DCX launches 8MW cooling distribution unit

DCX Liquid Cooling Systems has launched a new cooling distribution unit (CDU).

The company this week announced the new, second-generation facility distribution unit (FDU V2AT2). The new CDU delivers up to 8.15MW of heat transfer capacity.

Designed to support AI hardware, the company said the new CDU supports 45°C (113°F) warm-water cooling for Nvidia’s NVL72 GB200 / GB300 Blackwell and Vera Rubin architectures, which can remove the need for chillers on the heat-rejection side in many deployments.

“As the data center industry transitions to AI factories, operators need cooling system that won’t be obsolete in one platform cycle,” said Maciek Szadkowski, CTO at DCX Liquid Cooling Systems. “The FDU V2AT2 replaces multiple legacy 1.3MW CDUs and enables 45°C supply water operation. This new category of CDUs has minimum thermal loss with AT2 approach temperature, and provides multi-megawatt cooling at the hall level. That opens a clear path to Nvidia Vera Rubin architecture and beyond while simplifying cooling loop topology and significantly reducing both capex and opex of data center liquid cooling system.”

Poland-based DCX is a manufacturer of liquid cooling solutions for both direct liquid cooling and immersion cooling systems. The company provides in-rack, in-row, and hall-wide CDUs, cold plates, immersion cooling enclosures, dry coolers, modular containerized data centers, and enclosures for cryptomining.

Schneider Electric and Motivair launch new CDU

Schneider Electric has launched a new coolant distribution unit (CDU) for liquid-cooled data centers.

Motivair by Schneider Electric, this week, introduced a new 2.5MW CDU designed to cool high-density data centers.

The MCDU-70 is the highest-capacity CDU available from Motivair. The company said the CDU can support the cooling needs of GPUs and gigawatt-scale AI Factories.

With the addition of the MCDU-70, Schneider Electric’s end-to-end liquid cooling portfolio now offers CDUs ranging from 105kW to 2.5MW.

Motivair is based in Buffalo, New York, and produces units that pump coolant at high pressure close to the chips to cool the servers efficiently. Its products include cold plates, chillers, in-rack manifolds, hoses and cooling loops, and heat exchangers and dissipation units.

Applied Digital starts work on 430MW data center campus

Data center firm Applied Digital has broken ground on a new campus.

The company this week announced that it has broken ground on Delta Forge 1, an AI factory campus located in a “strategic southern US market.”

Delta Forge 1 is designed to support an initial 430MW of total utility power, enabling up to 300MW of IT load, with the ability to “scale considerably” in 2028.

The first phase will comprise two 150MW buildings on some 500 acres. Initial operations at Delta Forge 1 are expected to commence in mid-2027.

Applied said it was in discussions with a “prospective investment-grade hyperscale customer” over the campus, though details weren’t shared.

The company’s Polaris Forge 1 and 2 hyperscale campuses are located in Ellendale and Harwood, North Dakota. The former is set to be leased to CoreWeave while much of the latter is set to be leased to an unnamed “investment grade hyperscaler.”

Infrakey, City of Lacy Lakeview unveil 925MW data center district in McLennan County, Texas

Infrakey DC Parks and the City of Lacy Lakeview in Texas have unveiled the Lacy Lakeview Data District (L2D2). The public-private partnership covers around 520 acres in McLennan County, Texas, with a planned power capacity of up to 925MW.

According to the partners, the park is designed to provide an infrastructure framework that will support the development of hyperscale, AI-driven data centers. Previous reports have suggested that, at full build-out, the site could host up to six data centers with 1GW of capacity.

Instead of a single-tenant campus, L2D2 is expected to be structured as a district-scale infrastructure platform. The partners claim that the project will offer a coordinated stack that includes ERCOT grid interconnection with a district substation, behind-the-meter power generation and energy storage, microgrid controls and energy optimization, reclaimed water sourced through the City’s wastewater system, multi-path fiber connectivity, and optional district cooling.

Phase I of the project is expected to deploy 300MW of grid-based power, with 150MW available through an initial substation interconnection, and an additional 150MW of capacity slated for connection by the end of 2027.

The partners claim that at full build-out, the project will comprise more than $2 billion in infrastructure investment, with the district designed to scale over several years.

KDDI launches data center in Osaka, Japan

Japanese telecoms firm KKDI has launched the first phase of a new data center in Osaka.

The company this week announced that operations at the Osaka Sakai Data Center, located near the Osaka metropolitan area, started on January 22.

The former Sharp LCD screen manufacturing plant, which was acquired in April 2025, took just six months to convert to data center use.

The four-story facility, totaling 57,000 sqm (613,542 sq ft), is hosting Nvidia GB200 NVL72 hardware. It uses a combination of air cooling and direct liquid cooling.

KDDI said 100 percent of the site’s energy use is offset by renewable power.

The site will primarily serve KDDI, which will offer access to the site’s GPUs via its KDDI GPU Cloud service. The site will start accepting general service applications in April 2026, but the company named three companies as early customers.

KDDI Group company Medical Engineering Institute Inc., together with Takeda Pharmaceutical Company Limited, aims to use this data center from April 2026 in order to analyze medical data and explore new research projects using AI. By applying large language models to electronic medical records, the companies hope to help drug discovery, clinical research, and other medical research.

Morgenlot Inc, a research startup, will be using the site to aid fluid analysis in product design, particularly in the manufacturing industry.

