Google added two systems to its eighth-generation TPU lineup: TPU 8tx, a confidential computing system built on TPU 8t silicon, and TPU 8i Max, which puts two TPU 8i chips on a single package for long-context inference. They join the TPU 8t training chip and TPU 8i inference chip Google introduced at Cloud Next in
10ZiG Technology announced support for Amazon DCV on its thin clients and on PCs converted with its RepurpOS software, adding a remote display protocol built for graphics-intensive Windows and Linux workloads. 10ZiG is pitching the support at organizations moving off HP Anyware, which the company says is approaching end of life, and endpoints running DCV
Broadcom made three Kubernetes announcements at KubeCon North America 2026: its VM API project has been accepted into the CNCF Sandbox, it is co-chairing a newly formed Workload Conformance Working Group in the Kubernetes community, and more than 20 partner solutions have been validated to run on VMware vSphere Kubernetes Service (VKS), the Kubernetes runtime
IBM and Microsoft announced Imagine Enterprise, a platform for building, running, and governing AI agents that carry out industry-specific business processes. IBM Consulting contributes more than 6,000 prebuilt AI agents and assets built on IBM Consulting Advantage, its AI-powered delivery platform, along with the experience of building and running them. Microsoft provides the AI, data,
Druva announced Druva Agent Recovery, a set of AI Resilience capabilities that back up custom AI agents and restore them when they are deleted or unexpectedly changed. Support starts with Microsoft Copilot Studio, Microsoft’s platform for building custom agents. Druva is also widening its Power Platform coverage with Power Apps and Power Automate alongside existing
IBM has announced Nighthawk r3, its next quantum processor, designed to run 10,000 quantum operations accurately across 520 programmable qubits. The chip carries 2,034 physical qubits and adds 10 chip-to-chip interconnects, and IBM plans to link two of them into a single 1,040-qubit system in 2027. That is 4.3 times the programmable qubit count of
KIOXIA America and Dell Technologies have put 157PB of raw flash in a single 42U rack, using 640 KIOXIA LC9 Series 245.76TB SSDs across 16 Dell PowerEdge R7725xd servers running software-defined Dell ObjectScale. KIOXIA calls it the first 157PB object storage configuration in an industry-standard rack, a claim it bases on its survey of the
HPE has added hybrid flash and hard drive configurations to its Cray Supercomputing Storage Systems E2000, pairing SSDs with new 32TB SAS-4 HDDs in new SAS-4 enclosures that hold 74 or 100 drives. HPE rates a rack at up to 116GB/s of read throughput and 92GB/s of write throughput and calls the result the fastest
Lenovo has introduced two ThinkSystem V5 servers built on AMD’s 6th Gen EPYC 9006 SP7 processors: the SC755 V5, a Neptune liquid-cooled dual-socket node for HPC clusters, and the SR755 V5, an air-cooled single-socket 2U rack server. Announced ahead of SC26, the launch also covers the ThinkSystem DE6500 and DE6800 NVMe block arrays, validation of
Cornelis and AMD have launched a joint program to design, validate, and deploy HPC and AI-for-science systems that pair AMD EPYC CPUs and Instinct GPUs with Cornelis networking, sold through the OEMs and system integrators customers already use. The announcement is anchored by a production system: GWDG’s new EMMY P4 supercomputer in Göttingen, Germany, runs
Amazon