DUG offers 1MW immersion-cooled data center container

Australian firm DUG has unveiled a new, larger version of its immersion-cooled containerized data center.

The company has announced the DUG Nomad 40, a 40-foot version of its containerized offering. The pod utilizes DUG’s proprietary single-phase immersion cooling system, and reportedly offers a PUE of 1.05.

The Nomad 40 features 12 DUG Cool tanks, with up to 26 RU of immersed hardware per tank that can cool 84kW each. The pods are described as “Starlink compatible” for remote connectivity.

Previously known as DownUnder GeoSolutions, DUG was founded in Perth in 2003 to provide immersion-cooled high-performance computing (HPC) solutions for scientific data analysis and provides HPC-as-a-service to customers.

Clients include Murdoch University, Australian National University, University of Western Australia, Monash University, shipbuilder Austal Australia, Australia’s CSIRO research agency, crypto firm HODL Ranch, and a number of oil and gas firms.

DUG currently has HPC deployments in Houston, Texas; Kuala Lumpur, Malaysia; and Perth, Australia. The company is also planning another Australian site in Geraldton, some 400km north of Perth in Western Australia.

The company first announced plans to offer a containerized product last year. DUG has previously told DCD that the offering is targeted at Edge use cases where data needs to remain stored locally.

The original Nomad 10 offers 80kW of cooling and can host more than 80 immersion-cooled Nvidia H200 GPUs in a 10 ft container.

DUG has licensed out its immersion cooling tech to cooling solutions provider Baltimore Aircoil Company, Inc.

Airedale by Modine announces stainless steel extension of TurboChill DCS chiller for liquid cooled data centers

Modine (NYSE: MOD), a diversified global leader in thermal management technology and solutions, has announced the addition of a stainless steel variant to its flagship Airedale by Modine TurboChill DCS chiller range. The stainless steel design provides the hydraulic capacity needed for direct liquid cooling (DLC) systems, which are increasingly used in high-density AI deployments across hyperscale, colocation, and neo-cloud facilities.

Providing a robust stainless steel interface to the facility water system loop, this extension to the TurboChill DCS range significantly enhances circuit cleanliness and overall system reliability. The corrosion-resistant material safeguards the integrity of the cooling loop, reducing contamination and build-up on the cold plates. Designed for high-pressure environments, the TurboChill DCS Stainless Steel maintains structural stability under extreme thermal loads while enabling precise filtration and fluid management.

When integrated within optimized liquid cooling architectures, the TurboChill DCS Stainless Steel can facilitate the complete elimination of in-row coolant distribution units (CDUs), reducing system complexity and potentially improving PUE, while freeing up space in the data center hall for additional IT racks.

Suitable for global application, the stainless steel variant of the Turbocor compressor chiller meets Seismic Design Category D. Designed for durability and to meet customers’ ambitious efficiency, performance, and sustainability objectives, TurboChill DCS chillers are optimized for higher ambient temperature operation of up to 55°C (131°F), with low GWP refrigerant, R1234ze (GWP 1.37).

Accelsius claims 300MW customer win with new operator in Canada

Two-phase direct-to-chip liquid cooling firm Accelsius has secured a new data center customer win in Canada.

The company this week announced that DarkNX, a new digital infrastructure company, has entered into an agreement to deploy Accelsius’ NeuCool technology across a planned 300MW AI data center campus in Ontario, Canada.

The first phase is set to include two facilities at 65MW, with deployments scheduled for 2026 and 2027. Further details weren't shared.

DarkNX selected NeuCool for chip-level cooling alongside chiller systems from Johnson Controls for facility cooling. Accelsius said the project is expected to be the largest two-phase, direct-to-chip deployment to date.

Texas-based Accelsius was originally founded in 2022 by Nasdaq-listed industrial firm Innventure to commercialize technology developed by Nokia's Bell Labs that had not been productized. Johnson Controls recently invested in the firm.

Accelsius has deployed its technology at several data center labs, and recently posted on LinkedIn that its NeuCool Thermal Simulation Rack is headed to London for deployment at Telehouse's new Liquid Cooling Lab. It has also deployed a cooling system to an Equinix lab in Virginia and at Park Place Technologies' new facility in Cleveland, Ohio.

