BANDF AD
BestBrokers has collated data from Pitchbook, Crunchbase and more, and provided a table of the biggest funding rounds in AI in September 2025:
- Anthropic – $13 billion, $183 billion valuation
- Databricks – $1 billion, $100 billion valuation
- Groq – $750 million, $6.9 billion valuation
- Cognition – $400 million, $10.2 billion valuation
- Sierra – $350 million, $10 billion valuation
- Modular – $250 million, $1.6 billion valuation
- Lila Sciences – $235 million, $1.23 billion valuation
- Perplexity – $200 million, $20 billion valuation
- Baseten – $150 million, $2.15 billion valuation
- Upscale AI – $100 million, unknown valuation
- Invisible Technologies – $100 million, $2 billion valuation
- You.com – $100 million, $1.5 billion valuation
BANDF AD
It also has a graphic showing what it believes to be the largest AI funding rounds in the whole of 2025 so far:
…
Databricks launched Data Intelligence for Cybersecurity to help organizations defend against modern and AI-driven threats with more accuracy, stronger governance and greater flexibility. Data Intelligence for Cybersecurity seamlessly integrates with enterprises’ existing security stacks, unifying all data and leveraging an open partner ecosystem so security teams can fully harness the power of AI – spotting risks earlier, understanding the full context of an attack and responding with greater speed. Building on this foundation, Databricks Agent Bricks enables enterprises to build AI apps and agents that not only accurately analyze their data but also take safely-governed actions across every step of the security workflow. With this release, Databricks is also introducing partner integrations with leading providers, including Abnormal AI, Accenture Federal, ActiveFence, Alpha Level, Arctic Wolf, BigID, DataBahn, DataNimbus, Deloitte, Entrada, Obsidian Security, Panther, PointGuard AI, Rearc, SPLX, Theom, Varonis, and ziggiz, extending the power of Databricks and helping customers drive unified, measurable outcomes in their cybersecurity defense strategies.
…
BANDF AD
A blog by Dell’s Drew Schulke, VP of Product Management, explains Dell’s views on QLC flash. It says “Done correctly, QLC delivers equivalent performance to TLC while meeting the high bar of reliability primary storage customers demand. With >20 percent of PowerStore’s shipped capacity leveraging QLC, and the introduction of QLC on our mission-critical PowerMax platform we’ve shattered the legacy narrative, and our competition has some explaining to do.”
“The truth on QLC performance is that it can deliver sub-millisecond latency – the kind of performance expected for demanding transactional workloads. Why some vendors choose to deliver QLC with 2-4 millisecond response times is a question for them to answer.”
“Next month you will see new QLC-based offerings on both our PowerStore and PowerMax product lines. In both we will make available to a larger number of customers the TCO benefits of QLC with the help of our software-based intellectual property.”
…
BANDF AD
GoodData introduced a new AI platform combining AI Lake, AI Hub, and AI Apps into a comprehensive data intelligence system. AI Lake processes structured and unstructured data into a self-learning layer for accurate AI outputs. AI Hub provides tools for workflow management and compliance. AI Apps integrate secure AI agents into business processes. The platform offers governance, scalability, and flexible integration with open APIs and SDKs, allowing custom AI solutions. GoodData aims to help enterprises turn data into actionable insights, moving beyond traditional analytics to support AI-driven business applications with transparency and control.
…
Fortanix (data security for an AI world) and BigID (data security, privacy, compliance, and AI governance) announced an integration between Fortanix Data Security Manager (DSM) and BigID’s data discovery and classification capabilities to automatically trigger protection actions whenever sensitive data is classified. This streamlined workflow eliminates manual intervention while maintaining complete audit trails of all data protection activities.
…
BANDF AD
SaaS data protector HYCU announced the findings of the newly released HYCU State of SaaS Resilience Report 2025, an independent global survey of 500 IT business decision-makers. According to survey respondents, organizations now use an average of 139 SaaS applications, with many deploying well over 200. The introduction of every new app creates new data, new permissions, and new vulnerabilities. As SaaS portfolios grow, so does the exposure.
