Company also makes UltraServers with Nvidia GB300s available, and teases Trainium4

Initiative dubbed 'Manhattan Project 2.0' but unlike its predecessor, hinges on mobilizing private sector investment

The likes of OpenAI, AMD, and Cornelis Networks are taking part in what is really a modernization effort – and that's okay

The system is designed to handle data-hungry AI applications

Powered by hardware-accelerated RoCEv2, the CN6000 SuperNIC doubles the performance of its predecessor

The vendor is a big Ethernet backer

High-performance computers might be able to lighten the load

Neoverse integration means custom hardware from AWS, Google, and Microsoft can integrate with Nvidia GPUs

DRAM becomes a hot commodity as inference needs increase

Announces HLRS and LRZ will deploy the platform for Herder and Blue Lion systems

Kubernetes-based platform simplifies orchestration across thousands of GPUs

Azure infrastructure buildout goes distributed as OpenAI inference costs spiral

IPO suggestions follows refuted takeover by local IT giant

Deal underscores chip giant's push to fuse software and hardware to meet inference demand

Offering turns complex orchestration needs into simple Kubernetes pods

Vendor will underpin neocloud infrastructure serving Meta, OpenAI, and Microsoft workloads

Move brings AI-SuperNIC technology in-house as Arm looks to capitalize on surging data center demand

Claude developer Anthropic spoilt for inferencing choice as Ironwood adds to AWS Rainier options

MLPerf Inference tests see the new Azure ND GB300 v6 VMs achieve token performance that ‘fundamentally alters the calculus of AI efficiency’

Comes in both 256 and 9,216 chip configurations

The new offerings include the first data center APU for HPC and AI workloads

Deployment to support LANL's generative AI work

A huge generative AI system not built with GPUs or TPUs

TPU v5e doubles the performance of the company's in-house AI chip