Storage & Data Management Watch: Samsung’s AI Storage Strategy, Dell’s Knowledge Graph and the HDD Supply Crunch
Storage & Data Management Watch · October 14, 2026
Samsung Unveils AI Storage Strategy at OCP, Dell Expands Its AI Data Platform With a Knowledge Graph, Toshiba's HDD Expansion Rattles the Storage Market, and VAST Data Lands Strategic Investment From Hyundai and Kia
The Storage Layer Is Becoming the Bottleneck — and the Battlefield — of Enterprise AI
For the past two years, the AI infrastructure conversation has centered on GPUs. That is now changing. As AI models grow larger and inference workloads multiply, memory capacity and data storage are becoming the constraints that determine whether AI systems perform at their potential. Samsung Electronics made that explicit at the OCP Global Summit 2026, unveiling a three-part strategy covering high-performance memory, CXL-based memory expansion, and data storage and management solutions using next-generation SSDs. Shim Ho-jun, head of Samsung's application engineering team, explained that as AI models grow and key-value cache requirements increase, memory capacity expansion and storage technology are becoming increasingly important alongside memory bandwidth.
Dell Technologies is approaching the same problem from the software side. On October 13, Dell expanded its AI Data Platform with a Unified Semantic Layer, an Enterprise Knowledge Graph, and Knowledge Agents designed to help AI systems interpret business terms, identify relationships between information, and follow access controls. The company also announced storage security and multitenancy improvements for PowerScale, including support for up to 500 tenants in a single cluster. Dell's message is that the bottleneck is no longer compute — it is helping AI agents understand the data they need to work with.
In hardware, the supply-demand picture is tightening. Toshiba plans to invest approximately ¥60 billion (around $380 million) to double HDD production capacity by fiscal 2027, its first major investment in the business in years. The announcement sent Western Digital and Seagate shares down sharply — WDC fell as much as 14% and STX over 16% intraday — even though both companies have already sold out their 2026 HDD capacity, with only 5% allocated to the consumer market. The market reaction reflects concern that Toshiba's expansion could disrupt a supply-constrained environment that has been highly favourable for incumbent pricing. And VAST Data received a strategic investment from Hyundai Motor and Kia on October 13, aimed at strengthening physical AI data infrastructure for autonomous driving and robotics, with the company having been valued at $30 billion in its April Series F round.
Storage & Data Management at a Glance
| Development | Company | Category | Why It Matters |
|---|---|---|---|
| Next-gen memory and storage strategy | Samsung | AI memory & SSD | Memory capacity, not bandwidth, is becoming the AI bottleneck; CXL and KV cache optimization are the responses |
| AI Data Platform expansion | Dell | Data management & governance | Semantic layer and knowledge graph help AI agents understand business context; features arrive in 2027 |
| HDD capacity expansion | Toshiba | Storage hardware | Supply-constrained HDD market may loosen; Toshiba targets 30% share and 65TB drives by 2030 |
| Strategic investment | VAST Data | AI data infrastructure | Hyundai/Kia invest to strengthen physical AI data infrastructure for autonomous driving |
| OCI NetApp Storage Service | NetApp & Oracle | Cloud storage | Fully managed ONTAP service natively integrated into OCI; bridges on-premises and cloud data management |
| Wasabi funding round | Wasabi | Cloud storage | $70M raised at $1.8B valuation to expand AI-optimized storage and NVMe offerings |
| Unified data management platform | Precisely | Data governance | Platform makes enterprise data AI-ready across cloud, on-prem, and mainframe without migration |
Signal summary: Every development this week points to the same shift: the storage and data layer is becoming the constraint on AI performance and the control point for AI governance. Samsung is addressing the hardware side (memory capacity, SSD optimization). Dell is addressing the software side (semantic understanding, data access). Toshiba is expanding supply to meet demand. And data management vendors are racing to make enterprise data AI-ready.
The Stories Shaping Storage & Data Management
Samsung Unveils Three-Part AI Storage Strategy as Memory Capacity Becomes the New Bottleneck
What is happening: Samsung Electronics presented its next-generation memory and storage technology strategy at the OCP Global Summit 2026 in San Jose, outlining three key directions: high-performance memory, CXL-based memory expansion technologies, and data storage and management solutions using next-generation SSDs. Shim Ho-jun, corporate vice president and head of Samsung's application engineering team, explained that as AI models grow larger and the volume of data and key-value cache required for inference increases, memory capacity expansion and data storage are becoming increasingly important alongside memory bandwidth. Samsung also plans to jointly demonstrate a CXL-based memory expansion solution with AMD at the event.
Why it matters: The framing is significant. Samsung is saying, in effect, that bandwidth alone is no longer sufficient — the industry needs capacity. AI inference workloads require large KV caches, and those caches need somewhere to live. If memory cannot hold them, they spill to storage, which introduces latency. That is why CXL-based memory expansion matters: it allows systems to add capacity without adding a new CPU socket, effectively creating a tier between DRAM and SSD.
The competitive context: Samsung's announcement comes as HBM remains sold out and DRAM capacity is being reallocated toward AI memory, as tracked in Semiconductor Watch. The company's emphasis on SSDs for AI storage also reflects the growing role of enterprise SSDs as a KV cache tier — a trend that competes with HDDs for some workloads but complements them for others.
