Entertainment and media-focused storage supplier 45Drives announced a major expansion of SnapShield, its server-side cybersecurity platform. SnapShield uses real-time behavioral analysis at the storage server to detect ransomware-like activity. New Data Exfiltration Protection spots suspicious file access indicating potential data theft, while a new Centralized Management System provides enterprises and managed service providers a single pane to secure and monitor multiple servers, sites and customer environments.
Lakehouse analytics provider Cloudera inked a partnership with French AI lab Mistral. It integrates Cloudera’s hybrid data and AI platform with Mistral’s sovereign AI models, letting enterprises securely build, customize and run AI on private data. Mistral’s cutting-edge models and tools will connect to Cloudera’s platform, enabling customers to deploy custom AI models and applications across 30 exabytes of data. Enterprises gain deployment flexibility across public, private, on-premises and fully air-gapped environments while retaining consistent governance and data control. The partnership covers Mistral’s full model lineup including reasoning, chat, coding, document intelligence and voice, expanding Cloudera users’ AI options for enterprise data.
Commvault appointed James Hayes as its first Chief Public Affairs Officer, heading government and public affairs strategy for federal, state, international agencies and regulators. Hayes joins from IonQ and Tenable, where he managed federal advocacy, cybersecurity policy and emerging tech standard engagement. Before Dell acquired EMC, he ran EMC’s Washington DC Government Affairs Office.
CoreWeave launched Physical AI Field Engineering. The service pairs customer teams with domain specialists to build, validate and deploy AI across the full engineering lifecycle, from R&D to field operations. CoreWeave engineers collaborate with client teams to train models using existing customer data such as bench test results, simulation outputs, production sensors and live telemetry. Each model is validated against real system physics for practical reliability. Key components:
Strategy: Prioritize viable AI use cases and valuable datasets.
Simulation infrastructure: Deploy GPU, storage and simulation stacks for target workloads. Field engineers connect clients to CoreWeave’s broader Physical AI platform for large-scale full infrastructure design.
Real-world data: Convert scattered test, sensor and production data into predictive, anomaly-detection or fault-explanation models instead of leaving data signals unused.
Agentic learning: Translate model insights into physical-world actions — system recalibration, pre-emptive fault correction or robots executing custom trained skills built by client teams.
The Information reports China’s CXMT has begun limited production of HBM3E memory chips and plans volume expansion in 2027, marking a major milestone for China’s semiconductor sector.
d-Matrix announced a new multi-year collaboration with Nvidia, bringing its Raptor XPU into Nvidia’s widely adopted AI factory ecosystem. d-Matrix builds custom XPUs for inference decode, while Nvidia GPUs handle profiling. Raptor leverages SRAM plus stacked 3D DRAM atop compute. SRAM capacity scales well but cannot fully support the largest reasoning models and rising token consumption trends.
Its stacked 3D DRAM sits within die-to-die interconnect space. Using 3DIMC, it delivers 10× higher memory bandwidth and 10× better energy efficiency versus HBM4. 3DIMC stacks compute/logic dies face-to-face with custom DRAM dies via microbumps, removing traditional HBM PHY, interposer and long-distance data movement. d-Matrix stated: “We repurposed passive DRAM capacitor layers into an active stack. This exposes every small DRAM bank directly to the compute engine. With 3D DRAM, we can use the entire 3D surface for connections, running signals at hundreds of MHz at the DRAM base. This achieves 20 TB/s bandwidth per stack — 10× HBM4 — at 0.3–0.4 pJ/bit versus 3–4 pJ/bit.”
d-Matrix Raptor is scheduled to tape out before year-end, designed for NVLink Fusion and MGX rack-scale ecosystem integration. Hyperscalers and frontier AI labs are evaluating its unique memory stacking design; the product is backed by over 100 patents.
