Adata Technology expands its B2B portfolio with a new Business Applications hub, combining Trusta and Adata Industrial memory and storage solutions across enterprise server, intelligent edge, computer system, smart home, and portable & wearables—targeting search intent for AI infrastructure, data center, and enterprise SSD news. The hub helps enterprise customers, system integrators, and partners select configurations based on workloads, system architectures, and deployment environments. For Enterprise Server applications, Trusta delivers PCIe Gen5 enterprise SSDs in U.2, E1.S, and E3.S form factors, plus DDR5 R-DIMMs, addressing high-speed storage and high-capacity memory for AI servers and HPC. These products are qualified on platforms from AIC, MiTAC Computing, ASRock Rack, and Giga Computing, with selected items listed in public compatibility and AVL/QVL resources. Adata also introduces the AI Scaler Toolkit for AI deployment and management, integrating enterprise memory, storage, and software to accelerate real-world AI adoption. For Intelligent Edge, Adata Industrial provides industrial-grade memory and storage optimized for reliability, environmental resilience, and long-term availability in smart manufacturing, transportation, healthcare, and Edge AI. Solutions extend to Smart Home (smart cameras, gateways, appliances) and Portable & Wearables (embedded memory balancing performance and low power). As AI and edge computing drive growth, Adata will further integrate Trusta and Adata Industrial technologies, expanding from enterprise infrastructure to intelligent end devices and strengthening its global enterprise market presence.
Vdura Powers AI and HPC Research at New Mexico State University with Data Platform Deployment
1 d ago
Humain, a PIF-backed company delivering full-stack AI capabilities, and MinIO, the data and memory foundation for AI, announced a strategic partnership to co-develop the world’s most advanced AI-native platform, targeting enterprise and sovereign AI deployments in Saudi Arabia and globally. The collaboration combines Humain’s vision for sovereign AI infrastructure with MinIO’s expertise in high-performance object storage and persistent memory, creating a unified foundation for production-scale AI workloads.
Under the agreement, MinIO becomes Humain’s data and memory foundation partner, starting with Humain Fabric, Humain Brain, and Humain One—the models and products layer within the Humain AI Stack. The companies will jointly engineer next-generation AI-native platform capabilities, with MinIO leading the architecture, design, and development of Humain Fabric, an AI-native data platform for enterprise and sovereign AI initiatives. This platform underpins Humain ONE, Humain OS, Humain Brain, and Humain Create.
The partnership rests on a shared conviction: the future of AI will be defined by integrated platforms unifying infrastructure, compute, data, memory, models, agents, and intelligent applications. “AI is only as powerful as the data foundation behind it,” said Tareq Amin, CEO of Humain. “Our partnership with MinIO strengthens Humain with world-class AI-native data infrastructure while accelerating our vision of building an AI platform that enables enterprises, governments, and developers to deploy intelligent systems with confidence.”
Anand Babu (AB) Periasamy, Co-founder and CEO of MinIO, added: “We are at the beginning of a new era in AI, one defined not by individual models but by the platforms that put intelligence to work at scale. MinIO and Humain are co-developing Humain Fabric as that platform: unified, AI-native, and purpose-built for a new class of agentic data engineering.”
Deployments will begin in Saudi Arabia before expanding internationally, supporting organizations needing secure, scalable, production-ready AI infrastructure. The partnership includes joint R&D, strategic engineering collaboration, and global go-to-market efforts focused on advancing AI-native infrastructure for customers worldwide.
QSAN Technology, Inc., an enterprise data management and storage solutions provider, has begun trading its shares on Taiwan’s Emerging Stock Board (ESB), marking a key milestone as the company expands from enterprise storage into enterprise AI solutions to meet surging demand for on-premises AI infrastructure.
