IBM has entered a definitive agreement to acquire HRL Laboratories, LLC, a private research and development organization jointly owned by Boeing and General Motors. Financial terms were not disclosed. The transaction is expected to close by the end of the third quarter of 2026, subject to customary conditions and regulatory approvals.

Following the acquisition, Boeing and General Motors are expected to continue working with IBM on quantum applications and the development of advanced technologies.

The deal adds HRL’s silicon spin-qubit research to IBM’s quantum computing portfolio. IBM’s current quantum systems are based primarily on superconducting qubit architectures, while HRL has developed expertise in silicon-based spin qubits, quantum sensing, communications, electronics, manufacturing, and materials science.

Silicon Spin Qubits Add a Second Quantum Hardware Path

IBM is pursuing large-scale, fault-tolerant quantum computing through superconducting qubits. Its Quantum Starling system is planned for 2029, which IBM says will be 20,000 times more powerful than today’s quantum computers and capable of running 100 million quantum operations. The company’s longer-term roadmap includes the Blue Jay system, projected to handle 1 billion quantum operations in the mid-2030s.

HRL brings experience in silicon spin-qubit technology, an alternative hardware approach using electron spin states in semiconductor structures as quantum bits. Both superconducting and spin-qubit designs depend on advanced fabrication, cryogenic infrastructure, control electronics, interconnect technologies, and packaging. Silicon spin qubits are of particular interest because they may offer a path to leverage established semiconductor manufacturing techniques for future scaling.

IBM indicated that HRL’s work could help broaden its research into practical approaches for scaling quantum hardware over the next decade. The acquisition also gives IBM deeper capabilities in supporting technologies required for quantum systems, including cryogenics, qubit control, chip packaging, and high-speed communications.

Quantum Sensing and Materials Research Broaden Scope

Beyond quantum computing, HRL’s research portfolio includes quantum sensing and quantum materials. Quantum sensors can measure small changes in magnetic fields, motion, time, and other physical phenomena with high precision. Potential applications include life sciences, navigation, defense, industrial instrumentation, and scientific research.

IBM expects HRL’s work in materials science and semiconductor technologies to contribute to future generations of quantum devices and sensing platforms. New materials may improve qubit performance, sensor sensitivity, device reliability, and manufacturability across several emerging technology categories.

HRL also brings decades of development work for commercial and U.S. government customers in high-power and high-speed communications, electronics, advanced manufacturing, sensors, and materials science. IBM is positioning these capabilities as relevant to broader computing, communications, and mission-critical technology programs.

Potential Link to IBM’s Quantum Manufacturing Efforts

IBM stated that HRL could also create opportunities for closer collaboration with Anderon, the quantum wafer foundry it announced in May 2026 and described as the world’s first pure-play quantum foundry. Anderon is being established as a standalone IBM company with support from the U.S. Department of Commerce, aiming to provide scalable and repeatable quantum hardware manufacturing across multiple quantum computing modalities.

Adding HRL’s spin-qubit expertise could support future research into scalable spin-qubit fabrication and shorten development cycles between quantum-device research and manufacturing.

The post IBM Buys HRL Labs From Boeing and GM, Adding Silicon Spin Qubits to Its Quantum Roadmap appeared first on StorageReview.com.

The Korea-focused AI Summit in San Francisco on July 24 gathered leaders from NVIDIA, OpenAI, Broadcom, Anthropic, Samsung, SK Group, Hyundai Motor Group, and NAVER. Held during South Korean President Lee Jae Myung’s visit, the event produced announcements spanning AI factory capacity, HBM supply, semiconductor manufacturing, agentic AI research, and robotics infrastructure.

These announcements point to continued investment in Korea’s AI ecosystem, especially in sovereign AI capacity, faster computing, advanced memory, and real-world AI applications.

SK Group and NVIDIA Target 2GW of AI Infrastructure

SK Group and NVIDIA signed letters of intent covering AI infrastructure, including building AI factories and supplying next-generation memory. The companies said the potential partnership could be worth more than $500 billion, but did not share details about project timelines, commercial terms, or committed investments.

The first step is a planned 2GW AI cloud project in Korea with SK Telecom. This setup will use NVIDIA’s DSX AI factory design, Vera Rubin accelerated computing, and SK hynix HBM4 memory. The first AI factory is expected to be ready by 2027.

NVIDIA DSX brings together fast computing hardware, software, facility design, and partner technologies into a complete AI factory system. SK Group and NVIDIA plan to use this setup to support different types of AI work in South Korea and the wider Asia-Pacific region.

SK hynix will also start a long-term AI memory partnership with NVIDIA. Together, they plan to develop and supply new high-bandwidth memory to train large language models, run AI, and support both agentic and physical AI tasks.

Samsung and Broadcom Expand Memory and Foundry Cooperation

Samsung Electronics and Broadcom signed an agreement to work together on memory supply, foundry manufacturing, and advanced packaging. They expect this partnership to generate over $200 billion in value by 2030.

Samsung will provide high-bandwidth memory for Broadcom’s upcoming AI accelerator products. The agreement also includes the use of Samsung’s 2nm and smaller process technologies for Broadcom products, such as wireless broadband chips.

They also plan to explore advanced packaging methods using Samsung’s 2nm process, such as 2.5D and 3D integration. These new methods are important for AI and networking chips, where traditional designs limit performance, memory speed, interconnects, and energy efficiency.

NAVER, NVIDIA, and Brookfield Expand AI Factory Plans

NAVER, NVIDIA, and Brookfield shared plans to expand their NVIDIA DSX AI factory at NAVER’s GAK Sejong data center in Sejong, South Korea. The planned capacity has grown from 55MW, announced in June, to 200MW by 2028, a buildout expected to comprise roughly 100,000 GPUs.

The planned 200MW facility will use NVIDIA Vera Rubin and Blackwell platforms. NVIDIA DSX MaxLPS software will help improve token processing per megawatt. DSX OS will handle system management, automation, reliability, and support for multiple users.

The expansion is structured as a roughly $10 billion project. Brookfield has signed a nonbinding term sheet to serve as an exclusive capital partner, funding up to $9 billion. NVIDIA plans to invest $1 billion, subject to customary conditions, and NAVER will fund the remainder. NAVER also plans to deploy up to 1 GW of NVIDIA AI infrastructure.

NAVER also plans to use this infrastructure to develop open AI models. The company is growing its HyperCLOVA X model family with NVIDIA Nemotron 3 Ultra open models and its own training data. NAVER aims to launch an AI agent platform in Korea later this year, leveraging parts of the NVIDIA Agent Toolkit, including NemoClaw blueprints.

Separately, the company is building a Seoul World Model using its street-view and spatial data, along with NVIDIA Cosmos world foundation models. This project focuses on physical AI and spatial reasoning uses.

Hyundai Outlines Physical AI Development Strategy

At the summit, Hyundai Motor Group shared its physical AI strategy, which encompasses smart vehicles, robots, AI-managed factories, and plans for connected-city infrastructure.

Hyundai is using its manufacturing, mobility platforms, robotics, and data as the base for a system that can train and improve AI in real-world settings. The company mentioned its partnerships with NVIDIA, Waymo, Boston Dynamics, and Google DeepMind as key parts of its physical AI ecosystem.

One main project is the Hyundai and NVIDIA Robot Reference Platform. This platform will give universities, research groups, and startups a standard hardware and software setup for building and testing physical AI systems. Hyundai hopes this will make robotics development more accessible and speed up the introduction of new products to market.

Hyundai also announced plans for the Saemangeum AI Valley in Jeonbuk. The project will include AI data centers, robotics manufacturing, electrolyzer plants, and AI-powered hydrogen city infrastructure. Hyundai plans to invest about KRW 9 trillion, or around $6.1 billion. The robotics cluster will make Hyundai robotics products and offer manufacturing services to smaller companies.

Hyundai also plans to invest KRW 42 trillion, or about $28.5 billion, over the next ten years in industrial hubs in the Yeongnam region. These centers will focus on AI-powered manufacturing, aerospace, and green energy infrastructure.

NVIDIA and KAIST Establish Agentic AI Research Lab

NVIDIA and the Korea Advanced Institute of Science and Technology (KAIST) announced a joint AI research lab at the KAIST Kim Jaechul Graduate School of AI in Seoul, described as NVIDIA’s first joint AI research lab with a university in Asia. The lab will focus on agentic AI research and Korean-language models for academic, business, and national uses, led by NVIDIA researcher Hyunwoo Kim, who joins the KAIST faculty in August.

The partnership is worth $300 million, with up to $50 million per year in computing resources for the first five years. Local NVIDIA Cloud Partners will provide access to NVIDIA AI infrastructure for researchers involved in the project.

The program will support at least 10 KAIST researchers each year with funding and NVIDIA internships. NVIDIA also plans to hire Korean researchers for full-time roles. The research will use NVIDIA’s full AI platform and Nemotron open models to help move academic work into real-world AI projects.

The post Korea AI Summit: SK and NVIDIA Float a $500B Partnership as NAVER Triples Its AI Factory appeared first on StorageReview.com.

ASUS has announced a new server portfolio based on 6th Gen AMD EPYC 9006 processors, built specifically around the efficiency-optimized SP8 socket rather than the 256-core SP7 flagship. The lineup includes dual-socket RS700A and RS720A systems designed for maximum compute density, alongside single-socket RS500A and RS520A systems targeting more space-constrained enterprise deployments.

