Schneider releases new reference designs for Nvidia’s GB300 NVL72

Schneider releases reference designs for Nvidia’s GB300 NVL72. The company has announced new reference designs developed with Nvidia it says will “significantly accelerate time to deployment and aid operators” as they adopt high-density AI infrastructure.

The first reference design delivers a framework for integrated power management and liquid cooling control systems that includes interoperability with Mission Control, Nvidia’s operations and orchestration software, including cluster and workload management features.

The second reference design focuses on the deployment of AI infrastructure for AI factories of up to 142kW per rack, specifically Nvidia’s GB300 NVL72 racks, in a single data hall. The reference design includes information on four technical areas: facility power, facility cooling, IT space, and lifecycle software.

Schneider partners with Compass for pre-fab modules

Compass Datacenters and Schneider Electric have jointly engineered a pre-fabricated, modular solution for data center end-to-end space optimization (white space fit-out).

The companies said the Schneider Electric EcoStruxure Pod delivers all the elements of data center white space in a prefabricated pod solution that can be quickly deployed on-site with a simplified installation process.

The pod accommodates hot aisle containment, in-row, and rear door heat exchanger cooling, while integrating cabling, high-power busway, and a technical water loop for liquid cooling.

Schneider Electric has begun manufacturing the Prefabricated Modular EcoStruxure Pods for customers. The first units will be deployed at a Compass Datacenters campus later this fall.

Arista developing liquid-cooled network switches

Networking firm Arista is developing liquid-cooled switches and racks. As reported by NetworkWorld and Converge Digest, the company outlined its plans at the recent Hot Interconnects conference and its recent analyst day.

The company is reportedly developing an ORv3W-based data center server rack that supports approximately 120kW of power and can run up to 32 AI fabric switches along with fiber patch panels, management switches, and power shelves. The rack supports tight liquid fittings specifically for high-density switch cooling systems.

“A lot of our development is focused in this direction,” Arista co-founder Andreas Bechtolsheim said during the analyst day. “In addition, we are designing fully liquid-cooled switch racks that offer up to 32 of these fabric switches with patch panels, and management switches, and power shelves. And again, this is all optimized for the liquid-cooled data center to enable high-density switch configurations.”

Removing fans in liquid-cooled switches can reportedly save 5–10 percent of power, reduce vibration, and improve reliability.

While companies have previously trialed using liquid cooling for network switches, most network systems and rooms remain air-cooled. Network switch densities are still well below the TDP where liquid cooling would be mandatory, often requiring less cooling than CPUs or GPUs.

“Liquid cooling could potentially save another 10 percent for network infrastructure at the system level by eliminating active fans required for traditional air cooling,” Arista has previously said in a white paper, suggesting potential savings in large-scale AI clusters.

Microsoft increases Wisconsin data center investment to $7.3bn, says Fairwater site will be "world’s most powerful data center"

Microsoft said that it is "in the final phases" of building its $3.3 billion Mount Pleasant, Wisconsin, data center project.

At the same time, it said that it plans to spend an additional $4 billion over the next three years on a second data center of similar size and scale in the state. Microsoft said that it has multiple identical data centers under construction around the country.

The facility set to launch shortly is known as Fairwater, with Microsoft claiming that it will be the "world’s most powerful data center.“

The company did not share benchmarks to validate the claims, but said that it will house hundreds of thousands of the Nvidia GB200 GPUs, connected by enough fiber "to wrap the planet four times over." It said that it will have ten times the performance of today’s fastest supercomputers, but provided no further details.

While traditional Azure data centers are designed for thousands of independent workloads, this site is expected to be used as a single AI facility, both for training and inference.

The Fairwater site spans 315 acres and comprises three large buildings with a combined 1.2 million sq ft (111,483 sqm) under roofs. Construction required 46.6 miles of deep foundation piles, 26.5 million pounds of structural steel, 120 miles of medium-voltage underground cable, and 72.6 miles of mechanical piping.

Vantage enters Spain, to invest €3.2 billion in Zaragoza data center campus

US operator Vantage Data Centers is entering the Spanish market. The company has announced plans to invest €3.2 billion ($3.78bn) in developing a data center campus in Villanueva de Gállego, the Zaragoza area of Aragon.

The project will be developed in collaboration with the Aragonese company Desarrollos Ecoindustriales La Cartuja. It will be carried out in five phases over ten years, according to the Government of Aragon.

The data center will occupy 40 hectares of undeveloped industrial land parallel to the Huesca highway, next to the San Jorge University campus and opposite the San Miguel industrial estate.

Efficient energy supply

The campus will be in an eco-industrial park promoted by Vantage Data Centers, which plans to integrate local renewable energy generation facilities (wind and photovoltaic) and energy storage systems aligned with climate neutrality goals.

