Data Center Networking stories
Enterprises face rising pressure to secure and automate AI workloads as HPE ties networking, storage and software into a fuller stack with NVIDIA.
The additions broaden access to mixed quantum systems as OVHcloud seeks to build Europe's quantum infrastructure around its sovereign cloud.
The move could help hyperscalers cut cabling in dense AI clusters by half as optical links become central to NVIDIA's custom-chip strategy.
AI data centres are hitting copper limits, pushing Marvell and Nvidia towards optics as clusters grow larger and more distributed.
Demand for AI clusters is pushing networking hardware to the limit, and Marvell says its new chip could cut rack power use and latency.
The move could help ease bottlenecks as huge AI clusters struggle to keep hundreds of thousands of GPUs synchronised during network faults.
The USD $15 million order signals early commercial traction as the partners target lower-power, faster links between chips in AI data centres.
Its new fabric promises lower latency and more bandwidth for training, as Google links up to 134,000 TPU 8t chips across sites.
Customers will soon get faster, cheaper AI training and inference as Google Cloud adds new TPUs, GPU instances and networking.
The funding backs a push into AI data centres, where better network control could lift model utilisation and cut token costs.
F5 deepens Nvidia AI partnership, tying BIG-IP Next to BlueField-3 DPUs to boost token throughput and slash real-time inference costs.
Nvidia deepens its OpenNebula tie-up to automate multi-tenant 'AI factories', unifying GPUs, DPUs and networking under one control plane.
AMD is pitching EPYC chips as the orchestration engines for “agentic AI”, arguing CPUs must steer ever-busier accelerators in data centres.
Akamai debuts NVIDIA-powered AI Grid to run low-latency inference across 4,400 edge sites, extending its distributed Inference Cloud.
Equinix launches a global distributed AI hub and Fabric Intelligence layer to unify, govern and optimise multi-cloud AI infrastructure.
The UK lab trial aims to slash AI power use and cut accelerator idle time as demand for inference strains data centre networks.
Hardware-enforced tenant separation will help Visionbay scale Taiwan's largest GPU cluster without risking customer data leakage or network outages.
Federal buyers of AI and HPC systems will gain wider access to Cornelis’s CN5000 platform through new partners ASI Corp, CTG Federal and TVAR Solutions.
NBN Co and Nokia hit 230 Gbps on existing FTTP in a world-first trial, showing multiple optical technologies can share the same fibre.
Xscape Photonics raises $37m to double valuation and launch FalconX, an eight-wavelength redundant laser for AI data centre networks.