The new platforms support PCIe Gen6 connectivity, advanced DDR5 memory configurations, and E3.S NVMe storage. ASUS is positioning the systems for enterprise AI inference, virtualization, simulation, cloud infrastructure, and storage-intensive workloads.

Dual-Socket RS700A/720A Targets High-Density Compute

The dual-socket RS700A/720A series is the higher-density option in the new portfolio. The systems support 32 DIMM slots and MRDIMM memory configurations, allowing the platform to address workloads requiring substantial memory capacity and bandwidth.

ASUS also specifies support for up to 32 E3.S storage bays in a 2U chassis. Combined with PCIe Gen6, this configuration is aimed at high-throughput environments that require dense NVMe capacity, accelerator connectivity, or large-scale network expansion.

The RS700A/720A systems are intended for AI inference infrastructure, technical computing, complex simulations, and enterprise applications where compute density and memory bandwidth are key design considerations.

Single-Socket RS500A/520A Focuses on Deployment Efficiency

The RS500A/520A series uses a single AMD EPYC 9006 processor socket and is designed for mainstream enterprise infrastructure and rack-constrained environments. ASUS specifies a chassis depth of less than 800 mm, which may simplify deployment in locations with limited cabinet depth or edge-oriented infrastructure.

Despite its smaller footprint, the single-socket series retains support for PCIe Gen6 E3.S storage. ASUS also notes shared modular components with the larger RS700A and RS720A platforms, which could simplify platform qualification, service processes, and component inventory for organizations standardizing across both system classes.

Platform Engineering and Serviceability

ASUS has incorporated its DC-MHS modular architecture into the new systems. The design separates I/O, high-power module, fan, and storage components into distinct chassis zones. This architecture is intended to simplify service access and provide a more modular platform design across server configurations.

The company also includes its DIMM.2 implementation, which places M.2 storage in the DIMM area rather than a conventional motherboard location. ASUS states that this placement uses the airflow available around the memory subsystem to improve M.2 SSD cooling and reduce the risk of thermal throttling without adding dedicated heatsinks.

Thermal Radar 3.0 with PID control provides dynamic fan-speed management. The system adjusts fan behavior in real time based on thermal conditions to balance component cooling requirements against system power consumption and acoustic output.

ASUS also highlighted tool-less serviceability features intended to reduce maintenance time and support faster component replacement. This approach is increasingly relevant in AI and HPC environments, where sustained utilization and infrastructure availability can make scheduling service windows difficult.

Availability

ASUS has not published pricing or a firm release date for the RS700A/720A and RS500A/520A series, saying only that the platforms are coming soon.

The post ASUS Builds Its EPYC 9006 Server Line on AMD’s Efficiency-Optimized SP8 Socket appeared first on StorageReview.com.

VAST Data has expanded its collaboration with AMD around AI infrastructure for cloud providers and enterprises deploying training, inference, retrieval-augmented generation, and agentic AI services. The effort combines the VAST AI Operating System with 6th Gen AMD EPYC processors, AMD Instinct GPUs, AMD networking hardware, and ROCm software.

The collaboration reflects a broader infrastructure shift from training-focused AI clusters toward environments that must also support persistent context, high-concurrency inference, and multi-turn agent workflows. These deployments place increased emphasis on data movement, KV cache management, GPU utilization, and the ability to deliver low-latency access to large model and context datasets.

VAST positions its Disaggregated Shared Everything, or DASE, architecture as the shared data layer for these environments. The platform combines file and object storage, databases, event streaming, and data services under a unified global namespace. It also provides multi-tenancy and workload isolation capabilities for AI clouds operating concurrent customer and application workloads.

EPYC 9006 Platforms for VAST CBox and EBox Systems

VAST has selected 6th Gen AMD EPYC processors, formerly codenamed Venice, for its next-generation CBox and EBox platforms. The company plans to use the processors in its sixth-generation CBox and third-generation EBox systems, which underpin the VAST AI Operating System.

AMD EPYC 9006 support introduces PCIe Gen6 connectivity to the VAST hardware platform. VAST states that the new interface doubles generational I/O bandwidth, improving file and object storage throughput while reducing latency for data services such as databases, data warehouses, and event-streaming workloads delivered through VAST DataBase and DataEngine.

For AI infrastructure, higher I/O bandwidth can help reduce bottlenecks between compute, network, and NVMe storage resources. This is particularly relevant for model loading, retrieval pipelines, checkpoint access, and externalized KV cache workflows where GPU memory capacity alone is insufficient to retain active context.

Reference Architecture Combines Helios, VAST, and DriveNets

VAST, AMD, and DriveNets are also developing an AI infrastructure reference architecture built around AMD Helios rack-scale AI infrastructure, the VAST AI Operating System, and DriveNets AI Fabric networking.

The reference architecture is intended to provide deployment guidance for model training, inference, reinforcement learning, and KV cache workloads. It includes sizing and availability considerations for organizations building AI factories that require shared data infrastructure and high-performance networking alongside GPU compute.

VAST also cited expanded collaboration with inference software providers TensorMesh and EmbeddedLLM. The ecosystem effort is focused on production inference architectures for agentic AI applications, although specific product integrations and availability details were not disclosed.

KV Cache Offload Targets High-Concurrency Inference

A central component of the announcement is expanded KV cache support using AMD Instinct GPUs, AMD Infinity Context, ROCm software, and the VAST AI Operating System.

KV cache data stores intermediate attention-state information generated during inference. Retaining this data improves performance for multi-turn interactions and long-context workloads, but large cache footprints can consume significant GPU memory. Externalizing or tiering KV cache to a high-performance storage platform can free GPU memory for active workloads while retaining context for subsequent inference requests.

VAST reported early testing with an AMD Instinct MI355X GPU that showed a 9x improvement in time to first token and 9.7x higher token throughput when using VAST for KV cache offload in high-concurrency agentic AI workloads. The company noted that these results depend on the hardware baseline, workload characteristics, and storage configuration.

The integration also applies VAST lifecycle policies to KV cache data. This capability is intended to automatically expire and delete cached information, which is relevant when the inference context contains sensitive, personal, or regulated data.

Pensando Pollara 400 Connects GPUs to Shared Storage

The architecture uses the AMD Pensando Pollara 400 AI NIC to connect AMD Instinct GPUs with the VAST storage cluster. According to VAST, the NIC supports GPU-to-storage data movement through NFS over TCP and RDMA, enabling access to NVMe SSD-based VAST clusters for KV cache and broader inference data requirements.

The resulting design is intended to support AI environments that need to scale context management independently of GPU memory. Rather than treating storage as a separate persistence layer, VAST is positioning the platform as a data and execution layer that can manage models, databases, streaming data, file data, and AI context across distributed infrastructure.

For AI cloud providers, the approach targets higher GPU utilization and improved operational efficiency as deployments move beyond GPU rental and batch training into persistent inference and agentic AI services.

The post VAST Data and AMD Claim 9x Faster Time-to-First-Token With KV Cache Offload on Instinct appeared first on StorageReview.com.

Wistron has opened its D1 AI smart facility in Fort Worth, Texas, marking the company’s first manufacturing site in the United States and the first U.S. production and assembly location for the NVIDIA GB300 Grace Blackwell Ultra Superchip. The new site represents a $700 million investment, spans roughly 324,000 square feet, and has created more than 500 jobs so far, a figure Wistron expects to reach 1,000 by the end of the year. It expands Wistron’s footprint in AI infrastructure manufacturing and adds domestic capacity for NVIDIA-based systems.

The facility opening was led by Wistron Chairman Simon Lin and NVIDIA CEO Jensen Huang, with local and Taiwanese government representatives also in attendance. For Wistron, the Fort Worth plant serves as a strategic extension of its global manufacturing network and a key U.S. node for advanced AI system production.

At launch, the site is producing the NVIDIA GB300 Grace Blackwell Ultra Superchip, with plans to add the NVIDIA Vera Rubin Superchip. Wistron characterized the plant as a high-tech manufacturing operation built to support next-generation AI infrastructure rather than conventional server assembly. The company said the Texas facility is intended to serve local customers while expanding domestic assembly and test capacity for advanced NVIDIA platforms.

The Fort Worth location also reflects a broader shift in how AI infrastructure is built and delivered. Wistron is using NVIDIA accelerated computing and software, including the Nemotron and Cosmos models, along with the Omniverse and Metropolis libraries, to support digital twin-based factory design, workflow simulation, and operational optimization. That approach is intended to improve plant layout, process flow, and efficiency as production scales.

Wistron selected Texas as a manufacturing hub based on its logistics position, workforce access, and advanced industrial base. By placing AI servers and superchip assemblies closer to U.S. end markets, the company is aiming to shorten delivery cycles, improve regional support, and strengthen supply chain resilience for AI infrastructure deployments. The company is also building the site as part of a broader operational ecosystem that includes manufacturing and after-sales support.

Addressing Large-Scale AI Deployment

NVIDIA framed the announcement within the larger expansion of AI factory infrastructure, noting that demand for large-scale AI deployment platforms continues to accelerate. The two companies said the Texas operation is part of a broader effort to rebuild advanced manufacturing capacity in the U.S. while supporting skilled job growth and the development of a domestic AI supply chain.

Energy planning and factory efficiency were also central to the launch. Wistron said the Fort Worth facility is built around digital manufacturing and local operations, with an emphasis on energy optimization and smarter control of plant power requirements. As advanced manufacturing demand increases in Texas, that level of visibility and flexibility is expected to play an increasingly important role in plant operations.