The listing comes as generative AI and enterprise-specific LLMs drive demand for high-performance, secure, and scalable data infrastructure. QSAN leverages over two decades of storage expertise to develop integrated hardware and software platforms, including NVMe all-flash storage with proprietary management software, supporting AI training, inference, and RAG workloads. Chairman JP Chen emphasized that in-house R&D across hardware, firmware, OS, and software enables QSAN to help enterprises deploy AI within their own environments, strengthening its position as a Taiwan-developed technology leader.
Strategic partnerships with ASUS, Acer, and Gigabyte expand QSAN’s enterprise AI opportunities, combining storage and data management with server and computing capabilities for global markets. QSAN products are deployed in over 50 countries, and the company has received the SME Innovation Research Award, Rising Star Award, and two D&B Top 1000 SME Elite Awards, recognizing its technology innovation, business performance, and international competitiveness.
Entering the capital market marks the start of QSAN’s next growth phase. The company will continue investing in proprietary technology, strengthening enterprise AI solutions, selectively expanding into strategic global markets, and leveraging partnerships to capture opportunities from enterprise AI adoption.
IBM Completes Acquisition of HRL Laboratories to Accelerate the Future of Quantum
Dell's new Pro 5 14 business laptop, powered by Intel's Core Ultra Series 3 (Panther Lake) SoCs and featuring the first LPCAMM2 memory modules to cross our labs, targets mainstream enterprise users seeking high performance and modularity. The 14-inch Dell Pro 5 14 (model P514260) is built around Intel's top-tier Core Ultra X7 368H processor (4P+8E+4LPE, 5.0GHz) with integrated Arc B390 GPU (Xe3, 12 cores). It supports up to 64GB of LPDDR5X-8533 memory via a single LPCAMM2 module—a key differentiator for upgradeability—and a PCIe Gen5 x4 M.2 2280 SSD (1TB in our review unit). The laptop weighs 1.34 kg and measures 18.1 mm at its thickest point, with an aluminum chassis in magnetite gray. Display options include a WUXGA (1920×1200) 120Hz VRR panel rated at 500 nits and 100% sRGB, with a super-low-power IPS design consuming just 3.35W. Dell also offers WQXGA and OLED alternatives, though none support HDR. Connectivity includes two Thunderbolt 4 40Gbps USB-C ports, HDMI 2.1 (limited to 4K@60Hz without DSC), two USB-A 5Gbps ports, a rare 1GbE RJ45 jack (Intel I219-LM for vPro), and Wi-Fi 7 + Bluetooth 6.0 (Intel BE211). A 70Wh battery powers the system, and charging
Commvault, a leader in unified enterprise resilience, announced a new integration with CrowdStrike that makes Commvault cyber recovery actions available as native steps within Charlotte Agentic SOAR workflows, enabling joint customers to automate recovery actions as part of security workflows to accelerate response and forensic investigations. This AI-driven automation helps organizations detect, investigate, and respond to threats at machine speed, while reducing manual coordination between security and recovery teams.
The purpose-built connector allows security teams to incorporate Commvault recovery actions directly into Charlotte Agentic SOAR workflows. Key capabilities include: restricting access in Commvault to prevent unauthorized changes during active incidents; preserving clean recovery options by automatically suspending backup data aging policies to retain viable recovery points; and accelerating forensic investigations by restoring potentially compromised assets into Commvault Cleanroom for analysis without disrupting production systems.
“Security and recovery teams need to move quickly and in coordination during an incident,” said Vidya Shankaran, Field CTO, Commvault. “Our integration with CrowdStrike Charlotte Agentic SOAR makes Commvault cyber recovery actions available directly within security workflows, helping joint customers reduce manual handoffs and accelerate investigation and response.”
This integration follows previous Commvault-CrowdStrike collaborations: Falcon Insight XDR brought threat intelligence into Commvault Cloud, and Falcon Next-Gen SIEM extended visibility. The new Charlotte Agentic SOAR integration is available for joint customers through the CrowdStrike Marketplace. For more information, visit the Commvault and CrowdStrike joint partner page.
