JetCool releases new CDU Can chill up to 300kW

• Liquid cooling firm JetCool has released a new coolant distribution unit (CDU).

• The SmartSense CDU is available as both an in-rack and modular unit.

• JetCool says the units each have capacity of up to 300kW. The system is scalable up to 2.1MW in row-level configurations.

• The company specialises in ‘microconvective cooling technology,’ which means that cooling jets route fluid directly at the surface of a chip instead of passing fluid over it.

• JetCool’s portfolio also consists of air cooling solutions and cold plates.

• It is owned by American electronics manufacturing services and original design manufacturer Flex. The acquisition was made in November 2024. Terms of the deal were not announced.

Nvidia working with data center partners to build 800V HVDC power systems

• Nvidia is collaborating with data center equipment firms to develop new power systems to support incoming 1MW+ racks.

• Using 800V high-voltage direct current (HVDC), this new architecture will centralize power distribution in AI data centers, allowing power to be converted from AC to DC at the GPU level, within the server board, minimizing energy losses.

• Partners for the project include Infineon, STMicroelectronics, and Texas Instruments for the silicon; Delta and Flex Power for the power system components; and Eaton, Schneider Electric, and Vertiv for the data center power systems.

• The company said its 800V HVDC architecture addresses these challenges through a “holistic redesign,” removing the need for the multiple voltage conversions that take place at various stages between the grid and rack and freeing up “valuable space for more compute resources, allowing for higher-density configurations and improved cooling efficiency.”

• According to Nvidia, its approach could result in up to five percent improvement in end-to-end power efficiency and reduce maintenance costs by up to 70 percent, power supply unit failures, and lower labor costs for component upkeep.

• Cooling costs could also be reduced by removing AC/DC power supply units from inside racks.

OpenAI to open first international Stargate data centre in UAE

• OpenAI has unveiled its first international deployment of Stargate, with plans to open a 1GW data centre in the United Arab Emirates (UAE).

• The project will bring OpenAI together with its Stargate partners G42, Oracle, Nvidia, Cisco, and SoftBank to power AI workloads in the UAE, with the cash-rich country also investing in the project.

• Stargate UAE aims to go live in 2026 with an initial 200MW deployment. No timeframe was provided for when the full 1GW would go live.

• While the UAE was first in the pipeline for the OpenAI for Countries project, the team behind the project want 10 facilities total.

• In a bid to find its next sites, OpenAI’s chief strategy officer, Jason Kwon, will begin a road trip across Asia Pacific next week with plans to meet governments and potential private-sector partners.

Crusoe secures $11.6bn in debt and equity for OpenAI's Stargate data center campus in Abilene, Texas

• Crusoe has raised the second phase of its $15 billion joint venture to fund a 1.2GW AI data center campus in Abilene, Texas.

• The company secured $11.6bn in debt and equity for the site, with the funds managed by Blue Owl’s Real Assets platform along with Primary Digital Infrastructure.

• The campus is the flagship site for OpenAI's Stargate effort, a $500bn project to build a number of massive data centers across the US for the AI developer.

• Each of the eight buildings will operate up to 50,000 Nvidia GB200 NVL72s on a single integrated network fabric, Crusoe said. The data centers will use direct-to-chip liquid cooling via zero-water evaporation cooling systems that recirculate water through a closed loop system.

• The campus is itself part of the Lancium Clean Campus, owned by Lancium.

Flexential partners with UPenn for Nvidia DGX B200 SuperPOD deployment at Collegeville data center

• US data center operator Flexential has announced the phased deployment of an Nvidia DGX SuperPOD to support the launch of the University of Pennsylvania's (UPenn) new Penn Advanced Research Computing Center (PARCC).

• Comprising DGX B200 systems and an Nvidia Quantum-2 InfiniBand network solution, the system will be housed at Flexential's data center in Collegeville, Pennsylvania.

• In a statement, Flexential said the deployment would support PARCC's high-density infrastructure requirements, adding that the facility had been designed to meet the performance and efficiency demands of large-scale AI and high-performance computing workloads and certified as part of the Nvidia DGX-Ready Data Center program.

• Enterprise IT services provider Ahead was responsible for the off-site rack integration and on-site cabling, delivering pre-configured racks to the facility in order to reduce setup time and installation complexity.

• Located 25 miles northwest of UPenn's campus in Philadelphia, Flexential's Collegeville data center offers 7.2MW of capacity across the 203,705 sq ft (18,925 sqm) site. The company said the facility has built-in liquid cooling capabilities and maintains 100 percent service-level agreements on power, cooling, network, and bandwidth, and delivers more than 1,500 watts per square foot.

• GI formed Flexential in 2017 after merging Peak 10 and ViaWest. GI acquired Peak 10 back in 2014. Morgan Stanley is also an investor. Today, Flexential operates 40 data centers across 18 US markets and has more than 330MW of built and under-development capacity.

Saudi's DataVolt signs $20bn deal with Supermicro

• Saudi Arabian data center company DataVolt has signed a $20 billion agreement with Super Micro Computer (Supermicro).

• The agreement, described as a multi-year partnership, will see Supermicro delivering GPU platforms and rack systems to DataVolt's "hyperscale AI campuses" in Saudi Arabia and the US.

• Supermicro CEO Charles Liang added: “We are excited to partner with DataVolt and continue expanding our manufacturing efforts in the United States. By working together, we will bring cutting-edge AI and compute infrastructure, enabling the Kingdom’s vision of becoming a global hub for technology and innovation.”

