Cooling

Castrol launched Castrol CORE on September 29 in Singapore, a new service offering that integrates coolant products with design, system integration, testing, and maintenance for data centre liquid cooling. The service targets high-density AI infrastructure and is delivered through a network of hardware and service partners, with a dedicated team that includes specialists with backgrounds at firms such as Schneider, Vertiv, Eaton, IBM, and ABB. Castrol established its data centre and thermal management division three years ago in response to rising demand for cooling high-density servers. India is identified as a strategic growth market for Castrol CORE, with local stock available to support deployments.

According to an IndexBox report, the global liquid cooling manifolds market is forecast to grow at a compound annual growth rate of 12 to 18 percent from 2026 to 2035, driven by rising rack power densities in data centers, high-performance computing, and semiconductor manufacturing. Data centers are the largest end-use segment, accounting for an estimated 45 percent of global demand, with rack power densities exceeding 30 kW making traditional air cooling inadequate. Asia-Pacific and North America together represent 70 to 75 percent of global consumption, while the Americas source 60 to 70 percent of manifold requirements through imports. Standard-grade manifolds trade in a USD 80 to 200 unit price range, while premium specifications command two to three times that figure, and replacement cycles of three to five years generate recurring aftermarket revenue. Named market participants include Vertiv Holdings, Schneider Electric, CoolIT Systems, Asetek, and Boyd Corporation.

LG Electronics has joined the NVIDIA Partner Network as a Power and Cooling Solution Preferred Partner, gaining qualification for its 2.6-megawatt Coolant Distribution Unit as an NVIDIA DSX Ready CDU, which LG says is the largest cooling capacity among current DSX Ready CDU partners. The company has also completed NVIDIA infrastructure qualification for its 600 kW and 1 MW CDUs, covering cooling performance, failover capabilities, and reliability. The CDUs use direct-to-chip liquid cooling and are designed to work alongside existing air-cooling infrastructure. LG plans to display the qualified units at Data Centre World Asia 2026 on September 29 and 30 in Singapore, and aims to complete NVIDIA qualification for a 4.0 MW CDU by the end of 2026.

Trane Technologies, working with Eaton Corporation and Danfoss, has completed a laboratory demonstration of an 800-volt direct current cooling architecture, modifying an existing high-efficiency chiller to operate from an 800 VDC feed and delivering more than 1,000 tons, or 3.5MW, of cooling capacity. The company says the approach could yield up to a 2% improvement in system efficiency by reducing power conversion losses in compressor drives, pumps, and fans. In a 200MW data center, Trane estimates that efficiency gain could free up to 1.8MW of additional compute capacity, enough to support up to fifteen 120kW racks or up to three 600kW racks.

LG Electronics has been listed as a Power and Cooling partner in the Nvidia Partner Network, qualifying its 600-kilowatt and 1-megawatt coolant distribution units under Nvidia's standard requirements, and its 2.5-megawatt CDU as an Nvidia DSX Ready CDU under AI factory specifications. The designation gives LG greater access to customers building AI data centers on Nvidia technology, including earlier engagement at the design and planning stage. Lee Jae-sung, president of LG Electronics' ES Company, said the listing demonstrates that the company's thermal-management technology meets the requirements for next-generation AI data center infrastructure. LG entered the data center cooling market in 2011 and has since built a portfolio it describes as spanning chip-level cold plates, coolant distribution units, and chillers.

LG Electronics made its debut at Data Center World Asia 2026 in Singapore, unveiling a Hydronic Modular system that preassembles pumps, valves, emergency buffer tanks, and piping inside ISO-specification containers to reduce on-site installation work and allow cooling capacity to be added in phases as server deployments grow. The company also introduced 1MW and 2.6MW coolant distribution units for direct-to-chip liquid cooling of GPUs and CPUs, expanding a lineup that began with a 600kW model and has received quality standard approval from Nvidia. Additional products on display included an air-cooled oil-free inverter turbo chiller using magnetic bearings, a fan wall unit for dense server rooms, and immersion cooling technology. LG Electronics said it plans to combine its HVAC capabilities with LG Energy Solution's battery products and LG CNS's data center design and operations services to address AI data center design, construction, and expansion.

