design.
What Buyers Should Watch Through Early 2027
The next useful evidence concerns product readiness, customer engineering commitments, and transparent operating validation.
The first signal is greater detail about the equipment still under development. Over the coming months, buyers should watch for specifications, qualification information, ordering guidance, and updates to the stated availability window. These disclosures would make it easier to connect the architecture with a procurement schedule.
A product announcement alone would leave some questions open. Buyers would still need to confirm the configuration proposed for their facility and the commitments available for their delivery window. More precise documentation would strengthen confidence; an unexplained schedule change would weaken it.
This signal matters even for projects that will start later. Equipment decisions influence engineering work before installation begins. Earlier clarity allows teams to evaluate dependencies without assuming that a future product will match every preliminary requirement.
The second signal is a disclosed customer project with a clearly described stage. A named engineering selection would show something different from a completed installation. Buyers should look for enough detail to understand the relationship between the customer, the architecture, and the proposed equipment.
Useful disclosures would explain whether a project has entered design, procurement, construction, or operational acceptance. They would also identify the scope being supplied. That information would strengthen the case for adoption without requiring readers to infer deployment from an announcement.
A project cancellation or substantial redesign would deserve context rather than an automatic negative verdict. Infrastructure plans change for many reasons. The relevant question would be whether the change exposed a limitation in the architecture or reflected another project constraint.
The third signal is validation with enough operating context to support comparison. Readers should watch for descriptions of test conditions, operating transitions, maintenance scenarios, and measurement boundaries. Evidence becomes more useful when a buyer can judge whether the conditions resemble its own requirements.
Independent customer observations would add value, especially when they explain both successful behavior and remaining engineering work. Supplier results can also be useful when methods and limitations are clear. Neither should be reduced to an isolated efficiency figure without its operating context.
These signals would strengthen or weaken the same proposition: coordinated cooling design can make project integration more predictable. They would not automatically establish superiority over every competing design. Each new disclosure should be assessed against the specific claim it supports.
For engineering teams following the story, keeping the evidence connected is part of the work. Requirements, design revisions, supplier statements, and test results can diverge as a project develops. A searchable knowledge base can help teams retrieve the source behind a technical decision.
Treat the Daikin 60MW cooling design as a concrete proposal to investigate against your project’s requirements. Start with its boundaries, then examine the equipment schedule, site assumptions, and operating responsibilities. Ask prospective suppliers to explain where their proposal follows the reference architecture and where it departs. Keep those answers alongside the drawings and acceptance criteria your team will actually use. As product and customer disclosures emerge, revisit the assumptions that affected procurement or scheduling. That approach gives the announcement practical value today while leaving room for later evidence. Before committing to an AI facility plan, ask which cooling requirements have documented answers and which still depend on future engineering.