Nvidia Groq 3 LPX, the interactive AI inference accelerator now in full production, delivers world-record token generation speeds for agentic coding and latency-sensitive workloads, with Nebius becoming the first AI cloud to adopt the platform for its Token Factory inference service. An extension of the Nvidia Vera Rubin platform, Groq 3 LPX dramatically increases token generation rates for Vera Rubin NVL72 systems, enabling ultrafast responsiveness for agentic AI that requires massive token volumes across hundreds of inference steps. In Artificial Analysis benchmarking, Groq 3 LPX achieved a record 3,400 output tokens per second running Gemma 4 31B with a 100,000-token context—the fastest performance ever recorded for that model. This enables agentic tasks like coding to complete in minutes versus hours, providing 4x faster responsiveness than the nearest alternative platform. “Inference is the growth engine of AI,” said Jensen Huang, Nvidia CEO, noting that Vera Rubin extends the vision with workload-optimized AI factory configurations for the agentic AI era. Nebius plans to integrate Groq 3 LPX into its production inference platform, giving developers instant token generation through existing APIs without stack migration. Purpose-built AI inference cloud Groq also plans to be among the platform’s earliest adopters. The Vera Rubin platform, codesigned across seven chips and five racks, includes BlueField-4 DPUs, Vera CPU racks, Spectrum-6 SPX Ethernet, and STX storage to optimize multi-agent systems for highest throughput per watt and lowest-latency inference.
The AI infrastructure race has spent the past few years fixated on high-bandwidth memory as one of the most indispensable building blocks of accelerated compute...
Oxmiq Labs presented its High-Bandwidth Flash (HBF) specification for AI inference at Hot Chips 2026, positioning the technology as a capacity-tier alternative to HBM rather than a cheaper substitute. The company argues that HBF delivers 8 to 16 times the capacity of HBM at the same cost, targeting workloads where bandwidth demand is low and memory capacity is the bottleneck.
The HBF hardware specification spans three grades, with maximum user bandwidth ranging from 0.384 TB/s to 3.072 TB/s and UCIe rates climbing from 8 GT/s to 32 GT/s. Capacity tops out at 512 GiB on a 16-high stack. Oxmiq frames HBF within a memory technology landscape using alpha and beta metrics, where beta tracks cost and alpha tracks bandwidth, emphasizing that HBF is a distinct capacity point, not a cheaper HBM variant.
Oxmiq's analysis focuses on serving cost per token, breaking down the economics of hold and feed bandwidth. The company simulated a 72-GPU rack using a decode-centric Kimi-K2 1T model at FP4 with 1M tokens in and 1K out, finding that HBF buys roughly 14x the capacity for about 0.6x the bandwidth at the same rack cost.
Software constraints include 64 KB chunk access for maximum bandwidth, approximately 24 hours of power-on data retention at 85 degrees Celsius, and host-managed lifecycle handling. Since HBF is read-optimized and write-constrained, placement becomes a software problem. Oxmiq proposes a vLLM plugin using HBF in place of host CPU pinned memory for KV cache and MoE expert pools, with a GPU configuration reaching 2.2 TB capacity at 17.4 TB/s peak bandwidth.
The company concludes HBF wins only where bandwidth demand is low, such as MoE models with small batch sizes and long-context sparse KV scenarios, with attention sparsity models like DeepSeek Sparse Attention as genuine fits.