DarkNX launches

Founded this year, DarkNX describes itself as a developer of high-density modular data centers optimized for AI and GPU workloads.

Few details about the company or its planned projects are available. The company claims to have a proprietary immersion cooling technology known as AquaFlow.

Majestic Labs raises $100m for its all in one servers designed to eliminate the memory wall

An AI server startup founded by former Google and Meta executives emerged from stealth earlier this month following a $100 million Series A funding round.

Israel-based Majestic Labs claims to have “dramatically” improved AI infrastructure with its system architecture that aims to rebalance memory and compute, eliminating an issue known as the “memory wall” – the gap between processor speed and memory bandwidth.

According to the company, its all-in-one server is capable of handling the largest AI workloads that currently require multiple racks of servers and switches, delivering the memory capacity of ten or more racks in one server. Its approach reportedly allows companies to realise more than 50x performance gains, whilst reducing power and cooling costs associated with a large number of racks, Majestic claims.

The company’s founding team consists of Ofer Schacham, Masumi Reynders, and Sha Rabii, who built FAST (Facebook Agile Silicon Team) at Meta Reality Labs and GChips at Google.

Co-founder and COO, Reynders, added: “AI infrastructure is scaling at unprecedented speed, but the industry has not solved key fundamental architectural inefficiencies. Majestic addresses this by delivering immediate operational gains on today's workloads while maintaining full programmability and flexibility to adapt as AI evolves beyond transformer-based models.”

Cooling firm Xnrgy secures investment from Capital Bridge Group and Prologis

Cooling firm Xnrgy Climate Systems has secured new investment from Prologis and Capital Bridge Group.

Xnrgy, which provides liquid and air-cooling technologies as well as control systems, this week said it has completed its previously announced financing round with an investment from Capital Bridge Group and Prologis Ventures. The size of the round wasn’t disclosed.

Founded in 2017, Xnrgy designs and manufactures HVAC (heating, ventilation, and air conditioning) systems. The company offers CRAH (computer room air handler) and CRAC (computer room air conditioning) systems, supporting cooling capacities from 60kW to more than 1MW, as well as chillers, fans, and dehumidifiers.

Other investors in the round include Decarbonization Partners (a joint venture between BlackRock and Temasek), Climate Investment (CI), and Activate Capital.

Activate Capital previously invested in Xnrgy earlier this year. Idealist Capital and MKB invested in the firm in 2023.

Earlier this year, Xnrgy announced the construction of Mesa 2, a new 330,000 sq ft (30,660 sqm) facility dedicated to producing its next-generation air-cooled chiller. The firm’s operational footprint across Mesa, Arizona, and Montreal, Quebec totals nearly one million sq ft (92,900 sqm).

Capital Bridge Group is a growth-stage investment and advisory firm founded by Michael Ronen, who was previously with Fortress Investment Group. Before that, he was partner and COO of global TMT banking at Goldman Sachs and a managing partner at SoftBank's Vision Fund.

OpenAI partners with Foxconn to optimize data center design

OpenAI has partnered with Foxconn to optimize data center design.

The AI giant announced yesterday that its partnership with Foxconn, formally known as Hon Hai Technology, would focus on “design work and US manufacturing readiness for the next generation of AI infrastructure hardware.”

As part of the agreement, the pair will design and develop “multiple generations of AI data center racks” in order to keep up with the compute needs of new models.

Foxconn has also agreed to manufacture key data center components, including cabling, networking, cooling, and power systems, in the US.

Although the agreement does not contain purchase commitments or financial obligations, OpenAI will “have early access to evaluate these systems and an option to purchase them.”

This is not a new partnership – both companies are already working together on the Stargate project. In August, Foxconn confirmed that it would be manufacturing equipment for the project at a former General Motors factory in Lordstown, Ohio.

This announcement is also the latest in a series of US-based moves that the Taiwan-based electronics manufacturer has made in the past year. In July, the company partnered with the American subsidiary of Taiwanese electric motor company Teco to manufacture modular data centers in the US.

American businesses and the government have urged key Taiwanese semiconductor companies to move their manufacturing to the US amidst concerns that Taiwan’s security situation could jeopardize the global supply chain.