Top Location for AI Infrastructure Manufacturing

Simon Lin said the Fort Worth site will become a major location for AI infrastructure manufacturing in the U.S. over the next several years. He also emphasized that the pace of the AI market increases the importance of execution quality and responsible development alongside speed.

The company said the Fort Worth plant will act as the central hub for its U.S. manufacturing strategy, linking domestic operations with Wistron’s broader global production network and ecosystem partners. The investment is intended to deepen Wistron’s technical capabilities, improve supply chain efficiency and resilience, and position the company for the next phase of AI infrastructure growth.

The post Wistron’s $700M Fort Worth Plant Becomes the First U.S. Site Building NVIDIA GB300 Superchips appeared first on StorageReview.com.

Supermicro introduced its new H15 server portfolio, built around 6th Gen AMD EPYC 9006 Series CPUs. These servers are designed to work with AMD Instinct GPUs and AMD Pensando networking solutions. In addition, Supermicro is expanding its AMD GPU offerings with new PCIe GPU servers built for AMD Instinct MI350P GPUs and the previously announced Supermicro AMD Helios rack-scale platform, which incorporates AMD Instinct MI455X GPUs. This announcement includes details about the CPU platform, GPU and networking hardware, and the rack-scale system that integrates these components, targeting agentic AI, cloud, enterprise, and HPC workloads.

H15 Architecture and CPU Platform

The H15 architecture is based on Supermicro’s Data Center Building Block Solutions (DCBBS), which standardizes modular rack-scale deployment across its server range rather than designing each chassis individually. H15 systems support 6th Gen AMD EPYC 9006 Series processors with up to 256 cores and 512 threads per socket, which Supermicro says deliver up to 1.7x the generational performance of the prior lineup while offering more memory and I/O bandwidth. The company’s key message is simple: increased cores and memory bandwidth per socket let users run more AI agents or enterprise workloads per node without surpassing power limits, a critical factor as data center power availability often restricts rack capacity rather than compute density.

Six System Types Across the H15 Portfolio

Supermicro classifies the H15 lineup into six system categories, each tailored for specific deployment scenarios. The Hyper platform is the flagship dual-socket model, designed for enterprise applications, AI inference, virtualization, and cloud workloads, and features a thermal design capable of supporting top-tier AMD EPYC processors. CloudDC is available in both single- and dual-socket versions and adheres to the Open Compute Project’s Data Center Modular Hardware System (DC-MHS) standard, enabling interoperability with other OCP-compliant data center components rather than restricting users to a Supermicro-specific rack system.

GrandTwin is a high-density 2U, four-node system designed for scale-out workloads like object storage, virtualization, cloud services, and HPC. FlexTwin offers a more compact 1U, two-node, dual-CPU setup with liquid cooling, tailored for cloud-native and hyperscale environments where rack density and power efficiency take precedence over node flexibility. Supermicro’s Petascale Storage series rounds out its non-GPU portfolio with high-capacity all-flash platforms in 1U and 2U form factors, supporting up to 4.8PB per system for AI data lakes, large-scale analytics, and HPC storage. These platforms are built on software-defined storage rather than fixed hardware RAID.

The H15 8U 10-node SuperBlade is the rack-scale blade system in the portfolio, supporting single- and dual-socket blade configurations in air- or liquid-cooled options. It is designed for mixed CPU and GPU deployments used in HPC, AI inference, agentic AI, and enterprise computing, providing high blade density without requiring a full GPU rack.

New PCIe GPU Servers for AMD Instinct MI350P

Supermicro introduced two 5U PCIe GPU servers, the AS-5126GS-TNRT and AS-5126GS-TNRT2, featuring AMD Instinct MI350P PCIe GPUs. Each system can support up to 10 GPUs within a standard 5U air-cooled chassis, maintaining compatibility with data centers that already support such cooling and power configurations and avoiding the need for liquid cooling. The MI350P GPU provides up to 144GB of HBM3e memory and supports low-precision AI numeric formats, making it suitable for both inference and training workloads where memory capacity, rather than raw compute power, is often the limiting factor.

Open Ethernet Networking via Pensando Pollara 400

Supermicro’s MI350P systems utilize the AMD Pensando Pollara 400 AI NIC, an open Ethernet networking card designed for AI infrastructure. It manages front-end, storage, and scale-out traffic in MI350P deployments. Since it operates over standard Ethernet rather than a proprietary interconnect, Pollara 400 enables customers to scale their cluster from a single server to multi-rack setups without being locked into a closed networking stack. This flexibility is a key point that Supermicro and AMD highlight as a competitive advantage over interconnect options limited to a single GPU vendor’s ecosystem.

Supermicro AMD Helios: 72-GPU Rack-Scale Platform

The Supermicro AMD Helios Platform, created in partnership with AMD, is a liquid-cooled, 72-GPU rack system designed around AMD Instinct MI455X GPUs, 6th Gen AMD EPYC processors, AMD Pensando networking, and the AMD ROCm software suite. It targets large-scale AI training and high-throughput inference for organizations working with frontier-scale models and can scale from a single rack to larger multi-rack AI clusters. Helios and the new PCIe GPU servers serve different deployment needs: MI350P PCIe systems for enterprises expanding GPU inference or training within existing air-cooled setups, and Helios for organizations constructing dedicated, liquid-cooled AI clusters on a much larger scale.

Supermicro President and CEO Charles Liang explained that the H15 launch emphasizes building infrastructure tailored for performance, scalability, and efficiency, especially as agentic AI adoption increases. He highlighted the DCBBS architecture as key to achieving high performance while enabling flexible rack-scale deployment. Additionally, he mentioned Supermicro’s global services and U.S.-based supply chain as vital supports for customers deploying and scaling AI infrastructure.

AMD’s senior vice president and general manager of Compute and Enterprise AI, Dan McNamara, emphasized that enterprises scaling agentic AI require infrastructure that provides performance, efficiency, and deployment flexibility. He described how combining AMD’s EPYC CPUs, Instinct GPUs, and Pensando networking with Supermicro’s modular server and rack-scale designs can accelerate AI infrastructure deployment while enhancing resource utilization, energy efficiency, and total cost of ownership.

The post Supermicro H15 Servers Pair 6th Gen EPYC With MI350P GPU Systems and the Helios Rack appeared first on StorageReview.com.

AMD has outlined its role in two planned Oak Ridge National Laboratory supercomputing platforms, Lux and Discovery, positioned to support the U.S. Department of Energy’s Genesis Mission. The initiative brings together national laboratories, industry, academia, and other partners to apply AI, simulation, and scientific data to energy, discovery science, and national-security research.

Lux, under construction at Oak Ridge National Laboratory in Tennessee, is expected to become the first fully operational Genesis Mission platform. Funded projects are expected to begin using the system in October 2026. The system is intended to provide a unified environment for conventional high-performance computing workloads and cloud-native AI services.

Combining AMD GPUs, CPUs, and Networking

Lux will combine AMD Instinct MI355X GPUs, AMD EPYC CPUs, and AMD Pensando networking. AMD said the platform will support scientific workflows from instrument data capture through model development, simulation, and AI-assisted analysis. Planned integrations include the American Science Cloud, autonomous laboratory environments, digital twins, persistent AI services, and future workflows spanning CPUs, GPUs, and quantum processors.

The architecture reflects a broader DOE objective of shortening the path from experimental data and large-scale simulations to usable scientific insights. Initial target areas include energy research, materials science, medicine, advanced manufacturing, and national security. Potential use cases cited by AMD include critical-mineral recovery, fusion energy, hydropower, flood response, and electrical-grid resilience.

AMD also highlighted Frontier, the existing AMD-powered system at Oak Ridge, as an example of how to combine AI and HPC resources for research workflows. According to AMD, researchers have reduced a plant-imaging analysis workflow from 168 hours to less than one minute on Frontier. The system is also supporting materials research related to tritium production for future fusion-energy systems.

Discovery is planned as a subsequent exascale-class platform developed with the DOE, Oak Ridge National Laboratory, and HPE. Per ORNL, it will run 6th Gen AMD EPYC “Venice” CPUs and AMD Instinct MI430X GPUs on a DAOS-based HPE Cray Supercomputing platform, with delivery expected in 2028. That makes Discovery the flagship home for the MI430X, the sovereign-AI and HPC member of the MI400 series, paired with the same Venice CPUs coming to the broader server lineup. AMD described it as a long-term system that will extend the work of Frontier and Lux by more tightly integrating large-scale simulation, AI, and data-intensive scientific workflows.

Thomas Zacharia, senior vice president at AMD, characterized Lux and Discovery as open and interoperable platforms designed to combine simulation, AI, data, and emerging computing approaches. AMD said it is working with funded research teams to identify workloads suited to Lux and establish measurable, repeatable outcomes for Genesis Mission projects.

The post AMD’s Lux Becomes the First Genesis Mission Supercomputer, With Discovery to Follow in 2028 appeared first on StorageReview.com.

Wasabi Technologies and Megaport Limited have entered a strategic partnership to add Wasabi cloud object storage to Megaport Storage. The integration combines Wasabi’s object storage platform with Megaport’s private global network, targeting AI data platforms, analytics environments, cyber resilience workflows, and other large-scale cloud storage use cases.