Databricks has acquired Electric, integrating its PGlite WASM Postgres and real-time sync engine to extend Postgres capabilities from the lakehouse to the edge for AI agent sandboxes, addressing the need for distributed state and real-time data synchronization in agentic applications. Announced August 11, 2026, by Databricks engineers Stas Kelvich, Yan Leshinsky, Nikita Shamgunov, and cofounder Reynold Xin, the move brings purpose-built data primitives for agents: PGlite gives each agent a lightweight Postgres database running locally in sandboxes, browsers, or devices, providing ultra-low latency access to context. Electric’s real-time sync engine continuously synchronizes distributed state back to Databricks’ Lakebase, enabling teams of agents to collaborate without stale or conflicting data. PGlite has grown from 1 million to 13 million weekly downloads in 12 months, reflecting demand for embeddable Postgres. Unlike traditional applications with predictable queries, agents decide data needs at runtime, update context multiple times per second, run in sandboxed environments, and work in groups requiring shared, current views. Electric’s sync architecture, similar to collaborative apps like Google Docs and Figma, keeps agents in sync while centralizing control in Lakebase Postgres on cheap, durable object storage. Both Electric and Lakebase are built on Postgres, the open-source database standard for AI agents. PGlite was originally based on WASM Postgres work by Stas Kelvich (co-founder of Neon), which Electric turned into a production-ready embeddable database. The acquisition reunites these efforts and strengthens Databricks’ leadership in modern databases as agentic application demand accelerates. Developers can now build collaborative agentic apps on a single Postgres standard, run Postgres inside agent sandboxes, and keep teams of agents in sync with centralized governance.
Huawei Recognized as a Leader in Gartner Magic Quadrant for Enterprise Storage Platforms, 2026
9 d ago
This new Gartner report, Magic Quadrant for Enterprise Storage Platforms, 2026, belongs to a category of documents that, in the past, many end users and partners considered important when evaluating offerings, preparing an RFP or entering a purchasing process. But after several years of methodology changes and analyst team reshuffling, we increasingly question its real value, or at least the accuracy and consistency of its content. This Enterprise Storage Platforms (ESP) edition reinforces that feeling. But Gartner is a strong name… we expect and we asked several time for some reconsiderations. Our first observation concerns Gartner’s definition of an enterprise storage platform. In the market definition/description section, Gartner states: “Gartner defines enterprise storage platforms (ESPs) as the market consisting of products and services designed to unify support for diverse block, file and object storage workloads and use cases.” Our first reaction is that Gartner appears to associate the enterprise-storage qualification primarily with interfaces, essentially, what the storage platform exposes externally. Reading this paragraph and the mandatory features carefully, we don’t see any meaningful reference to the capacity such an enterprise platform should provide. This immediately raises a question: why are several brands offering these capabilities, sometimes with particularly rich and innovative products, absent from the report? We can mention Vast Data, Lenovo, Fujitsu, Nexsan and StorONE, and potentially others such as Synology, QNAP or Infortrend. Why is IEIT Systems regularly included while other comparable players are not? DDN has also disappeared, without a clear explanation. There is another inconsistency in the inclusion criteria. Gartner mentions “block and either file or object storage services”, which differs from the market definition above, where block, file **and** object are listed together. The first wording suggests that all three interfaces are required, while the inclusion criteria indicate that block plus only one of the other two is sufficient. If only a subset is mandatory, the number of vendors potentially eligible becomes significantly larger. The revenue threshold is also somewhat puzzling: $325M. Why $325M rather than $300M? Is this precise threshold designed to include certain vendors while excluding others? More importantly, obtaining revenue at this level of granularity is always difficult, sometimes even for the vendors themselves. How should revenue be counted for a platform capable of supporting all three interfaces but deployed by a customer with only one enabled? Should the entire product revenue qualify, or only revenue associated with deployments actually using the required interfaces? These are very different interpretations. When the term “enterprise class” is used, it has traditionally implied characteristics such as consolidation capabilities, scalability, connectivity, resilience and significant storage capacity. Yet these dimensions are barely reflected in the methodology. Gartner mentions minimum capacities of 100TB of raw block storage and 500TB of raw file or object storage, thresholds that again raise questions about some of the missing vendors. Gartner also seems to repeat a common confusion by associating primary storage with block storage. They are not synonymous. Primary storage describes the role storage plays within an organization and the mission it serves, not the protocol used to access it. Many organizations rely on file or even object storage as primary storage for critical workloads. At the same time, NVMe is mentioned while NVMe-oF is not, which is surprising for a report focused on enterprise-class platforms. Our readers may remember when SAN support, switch connectivity, host scalability and similar infrastructure characteristics were important criteria for qualifying a product as enterprise storage. And when we read the individual vendor assessments, some of the angles and conclusions taken are difficult to reconcile with Gartner’s own market definition and inclusion criteria. We let readers make their own judgment. Overall, we’re disappointed by this report. In our view, it doesn’t accurately reflect today’s enterprise storage landscape, the diversity of available technologies, or the real requirements and purchasing considerations of enterprise users. The report is available as a reprint from various players who participated to this study.