Phononic's Thermal Control System uses solid-state thermoelectric coolers to manage heat at individual components, including GPUs, high-bandwidth memory, switch ASICs, and optical engines, responding to thermal changes in milliseconds, which the company says is up to 600 times faster than conventional liquid cooling loops. By enabling supply water temperatures of up to 50 degrees Celsius, the system reduces demand for chilled-water production, with modeled savings of up to 14% in total data center energy consumption and approximately one million gallons of cooling-tower makeup water per day for a typical large-scale AI data center. The company says the system can deliver up to a 0.15 improvement in PUE for GPU HBM applications, and that higher operating temperatures can enable chiller-less or near-chiller-less operation for significant portions of the year in suitable climates. Phononic's Intelligent Thermal Fabric combines the thermoelectric hardware with real-time telemetry and software-defined controls to manage cooling across nodes.

Vertiv Holdings has agreed to acquire King Environmental Services, headquartered near Dublin, Ireland, to extend its fluid management and liquid cooling commissioning capabilities across Europe, the Middle East, and Africa. Financial terms were not disclosed; Vertiv said the deal is not expected to have a material impact on financial results and anticipates closing in the fourth quarter of 2026. KES, which also operates from a hub in Frankfurt, provides fluid-system flushing, filtration, water treatment, commissioning, thermal load testing, and electrical load testing for liquid-cooled data centers and critical power infrastructure. The acquisition follows Vertiv's December 2025 purchase of PurgeRite, which established a fluid management platform in North America.

ODATA is deploying Aligned Data Centers' DeltaFlow liquid cooling technology at its SP04 campus in Brazil and QR03 campus in Mexico, backed by a combined first-phase investment of US$630 million. The system, developed by Aligned in collaboration with Munters, uses a closed-loop circuit of pure, deionized water and is designed to operate without consuming water during cooling operations. DeltaFlow can also integrate with Aligned's Delta Cube air-cooling system, allowing both liquid and air cooling within the same data hall. DeltaFlow units are already being manufactured at Munters' facility in Araucária, Paraná, Brazil, with production intended to support future projects in Chile and Colombia.

Reflect Scientific has received an early-stage inquiry from a potential customer interested in deploying the company's waterless cooling technology in AI data centers. The inquiry signals growing interest in alternatives to evaporative and water-based cooling systems as data center operators face scrutiny over water consumption. Reflect Scientific's approach does not rely on water-based thermal management, which could reduce facility water use at high-density AI sites.

A new overview published by WSOC-TV details how AI and high-performance computing workloads are pushing cooling requirements far beyond what traditional data center designs were built to handle. The piece cites rack densities exceeding 100 kilowatts per rack in some AI deployments, compared to averages of 5 to 10 kilowatts per rack in legacy facilities. Engineers are increasingly turning to direct liquid cooling and immersion systems as the only viable path to managing heat at those densities.

Water Tech Online published findings drawn from water treatment work conducted across more than 200 data center sites, identifying recurring readiness failures in cooling systems including scaling, corrosion, and biological contamination that reduce efficiency and increase downtime risk. The analysis found that many facilities enter peak cooling seasons without baseline water chemistry data, making it difficult to respond to thermal load spikes. Operators are urged to implement continuous monitoring and pre-season treatment protocols rather than reactive maintenance schedules.

Data Center Dynamics published a technical overview of liquid cooling strategies being adopted as rack densities climb into the range required for GPU and AI accelerator deployments. The piece covers direct liquid cooling, immersion, and rear-door heat exchanger approaches, with attention to how operators choose between them based on facility vintage and power density targets. The article does not name specific deployments or dollar figures but addresses design trade-offs operators face as air cooling reaches its practical limits.

Trinity Biotech has launched a paid pilot program testing its proprietary cooling fluids inside AI data centers, Stock Titan reported. The pilot moves the company's technology beyond lab evaluation into revenue-generating commercial validation, a key threshold for enterprise adoption. Results from the pilot will determine whether Trinity Biotech's fluids can compete with established liquid cooling chemistries at the density levels demanded by current-generation AI accelerators.

Data Center Knowledge has published an analysis arguing that water usage effectiveness, the industry's standard water efficiency metric, is insufficient for measuring true water resilience at data center facilities. The piece proposes a broader assessment framework that accounts for source water availability, drought risk, local watershed stress, and supply chain dependencies. As water scarcity becomes a documented constraint in multiple US markets, operators and regulators are increasingly seeking metrics that reflect real-world risk rather than operational efficiency alone.

ET Datacenters reports that despite available technology, data centers across the industry seldom recycle or reuse cooling water because the economic trade-offs, including capital costs for treatment systems, operational complexity, and energy overhead, outweigh the savings in most markets. The analysis identifies the gap between water-efficiency goals and on-the-ground investment decisions as a structural problem rather than a technical one. Operators in water-scarce regions are the primary exception, where regulatory pressure or supply costs shift the calculation.