The service is available through the Megaport platform in supported regions and is designed to simplify the deployment of private, scalable object storage. Customers can provision dedicated connectivity to Wasabi storage through Megaport’s automation platform rather than relying solely on public internet paths. Megaport is extending private connectivity to Wasabi’s 16 global storage regions.

The combined offering is positioned for organizations building multi-provider cloud architectures, where compute, networking, and storage services may be sourced independently but require integrated connectivity. Use cases include AI training and inference data repositories, data lakes, cloud-native applications, media and content libraries, backup targets, and recovery environments.

Megaport’s network provides the connectivity layer, while Wasabi supplies the object storage platform. The companies state that the service is intended to provide more consistent network performance and predictable storage costs for workloads that move or access large data sets.

The partnership completes what the industry has taken to calling a neocloud stack: purpose-built AI infrastructure assembled outside the hyperscalers. Megaport’s software-defined network spans more than 1,100 enabled locations, its Latitude.sh acquisition brought GPU-accelerated compute, and Wasabi now supplies the object storage layer, giving the company all three pillars of an AI infrastructure platform for latency-sensitive, data-intensive workloads. The storage layer has drawn far less attention than GPUs in the neocloud wave, but the data feeding those GPUs has to live somewhere, move quickly, and not consume the budget.

Wasabi’s role is to provide the scalable storage layer for AI, analytics, enterprise data, and cyber resilience use cases. Through private links, organizations can move, manage, and access data across Megaport-connected infrastructure without placing all storage traffic on public internet connections.

The two companies plan further technical integrations and joint go-to-market work around AI infrastructure, cloud storage, and private network connectivity.

The post Megaport’s Neocloud Gets Its Storage Layer as Wasabi Powers Megaport Storage appeared first on StorageReview.com.

Chelsio Communications has introduced its seventh-generation AI Interconnect Platform, expanding its Ethernet offload portfolio with SmartNICs, storage controllers, and Data Processing Units. Built on the company’s Unified Wire architecture, the T7 platform targets AI training and inference clusters, disaggregated storage, cloud infrastructure, and high-performance networking environments.

The platform supports Ethernet speeds from 1GbE through 400GbE and combines native 400Gb Ethernet with Unified RDMA support for both iWARP and RoCEv2. Chelsio is positioning the architecture as a standards-based Ethernet fabric for AI deployments that need low-latency communication, high-bandwidth data movement, and hardware acceleration across compute and storage infrastructure.

The seventh-generation platform integrates networking, storage, and compute acceleration in a common hardware and software architecture. Chelsio states that the design maintains software compatibility with its T4, T5, and T6 product families, allowing existing users to migrate to higher-speed Ethernet and PCIe Gen5 deployments without a full software transition.

For AI and HPC workloads, the T7 architecture includes lossless Ethernet support, congestion-management capabilities, GPU-oriented data paths, and RDMA acceleration. These features are intended to reduce synchronization overhead in distributed training clusters and support GPU-to-GPU and GPU-to-storage transfers. Hardware offload and CPU-bypass capabilities are also designed to reduce host CPU utilization for network and storage I/O.

Storage acceleration is a central component of the platform. Chelsio lists hardware offloads for NVMe/TCP, NVMe over Fabrics, iSCSI, RDMA, and TCP Offload Engine functions. The company is targeting JBOF deployments, enterprise storage systems, and disaggregated storage architectures where protocol processing can affect CPU efficiency and application latency.

The platform also includes a programmable data path for workload-specific acceleration and inline processing. Its security feature set includes on-chip cryptographic acceleration for QUIC, TLS, kTLS, and IPsec. Virtualization and cloud features include SR-IOV, virtual-switch, and container offloads designed to help maintain multitenant isolation and network performance.

Product Families

The initial T7 product family includes SmartNICs based on Chelsio’s S7 silicon, storage controller adapters based on N7 Storage Controller silicon, and fully programmable T7 DPU adapters.

Available SmartNIC products include the S7250, S7450, and S7450-OCP, which provide dual- and quad-port configurations for 1/10/25/50GbE deployments. Chelsio positions these adapters for virtualization, embedded systems, edge environments, and lower-speed cloud infrastructure.

The S72200 and S72200-OCP are dual-port 40/50/100/200GbE SmartNICs for cloud, HPC, and AI infrastructure. At the top end of the SmartNIC range, the S71400 provides a single 400GbE port with configurable 4x100GbE and 2x200GbE modes for high-density fabric deployments.

Chelsio is also offering T72200 and T7450 storage controller adapters in PCIe and OCP form factors. Built on the N7 silicon, they provide hardware offloads for NVMe/TCP, NVMe-oF, iSCSI, and RDMA-based storage traffic, along with zero-copy and kernel-bypass acceleration for AI and storage workloads.

The T72200-DPU and T7450-DPU are Chelsio’s programmable DPU offerings. The adapters support stateful and stateless acceleration, direct data placement, and per-connection offload. Supported protocols include TCP/IP, UDP/IP, RoCEv2, iWARP, iSCSI, NVMe-oF, NVMe/TCP, NVGRE, VXLAN, TLS, IPsec, kTLS, QUIC, and RSA-related offloads.

Ethernet for AI Fabrics

Chelsio is positioning T7 as an Ethernet alternative to proprietary AI interconnect technologies. At 400GbE, the platform is intended to support distributed AI training and real-time inference workloads through RDMA, congestion management, protocol offloads, and GPU-optimized data movement.

The company expects these capabilities to improve GPU and storage utilization by reducing host-side protocol processing and supporting higher-throughput communication between compute nodes and storage systems. The approach also allows organizations to use Ethernet standards and existing operational practices as AI cluster requirements scale.

Availability

Chelsio SmartNICs and storage controller products are available through OEM, ODM, and distribution channels. The T7 DPU is currently offered as an evaluation platform for early testing, with production availability planned for December 2026.

The post Chelsio’s T7 Brings 400GbE RDMA SmartNICs, Storage Controllers, and DPUs to AI Fabrics appeared first on StorageReview.com.

WEKA has launched two connected products: NeuralMesh 6, its most significant software release to date, and WEKApod 3, a new generation of storage appliances designed and engineered by WEKA to run it. NeuralMesh 6 remains deployable on customer-selected hardware, while WEKApod provides the turnkey path, shipping with the software preinstalled. NeuralMesh 6 introduces native multi-tenancy, a combined file-and-object protocol stack, metadata-driven data mobility, always-on data reduction with contractual guarantees, Kubernetes-native operations, and integrated observability. WEKApod 3 is custom-designed hardware built to run this software at what the company claims is the highest capacity and performance density available in a single rack, with three configurations, Nitro, Prime, and Prime Max, targeted at maximum performance, balanced capacity, and maximum density, respectively.

Both announcements frame the release around a common market shift: as AI workloads move from training toward long-context, agentic, and retrieval-driven inference at production scale, storage and memory infrastructure, not just GPU count, increasingly determine cost per token and achievable throughput.

NeuralMesh 6: Multi-Tenancy, Unified Protocols, and Data Mobility

NeuralMesh 6 combines features that WEKA claims AI infrastructure operators have traditionally needed to piece together from various vendors: multi-tenancy, a unified protocol stack for files and objects, data mobility across sites, continuous data reduction, Kubernetes-native management, and integrated observability, all integrated into a single software stack instead of assembled from separate parts.

Multi-tenancy is split into two tiers that can be combined. Composable Clusters provide hardware-level isolation, with dedicated CPU, memory, and storage drives per tenant, intended for anchor tenants that need guaranteed resources and predictable performance. Virtual Multi-Tenancy adds VPC-style network isolation through WEKA’s Virtualized RDMA Data Fabric, supporting private VLANs, overlapping IP address spaces, per-tenant quality of service and encryption with independent key management, and independent LDAP or Active Directory authentication per tenant. Virtual Multi-Tenancy scales to more than 1,000 isolated logical tenants per cluster, with new tenant provisioning in under 30 minutes. The two tiers compose, so a single WEKA hardware cluster running 50 Composable Clusters can support up to 50,000 logically isolated tenants on the same physical infrastructure, enabling growth from dozens to tens of thousands without re-architecting the system.

On the protocol side, NeuralMesh 6 implements a native S3 stack in which the same physical data blocks are addressable via S3 and POSIX simultaneously, rather than via a gateway that translates between them. A file written via NFS or POSIX is immediately readable from S3, and vice versa, eliminating duplicate full-dataset copies that typically accumulate as data moves between training, fine-tuning, and inference stages. WEKA built the S3 implementation specifically for AI access patterns, supporting 2,000 to 5,000 concurrent S3 connections per node, roughly five times the concurrency of conventional S3 architectures, with S3 over RDMA enabling zero-copy transfer directly into GPU memory.

Data mobility is handled through metadata-first replication, making a destination environment immediately browsable rather than requiring a complete data copy before a workload can start. Data hydrates on demand, reducing WAN traffic and allowing organizations to place workloads wherever GPU capacity exists rather than where the data was originally written. This release adds asynchronous replication and remote caching as a first step toward broader federation and a global namespace across sites and clouds.

That replication capability is already underpinning real deployments: Sam Tabar, CEO of WhiteFiber, said NeuralMesh’s intelligent replication lets the company make datasets visible across sites and pull exactly the data each job needs to the next GPU allocation, the same architecture behind Project Redwood, the 111.2 Tbps cross-data-center supercluster we covered earlier this month.