• This year so far has seen DataVolt announcing plans for two new data centers in Saudi Arabia - a facility in Riyadh, and another 1.5GW campus in Neom's Oxagon.

• The company has committed to investing more than $5 billion in developing around 300MW of data center capacity in the country, and previously announced plans for data centers in Dhaka, Bangladesh, and Uzbekistan.

• While the committee reported in December 2024 that it found “no evidence of fraud or misconduct” on the part of management or the board, it did recommend Supermicro replace its CFO, a process which the company said it has started, with current CFO David Weigand retaining the role until a new candidate is found.

Nvidia launches NVLink Fusion to connect custom CPUs and ASICs with Nvidia hardware

• Nvidia has opened its NVLink interconnect technology up to external companies, with the launch of NVLink Fusion.

• Unveiled this week at Computex in Taiwan, the new offering allows customers to build semi-custom AI infrastructure that links non-Nvidia accelerators with Nvidia hardware in rack-scale solutions.

• Cloud providers will also be able to use NVLink Fusion to scale up AI factories to “millions of GPUs,” using any ASIC in combination with Nvidia’s rack-scale systems and networking platform.

• Nvidia’s NVLink provides high-performance interconnect technology that sits between the GPU and CPU, enabling much faster data exchange.

• The NVLink platform, which is currently in its fifth generation, provides 800Gbps of throughput and features ConnectX-8 SuperNICs, Spectrum-X, and Quantum-X800 InfiniBand switches, with support for co-packaged optics to be made available soon. It can also provide a total bandwidth of 1.8Tbps per GPU, 14x faster than PCIe Gen5, Nvidia said.

• Its first GPU interconnect specification was released in April 2025, providing 200Gbps throughput per lane and linking up to 1,024 accelerators per pod.

• NVLink Fusion silicon design services and solutions are available now from MediaTek, Marvell, Alchip, Astera Labs, Synopsys, and Cadence.

Nvidia and Foxconn plan Taiwanese AI supercomputer

• Nvidia and Foxconn are working with the Taiwanese government to build an AI supercomputer.

• Announced by Nvidia at Computex in Taiwan, the supercomputer will consist of 10,000 Nvidia Blackwell Ultra GPUs, with Big Innovation Company – a Foxconn subsidiary and Nvidia cloud partner – providing the AI infrastructure.

• The Blackwell systems featured in the supercomputer will include the GB300 NVL72, connected by NVLink, Nvidia Quantum InfiniBand, and Nvidia Spectrum-X ethernet networking.

• According to Nvidia, the Taiwan National Science and Technology Council will use the system to provide AI cloud computing resources to industries across Taiwan. Researchers from TSMC will also have access to the supercomputer, while Foxconn will use it to “accelerate automation and efficiency across its three core pillars — smart cities, electric vehicles, and manufacturing.”

• Young Liu, chairman and CEO of Foxconn, added: “Foxconn builds technology that underpins modern life, and now, we’re building computing infrastructure to scale the next generation of breakthroughs across Taiwan.

Fire breaks out at Oregon data center leased by Elon Musk's X

• A fire broke out on Thursday, May 22, at a data center in Hillsboro, Oregon, that is leased by Elon Musk's X social media firm, causing a major global outage.

• As reported by Wired, firefighters arrived at the Hillsboro Technology Park at 10:21 am to deal with the outbreak.

• According to the Wired report, the fire was related to a room of batteries. Hillsboro Fire and Rescue spokesperson Piseth Pich said that the fire had not spread to other parts of the building, but the room was full of smoke.

• According to Wired, X leases space in a data center building "linked" to Digital Realty.

• In a written statement to local publication Oregon Live, Digital Realty said of the fire: "The incident is under control, and the fire department has concluded its on-site response. All personnel were safely evacuated, with no reported injuries.

• Digital Realty declined to say whether X leases space in the facility directly affected by the fire.

• In late 2022, Musk shut down the Sacramento data center, moving as many as 2,573 server racks to data centers in Portland and Atlanta. Around that time, X suffered a seven-hour outage. Tesla later took over the lease at the site.

• Digital Realty's Portland, Oregon, data centers have a total footprint of 56,500 sqm (609,000 sq ft) split between two data centers.

GlobalWafers opens Texas wafer fab in Sherman, plans $4bn additional US investment

• Taiwanese chipmaker GlobalWafers has announced plans to invest an additional $4 billion in the US.

• The comments were made by Doris Hsu, GlobalWafers chairperson and CEO, at the opening of the company’s new $3.5bn fab in Sherman, Texas.

• GlobalWafers is set to receive $406 million in direct funding from the US government's CHIPS act to support the construction of two fabs, the newly opened facility in Sherman, Texas, and a second in St. Peters, Missouri, both of which will produce 300-mm wafers and silicon-on-insulator wafers.

• The company is also planning to convert its existing silicon epitaxy wafer manufacturing facility in Sherman into a silicon carbide (SiC) epitaxy wafer manufacturing site, producing 150mm and 200mm SiC epitaxy wafers.

• The 142-acre campus in Texas can accommodate up to six phases. No timeline has been set for these additional phases.

• Following that announcement, in March, TSMC announced that it would invest $100 billion into the US - with it unclear how much of the previous figure is included in the number - after President Trump criticized Taiwan for 'stealing' America's chip business, said that the CHIPS Act was a failure, and threatened tariffs.