Data Center Frontier examines how specialized valves controlling coolant flow have become a critical and often overlooked component in liquid cooling deployments for high-density AI workloads. As GPU rack densities climb, precise thermal management at the valve level determines whether cooling systems can maintain safe operating temperatures. Supply constraints and engineering complexity around these components are beginning to factor into data center construction timelines.

Intelligent Living surveys emerging cooling approaches designed to reduce water consumption in AI data centers, including closed-loop liquid systems, dry coolers, and hybrid evaporative designs. Water use has drawn increasing regulatory and community scrutiny as facilities scale to support large language model training and inference. Operators are under pressure to demonstrate measurable reductions in water usage effectiveness metrics.

A JLL analysis finds that liquid cooling has moved from an experimental option to a mainstream deployment strategy in data centers, driven by the heat output of modern AI accelerators. The report highlights direct liquid cooling and immersion systems as the formats seeing the fastest adoption. Facility operators are now expected to specify liquid cooling capacity at the design stage rather than as a retrofit option.

Haffner Energy has introduced a combined biomass power and cooling system designed specifically for data centers, offering an alternative to grid-connected electricity for both power and thermal management. The system uses biomass combustion to generate electricity and capture waste heat for cooling purposes. The product targets operators seeking low-carbon energy sources that can function independently of constrained utility grids.

The Brookings Institution has released an analysis examining water consumption by AI systems and data centers, covering the scale of usage, geographic distribution of impacts, and policy options for regulators. The report arrives as federal officials have publicly disputed the extent of data center water use. Brookings identifies gaps in current disclosure requirements and outlines potential regulatory frameworks. The analysis is positioned to inform legislative and regulatory debates at both state and federal levels as water scarcity concerns mount in data center-dense regions.

Researchers in Germany and Japan have built a working prototype of a cooling system for data centers that requires no electricity, according to Data Center Dynamics. The system uses passive thermodynamic processes to dissipate heat, potentially eliminating one of the largest power draws in a standard facility. Details on the technology's scalability and timeline to commercial deployment were not provided in initial reporting.

Law firm Foley Hoag's second installment in its data center growth series examines how water consumption, siting constraints, and electronic waste rules are converging into a new compliance frontier for operators. The analysis highlights that water withdrawal permits are becoming as contentious as zoning approvals in water-stressed regions. E-waste regulations, particularly in the EU and select US states, are adding end-of-life obligations that few operators have fully priced into project economics.

A Data Center Knowledge analysis examines why water reuse remains rare in data center cooling operations, despite growing public scrutiny of water consumption. The piece identifies capital cost, regulatory complexity, water quality requirements, and operational risk as the primary barriers preventing operators from implementing closed-loop or reclaimed water systems. Industry experts cited say that reuse is technically feasible but economically unattractive under current utility pricing structures.

The World Economic Forum published a framework for water circularity in data centres, describing models that recapture and reuse water consumed in cooling systems rather than discharging it. The piece highlights facility-level and district-scale approaches, including heat recovery loops and closed-cycle systems, and names specific projects in Europe and Asia as examples. The WEF framing positions water circularity as an emerging industry standard rather than a voluntary add-on.

An op-ed published in The Seattle Times argues that AI data centers are discarding large quantities of heat that could be recovered and redirected to district heating systems, reducing both energy waste and emissions. The piece cites examples from European cities where data center waste heat already supplies residential and commercial buildings. The author contends that US operators and regulators have been slow to adopt comparable heat reuse requirements.

South Korean data center operators are accelerating deployment of liquid cooling systems as AI GPU workloads generate heat loads that conventional air cooling cannot manage, according to Chosunbiz. The shift is being driven by the density of AI training and inference clusters, where modern GPUs can exceed 700 watts per chip and rack densities push beyond 100 kilowatts. Korean operators, including major domestic cloud providers, are evaluating both direct liquid cooling and rear-door heat exchangers. The transition is creating demand for new infrastructure investment in facilities that were originally designed for air-cooled servers.

AI workload requirements are driving liquid cooling design changes that extend from the chip package all the way to facility-level valve and flow control systems, according to analysis published by Data Center Frontier. Higher GPU thermal design power figures, including chips exceeding 700W, require more precise coolant flow management than earlier generations of direct liquid cooling. Vendors are now engineering cooling infrastructure as an integrated thermal system rather than a collection of discrete components.