The campus is also planned to use a closed-loop cooling system to ensure that its operation does not place additional demand on water resources.

CoreWeave plans £1.5bn UK AI data center investment, will deploy Nvidia GPUs at Scotland's DataVita data center

US neoloud CoreWeave plans to invest another £1.5 billion ($2bn) in UK AI data center capacity and operations.

The investment comes on top of a previous £1bn ($1.4bn) commitment, and includes a deployment at a DataVita data center in Scotland.

The news comes amid US President Donald Trump's visit to the UK, which has coincided with a slew of data center announcements, including a £5bn ($6.9bn) Google deal, a $678m BlackRock venture, and a new Vantage facility. An announcement is expected from Nscale and OpenAI this week.

CoreWeave will deploy Nvidia Grace Blackwell Ultra GPUs at one of DataVita's data centers in Scotland. The company operates the 8,000 sq m (86,111 sq ft) DV1 data center, supporting up to 40MW, and the 1,000 sq m (10,764 sq ft) DV2, supporting 1MW.

The deployment will use renewable energy and closed-loop cooling systems.

CoreWeave will deploy Nvidia GB300 GPUs and RTX PRO Blackwell Server Edition GPUs at other facilities in the UK. In January, the company took space in two Digital Realty and Global Switch data centers around London.

“Artificial Intelligence will drive incredible change in our country and has the potential to transform our public services and infrastructure,” the UK's Prime Minister Keir Starmer said.

Nvidia to invest $5bn in Intel, companies will develop custom CPUs together

Nvidia will invest $5 billion in Intel as part of an agreement to jointly develop custom data center and personal computing hardware.

In a statement, the two companies said they would work to connect Nvidia and Intel architectures using Nvidia’s NVLink technology, with Intel set to build custom x86 CPUs that will integrate into Nvidia AI infrastructure platforms.

For the personal computing segment, Intel will produce x86 system-on-chips (SOCs) that integrate with Nvidia RTX GPU chiplets for inclusion in PCs.

Subject to customary closing conditions, including required regulatory approvals, Nvidia will buy $5bn worth of Intel stock at $23.28 per share. Jensen Huang and Lip-Bu Tan, CEOs of Nvidia and Intel, respectively, will hold a joint conference later today to further discuss the announcement.

The news comes less than a month after the US government announced it had taken a 9.9 percent stake in Intel, with the state investing $8.9 billion in the chipmaker – $5.7bn of which came from funds that would have been awarded to the company through the CHIPS Act, and $3.2bn from funds under the Secure Enclave program.

Intel has already received $2.2bn from the CHIPS subsidies, but the claw-back and profit-sharing provisions in the subsidies will be eliminated, the government said at the time.

The deal comes after a turbulent time for the US chipmaker, where it has seen its market share eroded both in chip design and manufacture, damaging the capabilities of the nation's only leading-edge chipmaker.

Plans filed for 1GW data center campus in Sydney, Australia

A gigawatt-scale data center campus could be coming to Sydney, Australia. Project management firm Plan Project Management has filed a SEARS application with the New South Wales Government for the Mamre Road Data Centre Campus.

The AU$5 billion (US$3.3bn) project is described as a 1GW data campus totaling six four-story data center buildings spanning 400,000 sqm (4.3 million sq ft). Within the planning documents, the site is described as having capacity from 600MW up to 1GW across 24 data halls.

The 52-hectare site, at 706-752 Mamre Road in the Kemps Creek area of the city, would include 936 cooling units and 852 diesel back-up generators. Documents suggest the site would feature 7,488 cabinets for lithium-ion battery storage. At least one on-site substation is planned, and potentially up to four.

Australian real estate firm ISPT lists 706-752 Mamre Road on its website as Summit, a 52-hectare land holding masterplanned for 245,000 sqm (2.6 million sq ft) of industrial floor space, with construction forecast to commence in early 2026. The company acquired the site in July 2024 via its ISPT Core Fund.

The firm is developing a large industrial warehouse campus within the same park – known as the Kemps Creek Warehouse Logistics and Industrial Facilities Hub – which is described as separate from the data center proposals.

Microsoft signs ten year partnership with water tech firm Shayp

Microsoft has signed a ten-year partnership with Belgian water technology company Shayp to support the launch of its corporate water efficiency platform, Shayp 4IMPACT.

The initiative is designed to deliver verifiable and auditable water savings and will commence with deployments across schools and public buildings in Brussels and Paris, targeting more than 600 sites.

According to Shayp, its approach differs from more traditional watershed programs in that it deploys IoT-enabled monitoring technology within community infrastructure, allowing corporate partners to claim verified volumetric savings against their water stewardship goals. The company’s devices create digital twins of water consumption, detecting leaks and anomalies in real time, with sensitivity down to a trickle.