Everpure, a company revolutionizing storage and data management, announced it has been recognized as a Leader in the 2026 Gartner Magic Quadrant for Enterprise Storage Platforms.For the second consecutive year, Everpure was positioned highest in execution and furthest in vision; this two-time recognition follows 11 times Everpure has been named a Leader in other Gartner Magic Quadrant reports. As enterprises continue to invest heavily in AI models, applications, and compute, many discover that the true barrier scaling AI is fragmented data. To deploy AI successfully, organizations must shift to a Data Primacy architecture – where trusted, governed, and contextualized data, rather than proprietary application-centric data stores, serves as the foundation for agentic workflow. By unifying policy and governance, semantics and context directly with operational data, Everpure transforms enterprise data into an intelligent, trusted, real-time asset. “Data Primacy will be the future of IT architectures in the AI era, and we believe the Gartner recognition validates Everpure’s innovation to make enterprise data truly AI-ready. To scale AI, organizations need an IT architecture where trusted, governed, contextualized data powers every application and AI agent across the enterprise; this is precisely what we enable with the Everpure Platform and Everpure Data Intelligence,” said Charles Giancarlo, CEO, Everpure. Data Management Platform for the AI Era Gartner defines enterprise storage platforms (ESPs) as the market consisting of products and services designed to unify support for diverse block, file and object storage workloads and use cases. ESP products and services include appliances, software-defined storage (SDS) and storage as a service (STaaS). An ESP includes data management and data storage services provided through a centrally managed, multidomain control plane. They enable organizations to leverage AI-powered telemetry for platform structured and unstructured workloads. Everpure believes its positioning validates a broader industry shift toward this model. Everpure’s Data Primacy approach reframes data management for the AI era – reducing operational friction and helping enterprises manage information as a trusted asset across hybrid environments through a global data plane and intelligent control plane. Platform Innovations Built for the AI Era Recent updates to the Everpure Platform enable organizations to modernize infrastructure, automate management, and prepare enterprise data for AI workloads: Everpure Data Intelligence: Makes fragmented data usable for AI by discovering, classifying, contextualizing, and governing data at the source across Everpure, cloud, SaaS, and third-party environments Enterprise Data Cloud Enhancements: Unifies data, policy, and semantics across hybrid estates to streamline operations and enforce consistent governance AI Data Stream: Automates data discovery, preparation, and delivery pipelines, making real-time information immediately accessible to AI applications Cyber Resilience Innovations: Strengthens protection and speeds recovery from cyber threats to ensure continuous business resilience Customer Validation We believe the Gartner recognition is supported by customer feedback on Gartner Peer Insights. As of August 7, 2026, Everpure has an Overall Rating of 4.9 out of 5 stars in the Enterprise Storage Platforms market, with 89% of reviewers willing to recommend the platform, based on 202 reviews. “Easy-to-use Interface Stands out, Support Rated Highly. The intuitive GUI, performance of the Array and support provided by Pure is 5” Technical Lead Infrastructure, Finance “Truly been a game changer for my company. FlashArray has enabled my company to do more than we thought possible, help streamline our operations and minimize our infrastructure spend due to array efficiencies and feature functionality. We have had a 100% uptime on all of our arrays and no data availability issues.” IT Reviewer, IT Services Industry “If you need strong performance and support, Pure is the answer to your storage questions. Pure is not the cheapest, but the performance and the support (pre and post sales) has been world-class.” IT Manager, Manufacturing “Consistent high performance with seamless VMware integration and reliable replication.” Enterprise Storage Reviewer “A game-changer for our infrastructure’s performance. We noticed a significant improvement in the loading speed of databases and modern applications, allowing us to consolidate workloads effectively.” IT Specialist, Enterprise Infrastructure To learn more, access the full 2026 Gartner Magic Quadrant for Enterprise Storage Platforms report here.