- 65 percent of organizations were hit by a SaaS-related breach in the last 12 months with $405,770 as the average daily cost of SaaS downtime, adding up to $2.3 million over a 5-day recovery period.
- 87 percent admit they have at least one SaaS application at risk due to inadequate protection.
- Only 56 percent of SaaS applications are under the control of IT.
- 43 percent of respondents admit no one truly owns SaaS data resilience.
- 70 percent do not perform policy-driven backups for some apps .
- 74 percent do not have offsite data retention.
- 75 percent do not test resilience regularly.
This leaves the vast majority underprepared when cyber threats hit, integrations break, minor disruptions or data is deleted, accidentally or maliciously. Download the report here.
…
Index Engines will demonstrate ransomware-detecting CyberSense at IBM TechXchange 2025, October 6-9 in Orlando. It integrates with IBM FlashSystems’ immutable snapshots, using AI forensic analysis to detect ransomware corruption and support reliable recovery. It validates snapshots, reduces reinfection risks, and shortens restoration times. Sessions cover:
- October 7, 10:30-11:30 AM ET: Validated immutable storage recovery.
- October 7, 4:00-4:20 PM ET: FlashSystem integration and automation.
- October 8, 3:00-3:20 PM ET: 99.99 percent SLA for healthcare systems like EPIC.
…
InfluxData has released InfluxDB 3.5, available on both its Core and Enterprise products, and introducing:
- Explorer dashboards (beta): Designed for visualizing, querying, and managing data stored in InfluxDB 3 Core and Enterprise, the InfluxDB 3 Explorer UI now supports Dashboards, giving users the ability to save and revisit preferred queries in one place.
- Cache querying: Users can now perform ad hoc queries against InfluxDB 3’s built-in Last Value and Distinct Value Caches from Explorer to support exploratory analysis and everyday workflows.
- Ops upgrades: New features to simplify cluster management and improve oversight, as well as general improvements for production deployments.
This release fuels more powerful, controlled workspaces for time series monitoring and analysis.
…
Copenhagen-based SaaS data protector Keepit has secured $60 million in upsized and refinanced credit facilities from the Export and Investment Fund of Denmark (EIFO) and HSBC Innovation Banking. This includes $20 million in new funding and $40 million in refinanced existing facilities, building on a $50 million funding round from December 2024. The funds will support Keepit’s growth, product development, and expansion into larger client markets. Keepit provides secure, vendor-neutral cloud storage to ensure data access, compliance, and recovery for over 18,000 global customers.
…
Kioxia and Sandisk announced the operational start of Fab2 at the Kitakami Plant in Iwate, Japan, on September 30, 2025. The facility produces eighth-generation, 218-layer 3D flash memory using CMOS directly Bonded to Array technology to meet AI-driven storage demand. Production will scale up gradually, with significant output expected by mid-2026. Fab2 features an earthquake-resistant design, energy-efficient equipment, and AI-enhanced production processes. Part of the investment is subsidized by the Japanese government. The facility strengthens the 20-year Kioxia-Sandisk partnership, focusing on advanced 3D flash memory development to support applications like smartphones, data centers, and AI systems.
…
UK-based hyperscaler AI data center builder Nscale has closed a $433 million Pre-Series C SAFE, reinforcing commercial momentum and significant investor support following its recent $1.1 billion Series B. Backed by key investors such as Blue Owl Managed Funds, Dell, Nvidia and Nokia, along with other existing Series B and new investors, the funding underscores confidence in Nscale’s execution and growth. The close of the Pre-Series C SAFE comes only days after Nscale announced the largest Series B funding round in European history, raising $1.1 billion. “We’re overwhelmed by the interest we’ve received. It’s incredible to see the passion and confidence we have in Nscale is matched by key investors,” said Josh Payne, CEO and founder of Nscale. “This commitment to participating in our pre-Series C SAFE, just days after the close of our Series B funding, represents a powerful endorsement of our vision to deliver sovereign, scalable infrastructure for the AI era.”