Data Reduction up to 6X Capacity Savings

NeuralMesh 6 also enables data reduction, including fingerprinting, similarity hashing, deduplication, and compression, by default across every deployment, with a write overhead below 5%, up to 6x capacity savings on AI training data, and a contractual guarantee covering both reduction ratio and performance impact. A new Kubernetes Operator automates cluster deployment and lifecycle management for organizations running Kubernetes as their standard operating model, which WEKA says cuts deployment time from weeks to hours. NeuralMesh Observe, included at no extra cost with every deployment, provides SaaS-based multi-cluster dashboards, client-level diagnostics, and alerting routed to Slack, PagerDuty, or email.

WEKA reports using its Augmented Memory Grid feature in production, which expands GPU memory by speeding up persistent KV cache access to NeuralMesh-managed NVMe storage on Oracle Cloud Infrastructure. Benchmarks on OCI H100 infrastructure demonstrated 10x higher token throughput, 10x more concurrent users served, and 7x more tokens per GPU, which WEKA says is measured against DRAM-based alternatives. Pablo Selem, senior director of software development at OCI, characterized the approach as removing memory bottlenecks so customers can achieve higher throughput and more users from the same GPU footprint.

WEKApod 3: Custom Hardware Built Around the Software

WEKApod 3 is WEKA’s own hardware design rather than a reference architecture built on third-party OEM chassis. The company says a single WEKApod rack delivers 1.1 exabytes of effective capacity on a hardware foundation of 441.5 PB of raw capacity, making it the first single-rack system to exceed an exabyte of effective capacity. Per-rack throughput is rated at 10.2 TB/s with 210 million IOPS. WEKA reports 267% higher effective capacity density and 114% higher throughput density per rack unit than the next-best publicly available alternative in each category.

The design relies on a PCIe Gen 6 internal fabric, a cable-based drive interconnect rather than a backplane, NVIDIA ConnectX SuperNIC networking for Spectrum-X Ethernet connectivity, and a software-managed thermal architecture rated for 35°C ambient operation that throttles NVMe power under thermal stress instead of shutting down. WEKA has multiple patents pending on the chassis, drive interconnect, thermal management, and serviceability design. Serviceability features include hot-pluggable boot drives with a GUI-guided replacement process WEKA says takes about 10 minutes instead of a multi-hour maintenance window, along with headless, cloud-driven rack-scale deployment through NeuralMesh Home.

Three configurations for different workload priorities

WEKApod Nitro is designed for workloads where storage bandwidth is critical, ensuring GPUs remain saturated. It features a two-rack-unit, four-node chassis with four independent failure domains and 56 TLC drives, supported by dual-port NVIDIA ConnectX networking that delivers 800 Gb/s throughput. WEKApod Prime emphasizes balanced capacity and performance with an AlloyFlash blend of TLC and QLC drives, housed in a similar four-node, two-rack-unit chassis that supports 56 drives. WEKApod Prime Max maximizes capacity in a compact form: a two-rack-unit, two-node chassis containing 70 NVMe drives, using Micron’s 245.76 TB 6600 ION SSDs, combined with NeuralMesh’s object storage and data reduction techniques, to achieve an effective capacity of 1.1 exabytes in a single 56U rack.

AlloyFlash, the tiering feature that makes the Prime and Prime Max configurations viable, automatically routes latency-sensitive operations to TLC flash while directing bulk-capacity data to QLC, which runs roughly 30-40% cheaper per terabyte, without requiring customer configuration. This is the clearest point at which the NeuralMesh 6 software release and the WEKApod 3 hardware release are one product: the software’s tiering logic makes the higher-density hardware configurations usable at production performance levels, rather than just a larger capacity number on a spec sheet.

WEKA bases its hardware decision on current data center constraints: US data center construction dropped in 2025 for the first time since 2020, grid connection queues in major markets now take four to seven years, and Morgan Stanley forecasts a 49-gigawatt power shortfall in the US through 2028. This is in addition to ongoing NAND supply issues and longer OEM lead times. WEKA argues that storage that is inefficient in rack space and power directly competes with GPUs for limited physical resources. By managing its own hardware supply chain rather than relying on OEM channels, it believes it can offer more predictable pricing and lead times for customers planning large-scale infrastructure projects.

Jason Hardy, VP of Storage Technology at NVIDIA, said Spectrum-X Ethernet networking gives WEKApod 3 the high-bandwidth, low-latency fabric needed to keep the storage-to-GPU data path clear at scale. Steve McDowell, chief analyst at NAND Research, argued that inference at production scale is a different infrastructure problem than training, with tokens per rack, tokens per watt, and cost per inference at sustained load becoming the metrics that matter, a scorecard he says buyers should evaluate every vendor against. Jeremy Werner, senior vice president and general manager of Micron’s Core Data Center Business Unit, added that the new WEKApod architecture with Micron’s 245TB SSDs delivers 15.8 petabytes in a 2U footprint, preserving power and space for additional compute.

Availability

NeuralMesh 6 is expected to be generally available in the second half of 2026. Current WEKA customers can upgrade at no additional cost through standard channels. WEKApod Nitro, Prime, and Prime Max are now available for ordering through WEKA’s distributor and VAR network, with deliveries starting in fall 2026 and NeuralMesh 6 pre-installed. Additionally, WEKApod is now offered in configurable SKUs for the first time in this generation, allowing customers to choose chassis type, memory, drive capacity, and drive count, supporting systems from under 1PB up to 100PB or more in a single setup.

The post WEKA’s WEKApod 3 Breaks the Single-Rack Exabyte Barrier as NeuralMesh 6 Goes Multi-Tenant appeared first on StorageReview.com.

AMD has expanded its strategic partnership with Microsoft to cover GPUs, CPUs, networking, and software across Azure infrastructure. The collaboration includes Microsoft’s planned deployment of AMD’s Helios Rackscale Solution for AI inference workloads supporting frontier models, Azure AI services, and customer applications. The deployment makes Microsoft the first hyperscaler to publicly commit to Helios at scale, the clearest signal yet that AMD’s rack-scale platform is landing as a production alternative to NVIDIA’s rack-scale systems rather than a reference design.

Helios is AMD’s rack-scale AI platform, integrating Instinct MI455X GPUs, 6th Gen AMD EPYC Venice CPUs, Pensando networking technology, and the ROCm software stack. The platform is designed to provide an open architecture for large-scale AI training and inference, positioning AMD to offer an alternative infrastructure stack for cloud providers and organizations building large model environments.

AMD’s companion blog frames Helios as a system-level play rather than a component sale: a fully integrated platform designed around AMD compute, networking, software, power, and cooling, built to span training, inference, fine-tuning, and agentic workloads rather than optimizing for any single one. The division of labor is explicit: AMD supplies the silicon foundations, while Microsoft wraps them in Azure’s cloud operations, developer services, security, and AI platform services; the companies’ shared argument is that rack-scale infrastructure becomes more valuable when it arrives inside a complete cloud platform customers can consume at scale. The timing is no accident either: the announcement lands just ahead of AMD’s Advancing AI 2026 event, where Microsoft and AMD are co-presenting sessions on sovereign AI, AI economics, and production-scale infrastructure.

For Azure, the deployment targets inference workloads across Microsoft services and enterprise applications. Frontier model developers will be able to use AMD-powered Azure infrastructure for training and deployment, while enterprise customers can manage production AI workloads through Azure Foundry Managed Compute.

The companies are also extending their CPU collaboration with two new Azure VM families based on 6th Gen AMD EPYC Venice processors. Azure HDv2 is intended for agentic AI and data pipeline workloads, while Azure HXv2 targets semiconductor design use cases. These instances expand Azure’s existing AMD EPYC footprint across AI, data-intensive, and engineering workloads.

Networking is another component of the partnership. Building on Microsoft’s existing broad deployment of AMD Pensando DPUs, Azure is extending them into AMD AI backend networking infrastructure and select Azure services, and the companies are integrating Azure Boost with AMD technologies to improve networking performance, efficiency, and connection processing at cloud scale. That matters most for AI clusters, where east-west traffic between GPUs grows faster than anything else in the data center.

AMD and Microsoft characterized the expanded arrangement as an effort to provide scalable AI infrastructure spanning model development, inference, data preparation, search, and reinforcement learning. For AMD, the agreement broadens the role of its GPU, CPU, DPU, and software portfolio within a major hyperscale cloud environment.

AMD expects to begin shipping Helios systems to customers, including Microsoft, during the second half of 2026.

The post Microsoft Will Ramp AMD’s Helios Rack-Scale AI Platform at Scale on Azure appeared first on StorageReview.com.

NVIDIA has introduced Omniverse libraries, a set of software components designed to bring physical AI capabilities into existing 3D applications and prepare content for simulation. The libraries are intended to help developers, technical artists, and engineering teams build workflows that inspect scenes, validate assets, model physical behavior, and generate sensor data for robotics, factory automation, and autonomous-system development.

Physical AI systems typically require training and validation in simulated environments before deployment. That process depends on more than photorealistic 3D content. Simulation assets must include correct geometry, materials, scale, labels, sensor definitions, and physical properties such as mass, friction, and collision behavior. NVIDIA positions the new libraries as building blocks for AI agents that can assist with these preparation and validation tasks inside established design and content-creation environments.

The initial release includes ovrtx, ovphysx, and CAD-to-SimReady skills. The components are available on GitHub, alongside a Blender integration blueprint that demonstrates how developers can add agent-ready simulation features to an existing 3D application.

ovrtx provides NVIDIA RTX-based sensor simulation, enabling applications to generate virtual camera, lidar, radar, and other sensor outputs from 3D scenes. This capability allows developers to assess how a physical AI system would perceive a simulated environment before testing hardware in the field.