OpenAI took the Hot Chips 2026 stage to detail Jalapeño, an in-house inference ASIC and system built with Broadcom, positioning it as the best compute platform for OpenAI’s own inference workloads with superior throughput per kilowatt and latency against NVIDIA GB200 and GB300. The chip, designed around HBM4 and a spatial programming model, moved from initial RTL to tapeout in roughly nine months, with a late 2025 tapeout and Codex running in early 2026.
OpenAI frames Jalapeño as an inference platform rather than a raw accelerator, targeting state-of-the-art performance per watt at low latency for multi-chip workloads. Key metrics are time to last token (user experience) and tokens per joule (inference efficiency). Using the public InferenceX benchmark normalized to package TDP, Jalapeño at 700 watts is compared against GB200 at 1.2 kW, and GB300 and MI355X at 1.4 kW.
On GPT-OSS 120B, Jalapeño shows about 1.9x higher peak mixed tokens per second per kilowatt and roughly 1.7x lower end-to-end latency. On DeepSeek R1 670B MXFP4, gains widen to 1.7x higher peak mixed tokens per kW/s and 3.6x lower latency. On the 1-trillion-parameter Kimi K2.5, Jalapeño delivers about 1.5x higher peak mixed tokens per kW/s and 3.4x lower latency. Even in single-token mode against GB300 multi-token prediction, Jalapeño leads with 1.5x higher peak mixed-token rate per kW and 2.2x lower latency.
Architecturally, Jalapeño uses a single balanced chip where unused blocks are gated, avoiding idle accelerator power draw. Raw HBM4 bandwidth at 128-chip aggregate exceeds one petabyte per second, implying up to 10,000 tokens per second with speculative decoding, though real system rates are lower due to long-latency paths. OpenAI claims sub-millisecond token-to-token latency on frontier models and notes multi-token prediction could add 3x–5x latency improvement at iso-efficiency.
Terrapinn’s acquisition of the Future Memory and Storage (FMS) event delivered a refreshed edition with over 4,000 attendees, a redesigned visual identity, and a conference program dominated by the AI-driven “Memory Wall” challenge, alongside major product launches from Kioxia, Samsung, Micron, Sandisk, and SK hynix. The event team preserved the strong expo floor and rich conference agenda while introducing improvements. Notable exhibitor absences included Nvidia (which still keynoted), Phison, Swissbit, and Panmnesia; newcomers like Primemas joined returning names such as ScaleFlux, Winbond, and YMTC. Five core themes emerged: AI inference and the evolving memory/storage hierarchy; new NAND technologies (BiCS10, V-NAND, XL-Flash, HBF); PCIe Gen 5 and Gen 6 connectivity; CXL 3.2 memory pooling and sharing; and energy consumption/cooling. Key announcements: Kioxia unveiled the CM10 Series, the first PCIe Gen 6 enterprise SSD built on 332-layer BiCS FLASH Generation 10 TLC, plus GP1 ultra-high-IOPS SSDs and next-gen E1.S drives. Samsung outlined its 3D Memory vision for AI infrastructure and displayed 256TB BM1773 SSDs (E3.S/U.2), PM1763, V-NAND V10, HBM4, and CXL modules. Micron demonstrated with Astera Labs and Microchip. Sandisk showcased HBF, BiCS10 NAND, and an UltraQLC E3.S 256TB prototype; with SK hynix, it released the first OCP Technical Specification for High Bandwidth Flash standardization. SK hynix highlighted HBM, NAND, SSDs, and CXL memory. Silicon Motion introduced the MonTitan SSD Reference Design Kit for AI and agentic AI storage. FADU,
Dat
11 d ago
Dat
Samsung High Bandwidth Memory (HBM) at Hot Chips 2026: Samsung is presenting "Evolving HBM Base Die," detailing plans to transform the base die from a passive interposer into a more capable, SoC-like component for next-generation AI chips and data center memory. The talk covers how HBM splits work between stacked DRAM core dies (C-die) and the base die (B-die), which carries PHY and through-silicon vias (TSVs) for communication with the compute die. Samsung traces HBM bandwidth and capacity growth from ~1 GB/1 TB/s in original HBM to over 60 GB/6 TB/s in HBM5, noting that rising bandwidth forces fundamental base die changes.