Nscale signs 15MW lease agreement with Verne for data center capacity in Iceland

AI cloud provider Nscale has signed on to lease 15MW of capacity at Verne's data center campus in Iceland.

The agreement will see Nscale deploying 4,600 Nvidia Blackwell Ultra GPUs across Verne's campus through 2026. The GPUs will be 85 percent liquid-cooled and 15 percent air-cooled.

Verne's Iceland campus is fully powered by renewable energy, and due to the climate, it can utilize natural free cooling. According to a data sheet, the campus spans 40 acres and has more than 140MW of IT capacity. The facility is also Nvidia DGX authorized.

Philippe Sachs, chief business officer and president of EMEA, Nscale, added: “As compute demand grows, we’ve worked with partners throughout the world and the Nordic region to deliver sustainable solutions to meet that demand. The Nordics offer a uniquely sustainable foundation – abundant renewable energy and natural cooling. With our existing operations in Norway, we’ve seen first-hand how the region powers low-carbon, sovereign-grade AI infrastructure, Verne has been an exceptional partner – agile, technically rigorous, and aligned with our long-term sustainability vision.”

Verne operates data centers in Iceland, Finland, and the UK. The company was acquired by Ardian in 2024. Last month, Ardian raised $20bn for its infrastructure platform dubbed Ardian Infrastructure Fund VI (AIF VI).

Nscale has been expanding rapidly of late, having signed an agreement with Microsoft to provide the cloud giant with 116,000 Nvidia GB300 GPUs, as well as deploying a cluster in Portugal, and signing a deal to provide compute to Microsoft in Norway.

Dell updates its server, storage, and networking portfolios to support AI deployments at scale

Dell has announced a number of updates to its server, storage, and networking offerings to better support organizations looking to scale AI deployments.

In a statement, the company said the updates to its PowerEdge, PowerScale, and ObjectScale solutions will help boost performance, scalability, and data discovery capabilities.

Deepening its partnership with Nvidia, Dell has launched the PowerEdge XE8712 server, which the company claims will deliver the industry’s “highest GPU density in a standard rack,” with up to 144 Blackwell GPUs per Dell IR7000 rack. Available to customers from December, the XE8712 also features Dell’s Integrated Dell Remote Access Controller (iDRAC) monitoring software, OpenManage Enterprise (OME), and the Integrated Rack Controller (IRC) for advanced thermal controls.

Dell also unveiled its PowerEdge R770AP server, featuring Intel Xeon 6 P-core 6900-series servers. Also available from December, it features enhanced parallel processing, reduced memory latency, and increased PCIe lanes to support accelerated trading algorithms, scalable memory configurations, and improved network performance.

The air-cooled offering features high-core-count CPUs, large cache sizes, and support for CXL memory expansion, and will sit alongside Dell’s PowerEdge XE9785 and XE9785L rack-scale systems, launched in May 2025 and purpose-built for next-generation AI and HPC workloads.

The 10U XE9785 is also air-cooled, while the 30U XE9785L features direct liquid-cooling technology. Both offerings comprise dual-socket AMD Epyc processors and eight AMD Instinct MI355X GPUs per node. Combined with AMD Pensando Pollara 400 AI NICs and the Dell PowerSwitch AI fabric, Dell said the servers will deliver scalable compute, improved TCO, and reduced operational costs.

PDG breaks ground on 120MW data center campus in Jakarta, Indonesia

APAC operator Princeton Digital Group (PDG) has started work on a new data center campus in Jakarta, Indonesia.

The company this week broke ground on its 120MW hyperscale data center campus in Greenland International Industrial Center (GIIC), Bekasi Regency, Greater Jakarta.

Set to offer both liquid and air cooling, the first phase of the $1 billion JC3 campus is due to go live by Q4 2026.

“Indonesia is one of the fastest-growing digital economies in Asia Pacific, and we remain deeply committed to supporting its continued growth,” said Varoon Raghavan, co-founder and COO of PDG. “Backed by a proven track record of delivery excellence across the region, our JC3 campus is purpose-built to provide the scale, performance, and sustainability our global and local customers require as cloud and AI workloads accelerate.”

With the addition of JC3, PDG’s total capacity in Indonesia will reach approximately 230MW. Set on the same campus in Cibitung, JC1 and JC2 offer more than 25MW across a combined 14,780 sqm (159,090 sqm). JC2 launched last year.