The ovphysx library brings GPU-accelerated physics capabilities to simulation workflows. It supports modeling of collisions, mass, friction, motion, and other physical interactions required to evaluate robotic behavior and industrial processes in a virtual environment.

CAD-to-SimReady skills focus on converting CAD data into OpenUSD-based SimReady assets. The workflow is intended to preserve engineering content while adding the structure and simulation attributes needed for physical AI development, including robotics and autonomous-system testing.

Early Software Integrations

SideFX and PTC are among the software providers working with the Omniverse libraries. SideFX is evaluating OpenUSD workflows with ovrtx and ovphysx as part of its Houdini procedural 3D content-creation environment. The effort aims to enable agent-assisted workflows for generating procedural content, evaluating physics, and preparing scenes for simulation, while keeping technical artists in control of the underlying creative process.

PTC is integrating OpenUSD and ovrtx into its Onshape CAD and product data management platform. The integration is intended to connect cloud-native design workflows with physical simulation, allowing engineering teams to carry product content across CAD, PDM, collaboration, validation, and simulation processes without repeatedly reworking assets.

At SIGGRAPH 2026, NVIDIA demonstrated a SimReady Blender workflow built with the Omniverse libraries and the NVIDIA Nemotron Ultra open model. The reference implementation shows how developers can introduce RTX sensor simulation, physics, and validation into Blender-based workflows while retaining creator control. NVIDIA has made the Blender blueprint publicly available.

The company stated that these workflows can run locally across systems ranging from compact NVIDIA RTX Spark devices to NVIDIA GB300-powered DGX Station systems. NVIDIA expects RTX Spark systems from several OEM partners to become available in the fall, while DGX Station systems are available through multiple system providers.

Startups Build Agent-Assisted Asset Pipelines

Several startups are also applying Omniverse components to asset and scene-preparation workflows. Palatial is using CAD-to-SimReady skills to automate the creation and validation of SimReady assets from CAD files at scale. Lightwheel uses NVIDIA Content Agents and OpenUSD in its SimReadyGen technology to generate physically accurate simulation assets from text prompts.

ForgeCAD and MoonlakeAI are developing agent-powered 3D workflows built on Omniverse capabilities. Their work focuses on assisting with the creation, enhancement, and preparation of assets for simulations used to train and validate real-world AI systems.

The post NVIDIA Omniverse Libraries Put AI Agents to Work Prepping 3D Content for Simulation appeared first on StorageReview.com.

The Cloud Native Computing Foundation (CNCF) announced on July 16 that Broadcom has elevated its membership to Platinum. The move expands Broadcom’s participation in the CNCF ecosystem as enterprises operationalize AI, analytics, and other data-intensive workloads on Kubernetes-based platforms.

The announcement follows Broadcom’s recent contribution of Velero to the CNCF Sandbox. Velero is a Kubernetes-native platform for backup, restore, and migration operations. It enables platform teams to protect Kubernetes cluster resources and persistent data, support disaster recovery workflows, and move applications between clusters or environments. These capabilities are increasingly relevant for AI infrastructure, where reproducible deployment state, data protection, and workload portability are operational requirements.

Broadcom’s cloud-native lineage runs deep: the company originated Harbor, Antrea, Velero, and Contour, and it maintains components used throughout Kubernetes environments, including Cluster API, etcd, containerd, Kubernetes CSI, and controller-runtime. CNCF noted that Broadcom has ranked among the top three Kubernetes contributors over the past decade.

The Platinum membership positions Broadcom to further support the open-source technologies underpinning cloud-native platform operations, a scope that extends beyond container deployment to include resilience, security, and lifecycle management.

For infrastructure teams, the expanded participation reflects continued vendor investment in the Kubernetes control plane, container runtime, storage interface, and application mobility layers. These components are central to building and operating standardized platforms for modern applications and AI workloads across private, public, and hybrid cloud environments.

Additional information on CNCF membership is available at the CNCF membership page.

The post Broadcom Goes CNCF Platinum After Handing Velero to the Sandbox appeared first on StorageReview.com.

Dell Technologies closed the first quarter of 2026 as the top external enterprise storage vendor worldwide, according to IDC’s Worldwide Quarterly Enterprise Storage Systems Tracker released in June, posting 31.2% market share on 40.8% year-over-year growth, nearly double the 22.7% pace of the overall segment. By IDC’s count, the rest of the top five vendors combined did not outsell Dell for the quarter. While the share figure is the headline, Dell’s broader message is that the result reflects continued execution around three customer priorities: private cloud flexibility, AI infrastructure built on enterprise-owned data, and cyber resilience embedded into the platform.

That positioning is especially relevant for Dell because the enterprise storage conversation has shifted beyond raw capacity and performance. Customers increasingly want infrastructure that supports mixed hypervisor environments, AI data pipelines, and faster recovery from cyber events without requiring wholesale architectural changes. Dell’s current portfolio, with PowerStore at the center of the midrange discussion, is being framed around that convergence.

On the private cloud side, Dell is targeting enterprises that want both operational simplicity and infrastructure choice. Through Dell Private Cloud and the Dell Automation Platform, customers can deploy cloud software stacks from Broadcom, Microsoft, Nutanix, and Red Hat on disaggregated Dell infrastructure rather than being tied to a single hyperconverged model. At Dell Technologies World, the company added support for VMware Cloud Foundation 9.1, Microsoft Azure Local, and PowerStore integration with Nutanix AHV.

The practical value of that approach is that compute and storage can scale independently while lifecycle management remains automated. Dell is also making a cost argument, saying this model can reduce costs by up to 65% compared with traditional HCI approaches. For infrastructure teams trying to balance modernization with budget control, that flexibility is a meaningful part of the story.

PowerStore remains a key piece of that strategy. Dell is positioning PowerStore Elite as the storage foundation for organizations that need to consolidate block, file, virtual machine, and container workloads on a single platform while maintaining hypervisor flexibility. The company says the platform delivers up to three times the performance of prior models, includes a 6:1 data reduction guarantee, and supports non-disruptive upgrades. Existing customers can also modernize through mixed-generation clustering, which allows newer and older systems to operate together and reduces the operational friction that often comes with hardware refresh cycles.

That upgrade path is one reason PowerStore continues to resonate in the market. It gives customers a way to evolve their infrastructure without forcing a hard cutover while keeping pace with changing workload demands across virtualized environments, databases, and increasingly containerized applications. For a platform with broad enterprise deployment, continuity matters as much as peak performance.

Dell is making a similarly integrated case around AI. The company’s view is that successful enterprise AI initiatives depend less on public data sources and more on the ability to discover, govern, and operationalize internal data. Its AI Data Platform, part of the broader Dell AI Factory framework, is designed to index billions of unstructured files and orchestrate them through managed data pipelines to move customers from AI pilots into production more quickly.

Dell also highlighted GPU-accelerated analytics, noting that the platform can deliver SQL query performance up to 6 times faster when paired with NVIDIA Blackwell GPUs. Underneath that software layer, storage remains central. PowerScale and ObjectScale are positioned as the systems that turn fragmented enterprise data into AI-ready repositories. Dell said the new ObjectScale X7700 ultra-dense appliance offers up to 45% more HDD capacity than its predecessor, improving economics for large-scale object storage deployments. The company also said support for 245TB all-flash drives is coming, which would more than triple ObjectScale’s flash density.

Cyber resilience is the third pillar of Dell’s current storage narrative and is increasingly tied to both private cloud and AI deployments. Enterprise customers are no longer treating cyber recovery as a secondary planning exercise. Instead, the expectation is that ransomware detection, protection orchestration, and large-scale recovery need to be part of the platform itself.

Dell is addressing that with PowerProtect One, which brings data protection orchestration and storage under a single control plane. The company said this can reduce management overhead by up to 50% while preserving large-scale recovery capabilities and established data reduction efficiency. It also pointed to Cyber Detect, which integrates AI-based ransomware detection with PowerStore and PowerMax; Dell cites a 99.99% detection confidence from a third-party validation of the underlying Index Engines technology, with the goal of initiating recovery operations as soon as suspicious activity is detected.

Taken together, these elements help explain why Dell’s storage message is landing with enterprise buyers. Rather than presenting private cloud, AI infrastructure, and cyber recovery as separate product discussions, Dell is tying them back to a common infrastructure foundation. In that context, PowerStore plays an important role because it sits at the intersection of consolidation, modernization, operational simplicity, and resilience.

For customers already invested in Dell infrastructure, that consistency reduces friction. For prospective buyers, it strengthens the case that Dell’s storage portfolio is not just broad, but aligned with how enterprise IT priorities are evolving. The Q1 share gains and growth numbers may capture attention. Still, the more durable story is that Dell continues to pair scale with a portfolio strategy that maps cleanly to current enterprise requirements.

The post Dell Outsells the Rest of IDC’s Top Five Combined in Q1 External Storage appeared first on StorageReview.com.

Microsoft and 3M have announced a strategic partnership covering both data center optical connectivity and enterprise AI adoption. Under the July 15 agreement, Azure becomes the first announced hyperscale cloud provider to deploy 3M’s Expanded Beam Optical (EBO) technology across its data centers, while 3M adopts Microsoft’s AI and digital platforms across several of its own business functions. The companies say the arrangement pairs Microsoft’s hyperscale infrastructure and digital tooling with 3M’s materials science and precision manufacturing background, aimed at speeding AI adoption and building out the physical network layer that cloud and AI workloads depend on.