Key scaling limits include TSV pitch, I/O count, and PHY speed. Samsung shifts the B-die from older logic to 4nm starting with HBM4, enabling advanced logic integration for power reduction and active area minimization. The company introduces custom HBM (cHBM) using advanced logic to offload XPU functions onto the B-die. Phase 1 reclaims XPU area by replacing traditional HBM PHY with a die-to-die interface, shrinking footprint and improving energy efficiency, while a Heat Path Block reduces peak temperature by over 35% as power density climbs to 2.0 W/mm². Memory controller offloading and SRAM-based cell repair further enhance reliability.
Phase 2 adds SoC-level RAS with thermal, voltage, and aging sensors, plus on-chip self-test. It also enables external memory expansion via the B-die shoreline and offloads processing elements to cut die-to-die bandwidth. Phase 3 introduces zHBM, a true 3D vertical integration of XPU and C-die stack, eliminating the 2.5D interposer for ultra-low power consumption. Samsung targets these innovations to address AI model capacity demands, with context windows expanding ~30x per year, making high-capacity, high-bandwidth memory critical for next-gen AI inference.
Bosgame M5 AI Mini Desktop review highlights the AMD Ryzen AI Max+ 395 system with 128GB LPDDR5X memory, offering a cost-effective alternative to the Ryzen AI Halo platform at $500-$1,100 less. This AI mini PC targets enthusiasts seeking high-performance computing, AI workloads, and compact desktop solutions with extensive connectivity.
The Bosgame M5 features a sleek angled design with vertical mounting capability via included metal stand. Front I/O includes audio jack, dual USB 3.2 Gen2 10Gbps Type-A ports, USB4 port, SD card reader, and performance mode switch with three power settings. RGB light strip adds aesthetic appeal. Top and bottom vents require clearance for proper cooling.
Connectivity remains robust with rear ports including dual USB 2.0, HDMI, DisplayPort, additional USB4, USB 3.2 Gen2 Type-A, and audio jack. The Realtek 2.5GbE LAN port is noted as the system's primary weakness, with reviewers suggesting 10GbE would better compete with other Strix Halo systems. Users can add USB 10GbE NIC adapters for faster networking.
Internal access via bottom hatch enables upgrades. The system's aggressive venting design prevents stacking objects on top. This AI mini desktop targets professionals needing powerful AMD AI processing, memory-intensive applications, and versatile port selection in compact form factor. The Bosgame M5 delivers strong value proposition for AI development, content creation, and high-performance computing workloads, though networking enthusiasts may require additional hardware for maximum throughput.
Intel is exploring a return to memory technology, with CEO Lip-Bu Tan arguing that AI-driven bandwidth and latency demands are transforming memory from a commodity into a critical system-performance differentiator. This potential strategic pivot, decades after Intel exited mainstream DRAM and sold its NAND business to SK Hynix, could redefine the company's role in AI chip architecture and advanced packaging.
