Volume 343, Issue 3 · 6 articles from 5 issues
Vertiv's 2.3 MW CoolChip coolant distribution unit becomes the first CDU qualified as NVIDIA DSX Ready.
Vertiv announced that its 2.3 MW CoolChip coolant distribution unit (CDU) has been qualified as NVIDIA DSX Ready for AI factory infrastructure, the first CDU to achieve this specific qualification.
The designation lets AI facility builders evaluate cooling solutions against NVIDIA's DSX requirements with confidence.
Vertiv said the qualification reflects its focus on integrated data center infrastructure, where power, cooling, water, site, and grid constraints increasingly shape what builders can deploy.
The CDU manages liquid cooling for high-density computing systems across edge, colocation, and hyperscale AI environments.
Vertiv outlines five implementation priorities that decide whether bring-your-own-power data center projects keep their time-to-power edge.
Grid constraints and surging AI power demand are pushing data center developers toward bring-your-own-power (BYOP) generation, but Vertiv argues that choosing BYOP is only the start.
A new Vertiv paper sets out five success factors: early coordination with utilities and authorities, co-design of generation and facility infrastructure, early modeling and validation, prefabricated solutions, and a defined operating model.
It warns that AI training loads can cause rapid power swings that onsite generators struggle to follow, which, if unmanaged, risk instability, nuisance trips and control conflicts with UPS and battery systems.
As AI workloads increase power density, there is a growing need to evaluate the effectiveness of traditional air-cooling systems. Liquid cooling offers an efficient approach to managing the thermal requirements of high-density racks.
Our white paper, "Deploying liquid cooling in the data center," provides a clear roadmap for integrating liquid cooling into existing or new facilities.
Key topics include: - Step-by-step planning for liquid cooling implementation. - Technical considerations for retrofitting current infrastructure. - Performance metrics and operational benchmarks.
Adapt your cooling systems to meet the evolving requirements of AI workloads.
Get the white paper!
A Vertiv white paper offers a structured framework for deciding whether bring-your-own-power generation improves a data center project.
With grid connection waits exceeding a decade in some U.S. markets, interest in dedicated onsite or adjacent power generation for data centers is rising.
Vertiv cautions that BYOP is not a shortcut, since fuel, permitting, emissions, water, equipment lead times and economics create new constraints.
Its framework starts by assessing utility service and alternatives such as phasing, relocation, colocation and cloud capacity, and if BYOP remains viable, developers define its intended use, separate ownership and commercial decisions, and compare technology and fuel options iteratively.
Vertiv released a workload-based guide to help enterprises decide where AI workloads should run and what infrastructure they need.
Vertiv introduced the Vertiv Designing Infrastructure for Enterprise AI Guide, developed with research firm STL Partners and announced alongside Deloitte's Center for Integrated Research, giving business, IT and facilities leaders a workload-based framework for moving AI from experimentation to production.
It starts with the intended business outcome and AI workload, then evaluates deployment location, infrastructure requirements and facility readiness.
Organizations can map each workload to right-sized infrastructure models and reference designs to reduce deployment risk.
Vertiv and NVIDIA unveil an AI factory approach pairing reference designs with factory-built infrastructure, deployed by Polar in Norway.
At NVIDIA GTC Berlin, Vertiv and NVIDIA introduced an AI factory approach that pairs rapid configuration and virtual validation of pre-engineered reference architectures with factory-built deployment.
Vertiv says the approach can cut configuration and validation time up to 10x and, under defined assumptions, generate up to 30% more cumulative token output in the first 12 months.
It combines the NVIDIA Vera Rubin DSX AI Factory reference design and Omniverse DSX Blueprint with Vertiv SimReady power and cooling assets, integrated with Dassault Systemes' 3DEXPERIENCE platform.
Polar Data Centers is applying the approach to the first 41 MW phase of its AI data center at Heroya, Norway.