EBO Technology Moves Into Azure Data Centers

The core of the infrastructure side of the deal is 3M’s EBO technology, which Microsoft will implement within Azure data centers as it scales capacity for generative AI and other high-performance workloads. Traditional fiber connectors rely on direct physical contact between polished ferrule end-faces. This design is sensitive to dust and debris and requires regular cleaning and inspection to maintain signal quality. EBO instead uses lenses to expand and collimate the light beam across an air gap between connectors, thereby reducing the connection’s sensitivity to contamination and physical wear. According to the companies, this makes EBO connections faster to install, more tolerant of dust and handling, and less demanding to maintain than direct-contact designs, which matters in dense, high-churn data center environments where cabling is installed and serviced constantly.

Microsoft’s early use of the technology points to shorter deployment windows in specific data center scenarios, as well as stable optical performance in live environments where dust exposure and frequent handling during installation and maintenance are routine. For operators managing large fiber counts across AI clusters, that combination of faster installation and reduced maintenance overhead is the more practical selling point than the underlying optics itself.

3M is scaling production of EBO components to meet demand from hyperscalers and other data center operators building out AI infrastructure. The company has adapted its single-mode EBO design, developed over decades of work in materials science and precision manufacturing, for high-volume data center deployment. 3M was also involved in establishing the EBO Multi-Source Agreement (MSA), an industry effort to standardize the interface so that EBO components from different manufacturers can interoperate, which is typically a precondition for broader adoption across the supply chain rather than a single-vendor arrangement.

Cliff Henson, corporate vice president of Cloud Supply Chain at Microsoft, framed the EBO integration as part of a broader pattern of incorporating partner technology into Azure’s infrastructure to improve deployment speed, resilience, and scalability, providing customers with a more reliable environment for cloud and AI workloads.

3M Applies Microsoft’s AI Tools to Its Own Operations

On the enterprise side, 3M will roll out Microsoft’s AI and digital tools across parts of its transformation plan, including customer service, finance, sales, and marketing. The intent is to streamline internal processes, improve decision-making, and raise both customer satisfaction and employee productivity.

One concrete example already underway involves Microsoft Frontier Company engineers working alongside 3M’s Global Business Services group to automate customer order management. The two teams are building an AI-driven workflow that handles credit checks, delinquency assessments, and related system updates, with human-in-the-loop controls and a dashboard for real-time monitoring and approvals built into the process. The intent is to reduce manual processing, make the workflow more consistent, speed up cash flow, and free 3M staff to focus on higher-value work, while making the underlying process more scalable and auditable.

3M executive vice president and chief strategy officer Jon Van Wyck described AI as a lever for accelerating growth, improving customer experience, and increasing team effectiveness, and pointed to the Microsoft partnership as a way to optimize enterprise operations and build out AI infrastructure with an emphasis on practical, mutually beneficial outcomes rather than open-ended experimentation.

Next Up

Microsoft and 3M said they plan to continue the partnership through ongoing technical collaboration, direct engagement between engineering and commercial teams, and joint work across Microsoft’s data center and device ecosystem. The stated focus areas are reliability, faster deployment, higher interconnect density, and long-term scalability, drawing on 3M’s background in materials science, optical connectivity, and manufacturing.

Most notable is the EBO Multi-Source Agreement rather than the Azure deployment. A single hyperscaler adopting a new connector design is a data point; an MSA aimed at standardizing that design across vendors determines whether expanded-beam optics becomes a broader alternative to direct-contact fiber connectors industry-wide, rather than a one-off implementation limited to Microsoft’s own data centers.

The post Azure Becomes First Announced Hyperscaler to Deploy 3M’s Expanded Beam Optical Fiber appeared first on StorageReview.com.

NVIDIA has disclosed a comprehensive multi-track strategy to accelerate AI adoption in Japan, combining a new national AI infrastructure buildout with a broad ecosystem of industry-specific models and specialized edge hardware. The strategy moves Japan toward a unified platform approach that addresses localized language requirements, demographic shifts, and industrial workflows across the robotics, healthcare, and automotive sectors.

A central piece of this strategy is the launch of a national AI factory for physical AI, developed in collaboration with Noetra Corp. The facility will be architected with NVIDIA Vera Rubin NVL72 racks featuring 13,750 Vera CPUs and 27,500 Rubin GPUs. Managed via the NVIDIA DSX platform and scaled through Spectrum-X Ethernet networking, the 140-megawatt data center is intended to support the development of trillion-parameter-scale multimodal foundation models. This infrastructure will serve as the technical foundation for the Ministry of Economy, Trade and Industry’s FRONTia Project, which focuses on high-reliability AI for manufacturing, logistics, healthcare, and telecommunications.

An aggressive push into localized model development is keeping pace with the infrastructure buildout. Leading Japanese enterprises and research institutions are already using Nemotron open models, datasets, and libraries to build solutions for the domestic market. The Institute of Science Tokyo developed its Swallow family of models using Nemotron assets to prioritize Japanese language comprehension while maintaining standard capabilities in English and coding. In the telecommunications sector, SoftBank subsidiary SB Intuitions has developed the Sarashina generative AI models, with the Sarashina3 mini variant already selected by Japan’s Digital Agency.

Physical AI deployment is driving significant momentum across Japan’s industrial base, supported by the expanded Jetson Thor lineup. NVIDIA introduced the Jetson T3000 and T2000 modules to support mainstream robotics and embedded AI deployment. The T3000 delivers 865 FP4 teraflops from a Blackwell GPU in a package roughly half the size and power of the flagship T5000, and NVIDIA says it maintains similar inference performance for multimodal workloads; the modules are slated for availability in the first quarter of 2027, with emulation support arriving this month in JetPack 7.2.1. To support these modules, NVIDIA released Cosmos 3 Edge, a 4-billion-parameter world foundation model built on Nemotron and designed for on-device reasoning and robot policy execution. The model allows embodied systems to perceive their environment and generate actions locally on edge hardware, bridging the gap between centralized training and real-world execution.

In the automotive sector, Toyota is expanding its partnership with NVIDIA to connect AI across vehicles, infrastructure, and industrial operations. The vehicle side builds on last year’s announcement that Toyota will develop next-generation vehicles with L2++ advanced driver-assistance capabilities using NVIDIA DRIVE AGX and the safety-certified DriveOS platform. What’s new is on the factory floor and in the city: Toyota is bringing simulation to manufacturing using NVIDIA Omniverse libraries and Isaac Sim for factory and robotics workflows, while Woven by Toyota has developed a multimodal vision-language model for urban traffic intelligence, trained on NVIDIA H100 GPUs with Megatron-Core and designed to interpret real-world conditions and anticipate what happens next.

The healthcare track is equally robust, with Japanese innovators utilizing NVIDIA technology across surgical robotics, CT imaging, and drug discovery. Canon launched Japan’s first NVIDIA-accelerated photon-counting CT system, while Fujifilm commercialized Japan’s first whole-body CT system powered by NVIDIA Blackwell, using diffusion-based deep learning reconstruction to improve image quality. In the life sciences, the Tokyo-1 AI drug discovery consortium operated by Xeureka continues to expand, with Eisai joining this past April alongside Astellas, Daiichi Sankyo, and Ono Pharmaceutical, all advancing drug discovery with NVIDIA BioNeMo. Kawasaki Heavy Industries plans to use NVIDIA Holoscan IGX, Isaac for Healthcare, Isaac GR00T, and Cosmos to develop surgical-support, nursing-assistant, and hospital transport robots.

To accelerate the deployment of these vision-based systems, NVIDIA is expanding its Metropolis libraries. New tools such as VSS Blueprint 3.2, DeepStream 9.1, and TAO 7 enable developers to use coding agents to build and manage video intelligence systems. Industry leaders in Japan, including Hitachi, OMRON, and Shimizu Corporation, are leveraging these Metropolis tools to implement vision AI agents across factories, construction sites, and public infrastructure.

NVIDIA’s goal is to deliver an open, portable AI stack rather than a closed service. By providing models with open weights and recipes, NVIDIA enables Japanese organizations to maintain data sovereignty, meet local regulatory requirements, and deploy AI wherever their applications run. This layered strategy—from national Rubin GPU clusters to edge-optimized Jetson modules and specialized healthcare systems—positions Japan as a key hub for large-scale physical AI deployment.

The post NVIDIA and Japan Launch 27,500-GPU Vera Rubin AI Factory as Physical AI Push Spans Every Industry appeared first on StorageReview.com.

Axelera AI has launched Voyager Wingman, an AI-powered assistant designed to accelerate development with its Voyager Toolkit. The tool was first shown at CES in January and has since undergone extensive testing ahead of this broader release. Wingman lets developers use natural language to query the Voyager SDK and Axelera’s full documentation set, helping them build AI pipelines, debug issues, and troubleshoot applications without manually digging through reference docs. Axelera’s internal testing puts Wingman’s results 20-30% ahead of using Claude Code on its own.