…
Netlist has initiated new legal proceedings before the United States International Trade Commission (ITC) seeking exclusion and cease and desist orders against Samsung, Google, and Supermicro. The ITC investigation is based on the infringement of six Netlist patents. C.K. Hong, Netlist’s CEO, said: “Since its founding in 2000, Netlist has pioneered innovations in advanced memory technologies. With this action, Netlist is seeking remedial orders that direct US Customs and Border Protection to stop Samsung memory products that infringe on Netlist’s intellectual property from entering the country.”
Netlist asked the ITC to investigate whether Samsung, Google, and Supermicro infringe US Patent Nos. 12,737,366, 10,025,731, 10,268,608, 10,217,523, 9,824,035, and 12,308,087. Each of these patents reads on one or more of the following products: DDR5 memory modules, e.g. DDR5 RDIMM, UDIMM, SODIMM, and MRDIMM, and high-bandwidth memory (HBM).
…
Other World Computing (OWC) released SoftRAID 8.6, compatible with macOS 26. The update introduces SMART over USB support, allowing health monitoring of most USB and Thunderbolt-connected drives to detect potential failures early. It supports RAID 0, 1, 4, 5, and 1+0 configurations for flexible storage setups and enables sharing of Apple-formatted APFS and HFS+ volumes between macOS and Windows. The free Standard version offers basic data access and compatibility updates, while the Premium version adds advanced RAID options and health monitoring with alerts. SoftRAID 8.6 is available for download, with automatic updates for existing users.
…
Quantum announced its Scalar i7 RAPTOR tape library has won a Best of Show Award from TV Tech at IBC 2025 in Amsterdam. It was selected for delivering breakthrough advancements in archive scalability, density, and cyber resilience, making it a preferred solution for large-scale media archives. RAPTOR provides up to 2,016 slots and more than 60 PB of native storage with LTO-10 media offering unparalleled scalability. By delivering up to 200 percent more storage density than competing systems, i7 RAPTOR significantly reduces power, cooling, and floor space requirements, enabling up to 70 percent lower total cost of ownership. This marks the third Best of Show recognition for Scalar i7 RAPTOR in 2025. Earlier this year at NAB, the system received Best of Show honors from both TV Tech and TVBEurope.
…
Reuters reports Samsung Electronics and SK hynix have signed letters of intent to supply memory chips for OpenAI’s Stargate data centers. South Korea presidential adviser Kim Yong-beom said OpenAI was seeking to order 900,000 semiconductor wafers in 2029, and planned to set up joint ventures with Samsung and SK hynix to build two data centers in South Korea with an initial capacity of 20 megawatts. Samsung Electronics’ affiliate Samsung SDS signed a partnership with OpenAI to develop, build and operate AI data centers under the Stargate project, while also expanding enterprise AI services.Shipbuilder Samsung Heavy Industries and construction unit Samsung C&T will jointly work with OpenAI to develop floating offshore data centers to cut cooling costs and carbon emissions.
…
Sandisk announced licensed storage for Asus ROG Xbox Ally (X) handhelds at Tokyo Game Show: MicroSD cards in 512 GB, 1 TB, and 2 TB capacities with up to 200 MB/s reads via QuickFlow tech (limited to 104 MB/s in standard readers), and WD_BLACK SN7100X NVMe SSDs in 2 TB and 4 TB for internal upgrades. The SSD, akin to the existing SN7100, is DRAM-less and power-efficient for PCIe 4.0. MicroSDs are pre-orderable; SSDs arrive later.
…
Solidigm has a new AI Central Lab located near its HQ in Rancho Cordova, CA., with high-performing, storage-dense clusters to handle AI workloads in an environment similar to global large-scale data centers. Hardware highlights include:
- Highest-performing storage test cluster with D7-PS1010 SSD, which achieved the highest-ever per-node throughput measured in the MLPerf Storage (AI model training) test at 116 GB/s. This can be flexibly scaled to multiple nodes.
- Most dense storage test cluster with 192 x D5-P5336 122 TB SSDs packing 23.6 PB into 16U.
- Nvidia B200 and H200 GPUs, 800 Gbps Ethernet networking, and storage servers from leading vendors.
Workloads available for testing at the lab include:
- AI-specific and emerging workloads such as GPU-intensive AI model training and inference, KV cache offload and VectorDB tuning.