Founded in 1968 as a memory company, Intel's history includes early SRAM and DRAM production before shifting to microprocessors. Its later ventures—NAND flash and 3D XPoint-based Optane—were discontinued or divested, with the NAND unit becoming Solidigm. A new approach would likely avoid competing directly with memory giants Samsung, SK Hynix, and Micron in conventional DRAM or NAND. Instead, Intel's opportunity lies in integrating memory with processors through advanced packaging technologies like EMIB, EMIB-T, and Foveros Direct, which enable high-bandwidth, low-power data movement for AI accelerators.
Tan has proposed stacking memory directly with processors, potentially using HBM stacks, bonded cache dies, or specialized chiplets. Intel's 18A-PT process supports sub-5-micron interconnect pitches, and the hiring of former SK Hynix CEO Seok-Hee Lee adds credibility. However, no product, fabrication plan, or timeline has been announced. A full return to memory manufacturing would require billions in investment and risk distracting from Intel's foundry expansion. The more plausible strategy is a systems-engineering approach: co-designing memory interfaces, integrating third-party HBM, and optimizing packaging to keep AI processors supplied with data. This could make memory central to Intel's identity again without repeating past commodity-market battles.
Forrester Research’s new “The Forrester Wave: Object Storage Solutions, Q2 2026” report is drawing sharp criticism for omitting key industry pioneers like Cloudian, IBM, MinIO, and DataCore, raising questions about the report’s inclusion criteria and market accuracy. The analysis highlights that while established object storage vendors such as DDN, Dell, Hitachi Vantara, Huawei, NetApp, Scality, and VAST Data are featured, the absence of foundational players—including IBM’s Cleversafe-based COS, MinIO’s open-source platform, and DataCore’s Swarm (formerly Caringo)—undermines the report’s credibility. Forrester’s inclusion criteria, particularly around object-storage-specific revenue attribution and the "flat namespace" requirement, appear to favor vendors with broader unified or multiprotocol storage platforms, creating an inconsistent competitive landscape. The report also mixes on-premises products with cloud services, such as Oracle, and includes vendors whose object interfaces are merely S3-compatible add-ons rather than true object storage architectures. This approach, the critique argues, fails to reflect the true market dynamics and technical distinctions within the storage industry. The omission of notable open-source distributions like Ceph and Quantum’s ObjectScale further compounds the issue. While acknowledging the difficulty of crafting such reports, the analysis urges Forrester to relax its revenue thresholds and broaden vendor inclusion in future editions to better represent the evolving object storage market, which is critical for enterprise IT buyers evaluating data storage, AI workloads, and hybrid cloud strategies.
Cloudera’s new global survey, *The Great AI Re-Architecture*, reveals that legacy data architectures are forcing enterprises to overhaul IT infrastructure to meet the demands of scalable and secure AI, marking a fundamental shift toward hybrid cloud environments. The report, based on responses from 1,500 enterprise architects and data leaders, shows that while 77% of organizations actively use AI, 95% have delayed or canceled AI initiatives in the past year due to data governance, compliance, or regulatory hurdles. Consequently, 72% state their current data architecture requires significant changes to support future AI requirements, underscoring that existing infrastructure was not built for modern AI workloads.
The findings highlight a mass transition from legacy systems to hybrid architectures, driven by the need to bring trusted AI to data wherever it resides. Key data points show that 75% of respondents say AI has altered their data storage and architecture practices, while 84% report increased infrastructure costs from AI workloads. Governance remains a critical challenge, with 73% noting AI has made data governance more complex and 55% delaying over six AI projects in the past year due to compliance issues. As data becomes increasingly distributed—97% move data between environments monthly—consistent governance across cloud, private cloud, and on-premises is essential.
The survey also signals a shift away from public cloud dominance, as 66% of organizations have moved AI workloads back to private or on-premises infrastructure. Looking ahead, 25% plan to prioritize a hybrid-first architecture over the next two years, emphasizing flexibility over a single deployment model. Cloudera’s CTO, Sergio Gago, notes that success depends on building a data foundation that allows AI to run where it performs best without compromising control or security. The research was conducted by Wakefield Research across nine markets between June 5-22, 2026.