Backblaze has appointed Dan Spraggins as SVP of Engineering to scale its global storage cloud. He now oversees Backblaze’s engineering organization, architecting the distributed systems needed to handle the exponential data demands of AI and machine learning workloads.
…
Cohesity has hired Julien Mousqueton as Field Chief Information Security Officer (CISO) for Europe. He joins Cohesity following a ten-year tenure at Computacenter, most recently serving as CTO, directing technology strategy and innovation with a focus on cyber security and resilience. Mousqueton also serves as a reservist advisor for the Office Anti-Cybercriminalité (OFAC) in France, a specialist agency within the French National Police responsible for investigating cybercrime.
…
Commvault is partnering with Delinea, adding its privileged access management into the Commvault Cloud platform, strengthening security, supporting compliance initiatives, and streamlining data recovery. Commvault has also announced an agreement with Pinecone to help enterprises protect and recover vector retrieval workloads that power retrieval-augmented generation (RAG) and other AI use cases. We’re told it enables immutable backup, point-in-time-recovery (PITR), and customisable extended retention of vector data – all without impacting query latency.
…
The CTO Advisor, Keith Townsend, has devized a 4+1 AI infrastructure framework to help enterprises understand, design, and implement AI application stacks. The layers are:
- Layer 0: Compute and network fabric
- Layer 1: Data Plane – data storage and governance, context management and retrieval, and data movement and pipelines.
- Layer 2: Application Plane — infrastructure orchestration, application runtime, amd agentic infrastructure.
- Layer 3: AI application layer.
The +1 designation emphasizes that Layer 3 (Agent Applications) is the Value Plane; it consumes the lower layers to deliver business outcomes. Townsend cites VAST Data, Databricks, and Snowflake Arctic as representative layer 1 data storage and governance technologies. Weaviate, pgvector and FAISS/ScaNN are examples of context management and retrieval technologies. Dataflow, Fivetran, and Airflow are representative data movement and pipeline technologies.
…
Reuters reported Chinese DRAM maker ChangXin Memory Technologies (CXMT) is planning a $4.2 billion Shangfhai stock market IPO and is aiming to produce high-bandwidth memory (HBM) by the end of 2026. CXMT held a share of 4 percent of the global DRAM market in the second quarter, while Micron, SK hynix and Samsung together controlled more than 90 percent, according to data from research firm Omdia cited in the IPO prospectus.
…
Dell is supporting Nvidia’s Inference Context Memory Storage Platform (ICMSP) with its PowerScale, ObjectScale and Project Lightning (in private preview) storage products to offer scalable and performant KV Cache offload. It is integrating technologies including LMCache and Nvidia NIXL (NVIDIA Inference Transfer Library) into these storage engines to turn them into a high-speed extension of GPU memory. More details are available here.
…
Contract HDD prices rose 4 percent quarter-on-quarter, and to a 2-year high, towards the end of 2025 as US CSPs increased storage capacity in their data centres, focusing on nearline drives. Demand for SSDs is also rising, and rising spectacularly for both DRAM and HBM. Samsung and SK hynix are pulling cleanroom fab building plans forward as a result.
…
IBM is working with Nvidia to combine advanced storage and networking technologies for LLM inference. The pair say that integrating IBM Storage Scale’s global namespace and locality-aware placement with the Nvidia Inference Context Memory Storage Platform – powered by BlueField-4 on the Nvidia Rubin platform – delivers low-latency KV cache access, efficient resource utilization, and reduced TCO. This new IBM Storage Scale solution is purpose built for next-generation AI deployments on Nvidia Dynamo. Read more in an IBM blog.
…
Kioxia announced the DRAM-less BG7 PCIe gen 4 SSD using BiCS8 218-layer 3D NAND. With random read/write up to 1,000,000 IOPS and sequential read up to 7,000 MB/ps, it delivers approximately 10 percent and 16 percent performance improvement respectively compared to its BG6 Series predecessor, plus a 67 percent efficiency increase in sequential writes operations. It comes in M.2 2230, 2242 and 2280 form factors. There are 256 GB, 512 GB, 1024 GB, and 2048 GB capacities. Select PC OEM customers are evaluating samples now.
