Search

Vertiv Simplifies High-Density AI Infrastructure Deployment Across Data Center Environments | Stockwhiz AI

Vertiv offers a platform that supports NVIDIA MGX and OCP ORV3, usable in existing or new data centers, scalable from one rack to multiple racks without redesign, and available globally now.

AI vs traditional applications in data centers / Habr

Translated title. Original: ИИ против традиционных приложений в ЦОД / Хабр

The article compares AI data-center networks with traditional web-service networks, showing that both must be supported under one roof. It explains how AI training traffic is tightly synchronized and sensitive to the slowest messages, while traditional applications favor scalable, fault-tolerant TCP-based flows; misaligning these needs wastes infrastructure or hurts AI performance. It discusses traffic profiles, conflict points, and architectural criteria for network design in hybrid centers.

NVIDIA and Cisco Unveil USFP Optical Technology for Next-Generation AI Factories - TechAfrica News

NVIDIA and Cisco unveil USFP, a high-density 400Gb/s-per-lane optical module for AI factories, designed to increase compute scale within existing networks while reducing power usage. The partners, with TE Connectivity and Amphenol, plan an open specification to foster an ecosystem and ensure compatibility with NVIDIA Spectrum-X and Cisco Silicon One networks.

Cisco Expands AI Networking Portfolio with New 400G Switches

Cisco announces AI-friendly data center networking upgrades, featuring the Silicon One G302 51.2 Tbps switch and NVIDIA-backed 400G USFP optics, aimed at higher bandwidth, better power efficiency, and an open ecosystem for AI-scale deployments.

SuperComputing (SC26)

Vertiv invites visitors to Booth 2028 at SuperComputing (SC26) to see data center infrastructure for AI and high-performance computing, from individual racks to full environments.

Quantum computing company’s first breakthrough is in AI data centers | Semafor

The article discusses how data-center performance is limited by copper interconnects in AI training and how optical networking could overcome energy and heat inefficiencies. It highlights PsiQuantum’s device that routes light between racks without electrical conversions, potential benefits for latency and density, and mentions Google’s use of optical switching and industry moves toward standardization and investments in photonics.

For Vera Rubin: NVIDIA reduces specifications for HBM4 speed and NVLink interconnect – Hardwareluxx

Translated title. Original: Für Vera Rubin: NVIDIA reduziert Vorgaben für HBM4-Geschwindigkeit und NVLink-Interconnect – Hardwareluxx

NVIDIA reportedly lowered Vera Rubin system specifications for HBM4 and NVLink as AMD gains ground in memory and interconnect bandwidth, while NVIDIA relies on its ecosystem and integration to maintain a competitive edge.

This Week's Top Five Stories in the Data Centre Industry | Data Centre Magazine

Arelion completed a live field trial on the Amsterdam–London route using Nokia optics to assess extra capacity on existing fibre, targeting wholesale, neoscaler, and enterprise customers as they deploy AI and bandwidth-heavy workloads, as part of its push to strengthen its top-ranked global Internet backbone.

NVIDIA and Cisco introduce new pluggable optics form factor to scale AI factories | TahawulTech.com

NVIDIA and Cisco unveil Ultra-Dense Small Form Factor Pluggable (USFP), a 400Gb/s-per-lane optical module to scale AI factories, boosting bandwidth and efficiency, with industry partners and an open-spec collaboration planned to standardize the approach.

Tata Communications to Invest ₹4,350 Crore in 10,000-km AI Super Highway - Construction Mirror

Tata Communications plans a ₹4,350 crore investment to build about 10,000 km of AI-ready fibre, the AI Super Highway, to connect data centres across India with high-speed, low-latency links. The three-phase network expands from the Mumbai-Pune-Hyderabad-Chennai corridor to additional locations, adding ~6 Pbps of capacity by FY2036, funded by cash flow and linked to a global backbone for cross-border data movement. The project targets hyperscalers, neoclouds, AI platforms, GCCs, and enterprises, aiming to support large AI workloads and growing data centre demand.