Fortanix announced a partnership with CrowdStrike to secure AI across its full lifecycle. Combining CrowdStrike Falcon and Fortanix Confidential AI, the integrated system delivers end-to-end AI security: threat detection and response, plus protection for AI workloads, sensitive data and proprietary models active in memory. Fortanix Confidential AI creates trusted execution environments where AI workloads run within isolated, hardware-protected memory. Sensitive enterprise data, prompts, model weights and inference outputs stay encrypted in use, preventing exposure to cloud operators, privileged admins or malicious actors. Together, CrowdStrike and Fortanix deliver holistic AI workload security covering data, models and agents before, during and after execution.
SaaS data protection vendor Keepit has surpassed $100 million in annual recurring revenue. Its AI Truth Cloud strategy, which redefines data protection as the trusted foundation for AI-powered enterprises rather than mere loss insurance, drives broad customer adoption. Since growth equity firm One Peak’s initial investment in September 2020, the firm has grown more than 13 times with a CAGR above 55%, supporting organizations of all sizes from small single-user businesses to global enterprises on its independent, vendor-agnostic cloud infrastructure.
“Backup used to answer one question: Can we get the data back?” said Keepit co-founder and chief visionary officer Frederik Schouboe. “AI is making companies ask something bigger: Can we trust the data our business is running on? This means data protection is no longer a business cost, but a trusted business foundation.” The same infrastructure that mitigates ransomware and accidental data loss now provides enterprises a trusted data source for confident AI operation.
Reuters reports Micron is offering Taiwan employees severance packages worth 35 to 68 months of salary for fiscal 2026, with a minimum cash payout of NT$1.7 million. Local unions stated 80% of Micron’s Taiwan workers supported strike action, arguing existing bonus schemes fail to reflect record AI-driven revenue. Samsung and SK Hynix have rolled out similar compensation moves.
Open-source database vendor Percona announced a partnership with open-source observability firm Coroot to launch Coroot Percona Edition for Percona customers. Percona and Coroot deliver full-stack observability for cloud-native workloads instead of separate database or cloud infrastructure monitoring. Coroot automatically traces issues across stacks down to root cause and can open Percona support tickets directly for database-related findings.
Vector database and search provider Pinecone announced general availability of full-text search within Pinecone Database. It adds BM25 keyword ranking and text-match filters alongside dense vectors in the same index, letting agents retrieve exact order numbers or error codes missed by embeddings without separate systems to maintain. For search, recommendation, RAG or agent workloads mixing identifiers and exact phrases, Full-Text Search delivers:
Precise matches for queries embeddings get wrong: SKUs, error codes, case numbers, order IDs and exact quoted phrases.
BM25 keyword ranking across multiple text fields per index, supporting Lucene query syntax, fuzzy matching, plus tokenization and stemming for 18 languages.
Text-match filters to narrow semantic search results to matching documents. One query retrieves by identifier while ranking remaining results semantically.
Unified index and schema for text fields, dense vectors and sparse vectors, queried via the Documents API; no cluster tuning, patching or upgrades required.
Usage-based capacity by default, metered using the same read and write units as vector indexes. Dedicated Read Nodes are available for provisioned read capacity, alongside BYOC deployment for operation inside customers’ own cloud environments.
SK Hynix CEO Kwak Noh-Jung commented: "There are no clear signs of a memory downturn, so this situation will continue through the end of 2030. Even if a downturn comes, it will be closer to a gradual decline or flat trend rather than a steep drop. After 2030, I believe supply and demand will find balance and the entire AI industry can grow together."
Beijing Qianxing Jietong Technology Co., Ltd.
Sandy Yang/Global Strategy Director
WhatsApp / WeChat: +86 13426366826
Email: yangyd@qianxingdata.com
Website: www.qianxingdata.com/www.storagesserver.com
Business Focus:
ICT Product Distribution/System Integration & Services/Infrastructure Solutions
With 20+ years of IT distribution experience, we partner with leading global brands to deliver reliable products and professional services.
“Using Technology to Build an Intelligent World”Your Trusted ICT Product Service Provider!
আপনার বার্তাটি 20-3,000 টির মধ্যে হতে হবে!
Bengali