The platform builds on Shayp’s record of eight billion liters saved annually across more than 9,000 monitored buildings. The company claims users can achieve a return on investment in under a year, with an average 21 percent reduction in water bills.

Currently active in the Benelux region, France, Germany, and the UK, Shayp is expanding across Europe and exploring opportunities in water-stressed markets. Since its founding in 2017, the company said that it has saved an estimated 27 billion liters of water.

Water usage is a major concern amongst data center providers, due to the increasing demand for water associated with AI workloads. The company has undertaken several measures to reduce its water usage. At the end of last year, it announced plans to implement a zero-water evaporation design in its upcoming data centers.

The closed-loop system will first be piloted at its under-construction data centers in Phoenix, Arizona, and Mount Pleasant, Wisconsin, in 2026.

OpenAI announces Stargate UK, with up to 8,000 GPUs at Nscale data centers

OpenAI plans to deploy Stargate compute capacity in the UK. The generative AI company will take space in several Nscale data centers, initially "exploring" the uptake of up to 8,000 GPUs, with a potential to scale up to 31,000 GPUs.

OpenAI said that the sovereign capacity will allow its models to run on local computing power for specialist use cases such as public services, national security partnerships, and regulated industries like finance.

Nscale said that it will "significantly expand" its planned UK capacity for the contract, with the GPUs expected to be based across a number of sites, including a new 'AI Growth Zone' in Cobalt Park, in the North East. This week, Nscale separately announced a major contract with Microsoft to deploy 23,000 Nvidia Grace Blackwell Ultra GPUs at its planned Loughton, Essex, data center.

The company will also deploy 4,600 Nvidia GB300 GPUs as part of Nvidia's DGX Cloud, including for the DGX Lepton Marketplace.

In July, Nscale announced that it was working with Aker for a larger European Stargate project in Norway, featuring some 100k GPUs.

The news comes during US President Donald Trump's visit to the UK, and coincides with a flurry of data center-related investments in the country, including a £5bn ($6.9bn) Google deal, a $678m BlackRock venture, and a new Vantage facility. CoreWeave has also said it is investing £1.5 billion ($2bn) in UK AI data center capacity and operations.

Edged opens data center in New Albany, Ohio

Data center operator Edged has launched a data center outside Columbus, Ohio.

The company recently celebrated the opening of its newest data center in New Albany.

Located at 6526 New Albany Road East, the two-story facility delivers 24MW of capacity across 206,000 sq ft (19,140 sqm).

News that Edged was planning a data center in the area surfaced in late 2023, with the company breaking ground in August 2024.

Edged Columbus is fitted with a waterless cooling system from sister company ThermalWorks. The system supports up to 70kW per rack with air cooling and 200kW per rack with plug-and-play liquid cooling.

“This project would not have been possible without the support of the City of New Albany, who has been unwavering in its vision of bringing technology to the region in a thoughtful, sustainable way,” said Bryant Farland, president and CEO of Edged US. “This Edged facility is purpose-built for AI at scale and shows how advanced design can meet growing digital demand. With ultra-energy efficient systems and waterless cooling, we’re delivering reliable capacity while conserving resources and supporting the region’s economic future.”

The event was attended by New Albany Mayor Sloan Spalding, New Albany officials, local electric utility provider AEP, and project partners.

Microsoft is currently building in the New Albany area, with AWS and Google also having a large footprint around the Columbus area, including in Licking County’s New Albany. Meta also has a campus in New Albany.

OpenAI signs LOI to deploy 'at least' 10GW of AI data centers with Nvidia hardware, Nvidia to invest 'up to' $100bn in OpenAI

OpenAI has signed a letter of intent with Nvidia to deploy at least 10 gigawatts of the company's systems in its data centers.

At the same time, Nvidia will invest up to $100 billion in OpenAI as the new systems are deployed.

The first phase under the arrangement is expected to come online in the second half of 2026 using the Nvidia Vera Rubin platform.

Sam Altman, co-founder and CEO of OpenAI, added: “Everything starts with compute. Compute infrastructure will be the basis for the economy of the future, and we will utilize what we’re building with Nvidia to both create new AI breakthroughs and empower people and businesses with them at scale.”

OpenAI said that it will work with Nvidia as a preferred strategic compute and networking partner for its AI factory growth plans. The two companies will work together to co-optimize their roadmaps for OpenAI’s model and infrastructure software and Nvidia's hardware and software.

The deal comes despite the fact that OpenAI is developing its own AI chips internally, with mass production expected as soon as next year. It also uses Google's TPUs and AMD's GPUs.

The company is expected to spend some $350 billion on cloud services (as well as an additional $100bn for backup capacity), primarily on Oracle, but also across Microsoft and Google.