Graphene Manufacturing Group (TSXV: GMG; OTCQX: GMGMF) announced a new product called G FLUID, a graphene-enhanced liquid designed for immersion cooling in AI data center applications. The company did not disclose a commercial launch date or customer partnerships in the available announcement, but positioned G FLUID as a differentiated thermal management fluid for high-density AI racks. The product enters a market where liquid and immersion cooling demand is rising sharply alongside rack power densities.

NVIDIA has detailed a cooling breakthrough that allows its largest AI systems to operate using coolant at 45 degrees Celsius, above the threshold previously considered practical for liquid cooling at scale. The advance eliminates the need for chilling infrastructure in many deployments, reducing both capital cost and energy consumption for facility operators. NVIDIA says the approach is validated for its highest-density GPU configurations and is being adopted by early data center partners.

A startup backed by MIT researchers has developed a nuclear-inspired closed-loop cooling system designed to eliminate or dramatically reduce water consumption in data centers, MIT News reports. The system uses passive heat exchange principles derived from nuclear reactor thermal management and is being tested at small-scale facility deployments. The startup is targeting hyperscale operators facing water use restrictions in drought-affected regions.

Cogeneration and trigeneration systems, which generate electricity and simultaneously capture waste heat for cooling or heating, are seeing renewed interest as a solution for power-hungry AI data centers. The approach allows facilities to produce on-site power while using thermal byproducts to reduce mechanical cooling loads, addressing both energy cost and grid dependency concerns. Vendors and operators are revisiting decades-old industrial energy technology in the context of GPU-dense, high-heat AI workloads.

New refrigeration architectures are being developed and evaluated specifically to support rack densities reaching 200kW per rack, a threshold that conventional air cooling cannot address. Data Center Frontier's coverage details the engineering requirements and system configurations needed to manage heat at that scale, including direct liquid cooling loops and high-capacity chilled water infrastructure. The 200kW rack benchmark is increasingly cited as the planning target for next-generation AI training clusters.

Data Center Knowledge reviews the current state of water-free cooling technologies and assesses whether they are approaching commercial viability at scale. The report covers approaches including rear-door heat exchangers, dry coolers, and advanced air-side economization that eliminate or drastically reduce water consumption. Several operators facing water use restrictions in drought-prone regions are accelerating evaluation of these systems.

Trane and Eaton are jointly promoting integrated system designs that combine power distribution and cooling infrastructure for high-density AI data centers, presenting the approach as a way to reduce complexity and improve efficiency in facilities running dense GPU loads. The companies say the integrated architecture addresses the challenge of coordinating power and thermal systems that have traditionally been specified and installed separately. Neither company disclosed specific deployment sites or contract values.

Castrol announced that its ON PG25 cooling fluids have been validated for use in NVIDIA AI factory and data center infrastructure, signaling formal qualification within NVIDIA's ecosystem for immersion or liquid cooling deployments. The validation covers compatibility with NVIDIA hardware at the facility level rather than just component level. Castrol, a BP subsidiary, enters a market where fluid qualification is increasingly a procurement requirement for large-scale AI compute builds.

Malaysia is preparing to commercialize a palm oil-based dielectric fluid for single-phase immersion cooling in data centers, positioning the country's dominant agricultural commodity as a new input for AI infrastructure. The fluid is being developed as a biodegradable alternative to synthetic dielectric liquids currently imported from Western suppliers. No launch date or pricing was disclosed in the announcement.

A new MarketsandMarkets report projects substantial growth in Singapore's data center liquid cooling market through 2031, driven by rising compute density and regional AI infrastructure investment. Singapore has positioned itself as a Southeast Asian hub for hyperscaler facilities, making thermal management a critical operational and regulatory concern. The report covers immersion cooling, direct liquid cooling, and related thermal systems.

Nasscom has published an analysis of how liquid-cooled data centers manage thermal load across thousands of concurrent GPU threads, examining heat distribution, coolant routing, and load-balancing strategies at scale. The piece addresses the specific challenge of maintaining temperature stability when GPU utilization spikes unpredictably across large clusters. Liquid cooling adoption has accelerated as air cooling reaches its limits with high-density AI compute racks. Operators are increasingly treating thermal architecture as a core engineering constraint alongside power delivery.

Johnson Controls published analysis arguing that zero-water cooling systems are viable and preferable for data centers, citing advances in air-side and refrigerant-based thermal management that eliminate the need for evaporative cooling water. The company positions its approach as a response to regulatory and community pressure over water consumption at large facilities. The piece arrives as water usage effectiveness metrics face increasing scrutiny from state regulators and local governments.