Nokia and Hypertec partner on Nibi supercomputing cluster at Canada’s University of Waterloo

Nokia has partnered with server IT equipment manufacturer Hypertec for the deployment of the Nibi supercomputing cluster at the University of Waterloo in Ottawa, Canada.

The cluster will be integrated into the SHARCNET (Shared Hierarchical Academic Research Network) supercomputer, designed to support health, climate science, engineering, and AI research projects.

Comprising 134,000 Intel Xeon 6 (Granite Rapids) cores, 288 Nvidia H100 GPUs, and more than 25 petabytes of storage, the Hypertec-designed system is interconnected by Nokia 400G Ethernet. Nibi also features nine single-phase immersion cooling tanks and uses direct-to-chip liquid cooling for its GPU subsystems.

The use of immersion cooling has resulted in a reduction in up to 50 percent in cooling energy, when compared to air cooling, with the system offering 100 percent heat reuse, with the waste heat used to warm the Mike and Ophelia Lazaridis Quantum-Nano Centre, the university’s quantum R&D laboratory.

Nibi is Canada’s first large-scale immersion-cooled supercomputer and the first HPC cluster based on Intel Xeon 6 processors to be deployed using immersion cooling. It was built to replace the university’s previous flagship supercomputer, Graham, which was retired in late 2025.

The system, which ranked 186 on the most recent edition of the Top500 list, is available to staff, students, and research fellows at Canadian academic institutions. Nibi is supported by 19 academic partner institutions, making it the largest HPC consortium in Canada.

Global Switch signs eight year PPA with RWE to power Docklands data center in London

Global Switch has signed an eight-year Power Purchase Agreement (PPA) with German renewable energy developer RWE in the UK.

The PPA will see the data center firm offtake around 70GWh of power per year from RWE’s Brechfa Forest West onshore wind farm in Wales. The PPA, which came into force this month, will be attributed to Global Switch’s London Docklands data center.

Commissioned in 2018 and comprising 28 turbines, the wind farm has a capacity of 57.4MW.

“Our agreement with RWE is a critical next step on our journey to purchasing 100 percent renewable energy by 2030, and to the setting of new standards for what a sustainable, environmentally-conscious data center can look like.

Global Switch’s Docklands data center has 224MVA of power available and hosts H100 and H200 GPUs for CoreWeave, in what Global Switch claims is Europe’s largest deployment of the Nvidia hardware.

The company operates two sites in the Docklands - London East and London North. It announced plans for a third facility, London South, which will offer 40MW across 27,000 sqm (290,625 sq ft) in 2024. In June of last year, it launched a showcase of liquid cooling technologies at the Docklands site.

Nvidia invests another $2bn in CoreWeave

Nvidia has invested another $2 billion in AI cloud provider CoreWeave.

First reported by Bloomberg and confirmed by the companies, the investment aims to help CoreWeave speed up its build-out of more than 5GW of AI capacity by 2030, and will also see Nvidia giving the neocloud access to the upcoming Vera chip - Nvidia's latest CPU.

Nvidia has purchased CoreWeave Class A common stock at $87.20 a share. The investment builds on the September 2025 agreement between the two that would see Nvidia purchasing unsold compute capacity from CoreWeave for $6.3bn.

Nvidia will assist CoreWeave with the purchase of land and power for data centers, and will market CoreWeave’s offerings to cloud partners and large business customers.

According to CoreWeave's CEO Mike Intrator, the investment from Nvidia represents around two percent of its planned spend on bringing new infrastructure online - suggesting $200bn in anticipated spend. Intrator told Bloomberg: "This year, we’re going to deliver an enormous amount of infrastructure, and that’s just going to accelerate over the next three years." It is unclear what timeline the $200bn spend could refer to.

In addition to investing in CoreWeave, Nvidia has also committed to giving the neocloud access to the Vera CPU. Typically, Nvidia CPUs have been offered in systems in combination with other chips.

CoreWeave will be among the first to gain access to Vera, but Huang declined to comment on the other companies set to receive early deliveries, though he told Bloomberg, "there are going to be many.“

Global Switch suffers battery room fire at London data center

Global Switch suffered a small fire at its London data center over the weekend.

The London Fire Brigade (LFB) announced over the weekend that eight fire engines and around 60 firefighters were called to a “small fire” at a commercial building on Nutmeg Lane in Blackwall in the early hours of Saturday (January 24) morning.

A fire incident was reported to the Brigade’s 999 Control officers at 04:52am and “quickly extinguished” by 6.43am.

The fire occurred at Global Switch’s London data center. Local press reports that the incident occurred in the building’s basement.

There were no reports of any injuries. There was reportedly no disruption to customer services.

The scale of the brigade’s response was seemingly disproportionate to the actual size of the fire due to the fact that the fire was in a high-rise building. Crews had been mobilized from Poplar, Millwall, Plaistow, Stratford, East Ham, and East Greenwich fire stations to the scene.

Global Switch’s Docklands data center has 224MVA of power available and hosts H100 and H200 GPUs for CoreWeave, in what Global Switch claims is Europe’s largest deployment of the Nvidia hardware.

The company operates two sites in the Docklands - London East and London North. It announced plans for a third facility, London South, which will offer 40MW across 27,000 sqm (290,625 sq ft) in 2024. In June of last year, it launched a showcase of liquid cooling technologies at the Docklands site.