Google opens data center in Winschoten, the Netherlands

Google has launched its new data center in Winschoten, Groningen, the Netherlands.

The data center expands the capacity of the company's europe-west4 cloud region, which has existing data centers in Eemshaven.

Google broke ground on the data center, which is its third in the country, in December 2023. At the time, it was reported that the company would invest €600 million ($655.1m) into the new development, located at Hoogebrug 3. As of 2024, Google said it had invested more than €3.8 billion ($4bn) in data centers and digital infrastructure in the Netherlands.

The new facility in Winschoten is designed to support waste heat reuse for future district heating networks and has solar panels on the roof. It also uses "advanced air-cooling technology" to help reduce water usage.

The cloud company has two other data centers in the country: Eemshaven, which opened in 2018, and Middenmeer, which opened in 2020. The company has also bought land at the Westpoort industrial estate located west of Hoogkerk, in the municipality of Groningen, for a data center, which broke ground in April 2024.

Across the company's data centers in the country, Google employs around 700 people.

To date, Google says it has purchased 1GW of renewable energy generation capacity in the Netherlands, including a PPA with Shell signed in May of this year. The company also previously invested €45 million ($52m) in the construction of a water treatment plant that processes water from the Eemskanaal and supplies it to companies in the Groningen port areas for cooling, including Google's data center in the Eemshaven.

Schneider Electric and Switch pen $1.9bn cooling deal

Switch Data Centers and Schneider Electric have agreed an expanded $1.9 billion contract that they say is North America’s largest-ever data center cooling deal.

The supply capacity agreement (SCA) between the two firms will see Schneider supply prefabricated power modules for Switch’s data centers, as well as its Uniflair chillers. It will be the first time the chillers have been deployed in the US.

Switch Data Centers and Schneider Electric have agreed an expanded $1.9 billion contract that they say is North America’s largest-ever data center cooling deal.

The supply capacity agreement (SCA) between the two firms will see Schneider supply prefabricated power modules for Switch’s data centers, as well as its Uniflair chillers. It will be the first time the chillers have been deployed in the US.

The SCA contract is designed to provide guaranteed capacity to the end user, in this case Switch, while also allowing some flexibility for how this is delivered. The companies say this is useful in the rapidly evolving AI infrastructure market.

Uniflair chillers use oil-free, variable-speed centrifugal compressors and integrated free cooling to match capacity to real-time IT load, preventing overcooling and reducing run hours.

The prefabricated power modules come with standardized, pre-tested layouts that optimize airflow and containment to increase economizer hours and reduce cooling energy compared with traditional builds.

The agreement is the largest cooling service engagement Schneider Electric has ever undertaken.

Switch's traditional proprietary designs feature hot-aisle containment, but the company has now developed a liquid cooling offering too, and is hosting an Nvidia GB300 NVL72 deployment on behalf of AI cloud provider CoreWeave.

TRG breaks ground on 24MW data center in Houston, Texas

US operator TRG Datacenters has broken ground on a new facility in Houston, Texas.

The Tallvine Partners-backed company this week announced it has broken ground on HOU2, the second data center on its Spring, Texas campus.

The purpose-built HOU2, located at 2626 Spring Cyprus Road, will add 24MW across 110,000 sq ft (10,219 sqm). The facility is due to go live in Q4 2026. The facility will use a closed-loop, dry-chiller cooling system.

TRG said its HOU1 data center has maintained 100 percent uptime since going live in 2018, serving more than 160 customers, including multiple Fortune 500 businesses.

Attendees of the groundbreaking ceremony last week included representatives from CenterPoint Energy, Walker Engineering, HTS, Thomas Craig Construction, and Encore Concrete Construction.

Traditionally a retail colocation provider, HOU2 marks TRG's entry into the single-tenant/wholesale market. The facility is being offered for pre-lease as a dedicated data center, select few private data halls, and/or anchor tenant + retail colo configuration.

The site will be able to host high-density AI racks, including Blackwell GPUs. The company has operational Blackwells at its HOU1 facility.

Investment firm Tallvine Partners acquired TRG earlier this year, with plans to expand the company’s footprint across the US. The company aims to develop in Dallas, Texas, in the future.