Building AI applications for dedicated hardware involves a fair amount of overhead beyond the model itself: exporting and compiling models, configuring pipelines, tuning for performance, and validating outputs all take time. Wingman is meant to sit atop that process as an assistant, with direct access to Axelera’s SDK, documentation, and the full software repository, rather than requiring developers to piece that context together themselves. Axelera, a European AI semiconductor company behind the Metis and Europa edge inference platforms, says its hardware is deployed across more than 500 customers spanning telecommunications, aerospace, and enterprise sectors.

Core Functionality

Wingman is built around four areas where Axelera says developers tend to spend the most time. The first is pipeline construction, where developers can describe the application they want, including models, pre- and post-processing steps, or multi-stage pipelines, and Wingman helps assemble a working computer vision pipeline from that description.

The second is performance tuning. Wingman can suggest specific optimization techniques and compiler configurations to maximize throughput on Axelera’s hardware, rather than leaving developers to work through the compiler options on their own.

Third is troubleshooting. Wingman is built to recognize common failure patterns, including configuration errors, device enumeration issues, and failed model compilations, and to point developers toward a fix rather than just surfacing the error.

The fourth area is documentation access. Developers can ask direct questions about supported operators, APIs, runtime behavior, or configuration syntax and receive a contextual answer with examples, rather than searching through reference material.

Access and Deployment

Voyager Wingman is now available as a web-based chat interface and a standalone app. A plugin or skill for existing AI coding frameworks is planned for a later release. Across all of these access points, Wingman is designed to automatically stay current with Toolkit updates, new examples, and documentation changes, without requiring developers to download or manually update anything.

The web-based chat gives developers access from any platform via a single link, simplifying onboarding for teams that don’t want to install anything locally. Once the plugin or skill ships, it’s intended to integrate into existing developer workflows without requiring changes to their setup or exposing source code externally.

Availability

Voyager Wingman is accessible through the Axelera Developer Community and Customer Portal. The chat version runs on a freemium model, giving developers a free credit allowance to try the assistant before committing to paid usage. The standalone app is free to use on a bring-your-own-key basis, meaning developers supply their own API credentials for the underlying model.

Bram Verhoef, VP of Customer Engineering and Success at Axelera, said the platform’s goal is to make edge AI development on Axelera hardware accessible to developers of all experience levels, not just specialists already familiar with the Voyager Toolkit.

The post Axelera’s Voyager Wingman Pitches 20-30% Better Results Than Claude Code for Edge AI Development appeared first on StorageReview.com.

Cloudera and VAST Data have entered into a partnership to build a unified AI factory architecture for enterprises running continuous AI training, inference, and analytics workloads. The joint offering combines Cloudera’s containerized data services with the VAST AI Operating System, targeting a problem that has become increasingly common in enterprise AI deployments: expensive GPU clusters sitting idle while they wait on data.

That idle-GPU problem, often called GPU starvation, tends to show up when organizations bolt AI workloads onto data architectures that were never built for the continuous, high-throughput demands of modern AI pipelines. Data preparation, training, inference, and analytics all compete for the same infrastructure, and when the data layer can’t keep pace, the accelerators stall. Cloudera and VAST are positioning their combined stack as a fix for that bottleneck, aiming to keep GPUs fed with low-latency data throughout the full AI lifecycle.

How the Architecture is Built

On the Cloudera side, the company brings its lakehouse architecture, which packages data engineering, streaming, analytics, machine learning, and AI services into portable containers that can run across hybrid and multi-cloud environments. That gives customers a consistent operational model regardless of where the workload actually executes.

VAST contributes its Disaggregated Shared Everything (DASE) architecture as the underlying data infrastructure, scaling to exabyte levels while integrating vector database services with NVIDIA cuVS for GPU-accelerated vector indexing and search. The VAST AI OS is built on the NVIDIA AI Data Platform reference design and is designed to take latent enterprise data and turn it into an AI-ready state that can be consumed directly by training and inference pipelines. Cloudera then layers its data engineering, governance, and AI services on top of that foundation.

The combined stack is meant to cover the full path from raw data ingestion through model deployment, with consistent operations across data centers, private cloud, and public cloud. Cloudera and VAST say the architecture eliminates GPU starvation through high-bandwidth, low-latency data pipelines, which should, in turn, improve sustained GPU utilization and compute efficiency. The platform is also built to handle structured, unstructured, and multimodal datasets at scale, with enterprise governance and compliance controls intended for private and sovereign AI deployments.

NVIDIA Integration and Inference

The partnership leans heavily on NVIDIA’s stack. Alongside the AI Data Platform reference design underpinning VAST’s AI OS, the companies are integrating NVIDIA AI Enterprise software into the joint architecture. Cloudera’s AI Inference Service uses NVIDIA NIM microservices to enable organizations to deploy and scale models directly on their own data, including NVIDIA’s newer Nemotron open models.

On the data engineering side, customers can accelerate Apache Spark workloads using NVIDIA cuDF, which integrates transparently with Cloudera Data Engineering. That allows Spark jobs to tap into VAST’s high-throughput data services with GPU-accelerated processing, which should help on data-heavy stages of AI pipelines rather than just training and inference.

For regulated industries in particular, Cloudera and VAST are framing this as a “silicon-to-application” solution, covering everything from the underlying NVIDIA hardware and software up through production AI applications, deployable on-premises or in the cloud depending on data residency and compliance requirements.

Availability

The companies say the partnership combines 60 exabytes of customer-managed data across their installed bases, giving both vendors a large pool of existing enterprise customers to target as demand for private AI infrastructure grows.

The joint Cloudera-VAST AI factory solution is available now through both companies’ enterprise sales teams and partner networks. Cloudera and VAST plan to expand the portfolio through 2026 with additional reference architectures, validated deployment patterns, and industry-specific solutions.

The post Cloudera and VAST Data Take Aim at GPU Starvation With Joint AI Factory Stack appeared first on StorageReview.com.

IBM has expanded its Power server lineup with new software to automate infrastructure management and application development. The announcements include IBM Power Autonomous Operations, an agentic control layer for system management, and the IBM Bob Premium Package for i, an AI-driven development assistant for IBM i environments. IBM also introduced the Power S1112, a compact single-socket Power11 server designed for on-premises AI inference.

The releases build on last year’s Power11 launch, which IBM positioned around availability, resiliency, and scale across on-premises and IBM Cloud deployments. The company frames these newer additions as extending autonomous IT capabilities across the platform, from application code through runtime operations.

IBM cited its Institute for Business Value 2026 Tech Leader Study, which projects enterprises will deploy an average of 1,661 AI agents by 2027, a 38% increase that IBM says leaves tech leaders managing hundreds of thousands of autonomous decisions daily, beyond the reach of manual governance. IBM’s stated rationale for these releases is that closing that gap requires infrastructure capable of self-management, freeing IT teams to focus on application work rather than routine operations.

Hillery Hunter, General Manager for IBM Power and CTO at IBM Infrastructure, said the goal is to let enterprises pursue rapid AI deployment without trading off system stability, positioning increased automation in Power as a way to handle routine availability, optimization, and security tasks while preserving control and resilience.

Power S1112: Compact Server for Local AI Inference

The Power S1112 is a single-socket Power11 system intended for organizations that are extending AI workloads to the edge or on-premises locations outside the data center. It uses Power11’s on-chip Matrix Math Acceleration (MMA) to handle inference locally with reduced latency.

IBM’s performance claims put the S1112 at twice the per-core performance of the Power S914 and three times that of the older Power S814, based on published CPW results for 4-core configurations. IBM also claims up to 69% better energy efficiency than the S914 in a smaller physical footprint; by IBM’s own footnote math, a 10-core S1112 delivers 539 CPW per watt against 319 for an 8-core S914.

Alongside the new server, IBM Technology Lifecycle Services is launching Power Expert Care Premium Essentials, an incident-focused support tier exclusive to the S1112. It includes priority access to IBM specialists, faster response times, and automated support tooling.

Power Autonomous Operations: Automated Infrastructure Management

Power Autonomous Operations is IBM’s new control plane for automating day-to-day infrastructure tasks across Power environments. The platform continuously monitors system behavior to identify and address issues such as capacity constraints before they cause disruptions.

IBM reported internal testing results from an eleven-system Power environment, where the platform resolved a capacity-related issue in 3.33 minutes, compared to 52.59 minutes using traditional manual workflows, a roughly 15x reduction in resolution time. IBM says this is intended to eliminate hours of manual dashboard monitoring and analysis each month.

The platform includes an embedded AI agent that supports natural-language, chat-style interaction to manage and tune Power environments, which IBM says reduces the need for deep domain expertise for routine tasks.

IBM Bob Premium Package for i: AI-Assisted IBM i Development

IBM i remains widely used for core business operations across industries, but modernizing IBM i applications has historically required specialized RPG skills, limiting the pool of available developers. IBM Bob is positioned as an AI-driven development assistant that provides an agentic software development lifecycle experience for IBM i, aimed at helping engineers understand existing codebases, implement changes, and onboard onto IBM i development more quickly. Early adopters are already reporting results; Heartland Co-Op estimates new-to-platform developers understand complex IBM i applications 60% faster with Bob.

The Premium Package for i builds integrated support for IBM i standards into the development lifecycle, with the goal of expanding the number of engineers who can work on and modernize IBM i applications.

Availability

The Power S1112 is expected to be generally available on July 24. Power Autonomous Operations is scheduled for general availability on September 23, 2026. The IBM Bob Premium Package for i became generally available on June 24, 2026.

The post IBM Power S1112 Brings Local AI Inference to the Edge as Power Goes Autonomous appeared first on StorageReview.com.