…
Lightbits Labs, which produces fast block storage SW, has product launch plans for 2026, according to CEO Eran Kirzner talking on a CUBE podcast.
…
OWC announced an 8TB version of its OWC Envoy Ultra Thunderbolt 5 SSD, plus a major capacity expansion for its ThunderBlade X12 production shuttle RAID SSD – doubling max capacity to 192TBs. This makes the ThunderBlade X12 the first and only Thunderbolt 5 RAID shuttle drive to deliver this level of capacity and performance in a compact form factor.
…
At the end of 2025, Bloomberg reported “Data storage companies dominated the S&P 500 Index in 2025, with Sandisk shares soaring almost 580% to make them the benchmark’s best performer, with Western Digital in second and Seagate in fourth.” WD stock value increased 278.27 percent in 2025 while Seagate’ stock’s makret cap rose 222.68 percent. The S&P 500 Index itself went up just 14.78 percent in the year. [Micron, not mentioned by Bloomberg, was the third best performer, with a 245 percent rise.]
…
A Korean media report suggests SK hynix is working with Nvidia and Phison to develop a 100 million IOPS version of its 2027 AIN P high-performance SSD. An internal Nvidia “Storage Next” project is mentioned, with a prototype possible later this year.
…
SK hynix is showing 16-Hi HBM4 with 48GB for the first time at CES 2026 in Las Vegas. This is a follow-on to its 12-layer HBM4 product with 36GB, which demonstrated the industry’s fastest speed of 11.7Gbps, and will also be exhibited. It will also will present customized cHBM optimized for specific AI chips or systems, PIM-based AiMX (Accelerator-in-Memory based Accelerator), CuD (Compute-using DRAM) which conducts computing in memory, CMM-Ax that integrates computing capabilities into CXL memory, and its Data-aware CSD (Computational Storage Drive). The cHBM technology integrates some functions located in GPUs and ASICs into the HBM base die.
…
Snowflake has expanded its collaboration with Google Cloud to bring Google’s Gemini 3 models natively into Snowflake Cortex AI, enabling customers to build and scale generative AI applications, including intelligent data agents, without moving or copying data. Snowflake and Google Cloud are deepening go to market alignment through joint sales, co-sell opportunities, and availability via Google Cloud Marketplace, while continuing global expansion with new Snowflake deployments on Google Cloud in Saudi Arabia and Melbourne in Australia. Enterprises such as BlackLine and Fivetran are already using the integration.
…
Strada’s peer-to-peer collaboration platform enables video content professionals to access, share, and review large video files stored on local drives anywhere in the world without needing to upload anything to the cloud. In mid-December it announced a significant investment from Other World Computing (OWC). Over the coming months, OWC and Strada will commence co-marketing initiatives, including events and social media collaboration, and will extend special incentives to customers who purchase packages of OWC and Strada products.
…
TrendForce tabulated the top five enterprise SSD suppliers in 2025’s third quarter;
It said: “The enterprise SSD market experienced robust growth in the third quarter of 2025. TrendForce’s latest reports reveal that AI demand continued to expand from training to inference tasks. Meanwhile, North American CSPs simultaneously scaled out both AI infrastructure and general-purpose servers. This led to a significant boost in enterprise SSD shipments and prices, resulting in a 28 percent increase in combined revenue for the top five brands that reached approximately US$6.54 billion and set a new record for the year.”
…
ReRAM developer Weebit Nano has licensed its ReRAM technology to Texas Instruments, the number one analog semiconductor vendor and a major supplier to the automotive market. TI will manufacture devices with Weebit ReRAM at advanced nodes for embedded processing applications.
…
China’s YMTC may be planning high-bandwidth flash (HBF) technology using its Xtacking 3D NAND string stacking methodology.
…