- Benchmarking power consumption of different configurations.
- Feeding data to GPUs, keeping them as busy as possible.
- Helping translate SSD specifications into system-level, industry-relevant AI efficiency metrics such as tokens per dollar and tokens per watt.
…
America’s Test Kitchen, a longstanding StorONE enterprise customer, is benefiting from StorONE’s storage software, SSDs, and WD’s Ultrastar disk drives, and praises the unmatched stability and simplicity the system delivers, maximizing performance and longevity by moving data seamlessly between flash and HDD tiers. Dustin Brandt, Director of IT at America’s Test Kitchen, says: “The system runs flawlessly – we simply don’t need to worry about storage. StorONE’s platform is incredibly stable and simple to operate, and Western Digital’s Ultrastar HDDs have been rock-solid. This frees up our time to focus on what really matters to our business.”
…
DDN subsidiary Tintri announced GA of its VMstore T7000 hardware system for hypervisor, database and container ecosystems, bringing new features to enable enterprises to deploy virtualized environments, alongside the latest version of Tintri’s TxOS operating system, available for all Tintri VMstore solutions. The VMstore T7290 expands Tintri’s T7000 portfolio, featuring 40 percent better performance and support for 30 TB drives that double system capacity over past generations. The T7290 scales up to 2.58 PB in a single system and over 165 PB in scale-out configuration, enabling the management of hundreds of thousands of VMs, persistent container volumes and databases.
The T7290 also delivers enhanced security with Glas-DP, Tintri’s built-in Data Protection and Recovery Suite, and dual custody snapshot locking. It offers simultaneous multi-hypervisor and container support, integrating with vSphere, Hyper-V, Citrix Xen, Red Hat RHEV, plus expanded support for Platform9, OpenStack and Red Hat OpenShift.
TxOS 6.0 introduces enhanced observability, automation and self-service for application-centric infrastructure, allowing IT managers to simplify administration, reduce complexity and focus on business outcomes while managing all workloads from a single console. Tintri VMstore T7290 and the TxOS 6.0 update are available directly from Tintri and certified partners now.
…
The GA of the latest VAST AI OS version includes the VAST DataBase now delivering enhanced performance through its new Sorted Tables feature, making near-logarithmic query speeds across massive datasets possible. Logarithmic time means the query time increases by a constant amount each time the dataset size doubles. For example, if a database has 1,000 records and a query takes 10 milliseconds, doubling the dataset to 2,000 records might only increase the query time to ~11 milliseconds. Databases use indexes (e.g., B-tree or B+ tree indexes) to organize data in a way that allows for fast lookups, inserts, and deletes. Instead of scanning every record (linear time, O(n)), the database traverses a tree structure, reducing the number of operations to a logarithmic scale.
Suppose a database has 1 million records. Without an index, finding a specific record might require scanning all 1 million (O(n)). With a B-tree index, the database might only need to check ~20 nodes (log₂(1,000,000) ≈ 20) to locate the record, resulting in a much faster query. In real-world scenarios, VAST has seen point queries improve by up to 100x and Trino queries on 10B+ rows speed up by 95x – reducing latency from 12 seconds to just 144 ms.
…
VAST Data is partnering with Shonfeld Data Services (SDS), the dominant colocation and cloud provider in Israel, to power Israel’s sovereign AI infrastructure. SDS is building a next-generation AI cloud ecosystem to serve both domestic and international enterprises, and integrating VAST’s AI Operating System. There will be dozens of petabytes of VAST-powered data infrastructure and thousands of Nvidia Blackwell GPUs and Nvidia networking designed to serve large-scale enterprises that require uncompromising performance, security, and scalability to fuel their AI training and deployment workloads.
…
AI and HPC parallel file system supplier VDURA has joined the Ultra Ethernet Consortium (UEC). The UEC, which has welcomed more than 40 new members in the past year and a half, is building an open, Ethernet-based, high-performance architecture designed to address the challenges of accelerated computing. UEC brings together the full ecosystem of compute, networking, and storage to deliver advances in bandwidth, latency, and scalability needed for AI training, fine-tuning, and inference at scale.