Editorial question: Does CXL-based memory expansion become a standard architectural component in AI servers, or does it remain a specialised solution for capacity-bound workloads? Connect to: Semiconductor Watch · AI Infrastructure Special Report · Hardware Watch
Dell Expands AI Data Platform With Semantic Layer and Knowledge Graph to Help Agents Understand Business Data
What is happening: Dell Technologies announced an expansion of its AI Data Platform on October 13, adding a Unified Semantic Layer, an Enterprise Knowledge Graph and Knowledge Agents. The Unified Semantic Layer helps AI agents consistently understand business terminology across systems — for example, recognising that one database's "client" and another's "account" may refer to the same concept. The Enterprise Knowledge Graph maps relationships between business data, and Knowledge Agents use this context to provide guidance on specific topics with business-defined rules for data access, quality requirements, and spending limits. Dell also announced improvements to PowerScale storage security and multitenancy, including support for up to 500 tenants in a single cluster, encrypted file traffic, and more granular access controls, scheduled for November 2026.
Why it matters: The announcement addresses a practical challenge that is limiting enterprise AI deployment: AI agents do not understand business context. An agent that can generate text is not the same as an agent that knows what a "client" is in your CRM, how it relates to an "account" in your ERP, and who is allowed to see what. Dell's response is to build the semantic and relational context layer that agents need — not as a separate product, but as part of the data platform the enterprise already uses.
The availability caveat: The main context-building features — Semantic Layer, Knowledge Graph, and Knowledge Agents — are not available until the first half of 2027. Businesses evaluating Dell's platform will need to weigh those timelines against their own AI deployment plans.
Editorial question: Does the semantic and knowledge-graph layer become a standard component of every enterprise data platform, or does it remain a differentiator for vendors that build it early? Connect to: Enterprise Software Watch · AI Watch · Data Analytics Intelligence
Toshiba to Double HDD Production Capacity by 2027, Rattling a Supply-Constrained Market
What is happening: Toshiba plans to invest approximately ¥60 billion (around $380 million) to expand its HDD production facility in the Philippines and double capacity by fiscal 2027 — its first major investment in the HDD business in years. The announcement sent storage stocks sharply lower: Western Digital fell as much as 14% and Seagate over 16% intraday, though both stocks later recovered somewhat.
The supply context: Western Digital's 2026 HDD capacity is already sold out, with only 5% allocated to the consumer market. Seagate's nearline capacity is also fully booked, with some customer demand planning extending to 2029 and beyond. SSDs, while faster, cost approximately 20 times more per unit than HDDs, driving data center customers toward HDDs for capacity-oriented workloads.
Why it matters: The HDD industry is highly concentrated — only Seagate, Western Digital and Toshiba are major suppliers — and any capacity change by one player affects the entire market balance. Toshiba's expansion is a bet that AI-created data volumes will not disappear and that HDDs remain irreplaceable for exabyte-scale storage. The company also plans to introduce new products at the Philippines facility that increase per-drive capacity by up to 40%, and targets 65TB drives by 2030 and 100TB-class products eventually. IDC projects global data generation will reach 718 zettabytes by 2030, roughly four times 2024 levels, with about 60% expected to be stored on HDDs.
Editorial question: Does Toshiba's capacity expansion ease the HDD shortage and moderate pricing, or does AI-driven demand absorb the additional supply without changing the supply-demand balance? Connect to: Hardware Watch · AI Infrastructure Special Report · Semiconductor Watch
Hyundai and Kia Invest in VAST Data to Strengthen Physical AI Data Infrastructure
What is happening: VAST Data announced on October 13 that it received a strategic investment from Hyundai Motor and Kia, aimed at strengthening AI data infrastructure for autonomous driving and robotics. The investment size was not disclosed. VAST Data was valued at $30 billion in its April 2026 Series F round, which included participation from Fidelity, NEA, and Nvidia.
Why it matters: This is a notable example of a strategic investor — not a financial one — putting capital into a data infrastructure company for a specific operational purpose. Hyundai and Kia are not investing for financial return; they are investing to ensure that the data infrastructure they need for autonomous driving and robotics development exists and is optimized for their workloads.
The physical AI dimension: Autonomous driving and robotics generate enormous volumes of sensor data — lidar, camera, radar, telemetry — that must be ingested, stored, processed, and used for model training. This is a different data profile from enterprise AI: it is high-velocity, high-volume, and often needs to be processed at the edge before being consolidated for training. VAST Data's AI operating system is positioned as the data foundation for exactly this kind of workload.
The strategic significance: As physical AI moves from research to production, the companies building autonomous systems increasingly need to control their data infrastructure rather than depend on general-purpose cloud services. Hyundai and Kia's investment is a bet that owning that layer — or at least having a strategic relationship with the provider — is a competitive advantage.
Editorial question: Does physical AI data infrastructure become a distinct market category, separate from enterprise AI data platforms, with its own vendors and requirements? Connect to: AI Watch · Startup Funding Watch · AI Infrastructure Special Report
NetApp and Oracle Launch Fully Managed Cloud Storage Service for AI and Enterprise Workloads
What is happening: NetApp and Oracle announced a new fully managed storage service that natively integrates NetApp's enterprise storage capabilities into Oracle Cloud Infrastructure (OCI). The service — OCI NetApp Storage Service — aims to simplify cloud migration and management for critical AI and enterprise workloads, allowing customers to use familiar ONTAP data management functions for databases, enterprise applications, virtualisation, EDA/HPC, regulated applications, and AI data pipelines. Customers can manage the service through the OCI console, SDKs, ONTAP APIs, and existing operational workflows. The service is expected to be generally available within the next 12 months.
Why it matters: This is a significant bridging play. Enterprises that have invested in NetApp's ONTAP ecosystem — and the operational expertise that comes with it — have historically faced friction when migrating workloads to the cloud, because cloud storage services use different management paradigms. The OCI NetApp Storage Service removes that friction by making ONTAP a native service inside OCI. For regulated industries in particular, the ability to maintain consistent data management, high availability, security, and predictable performance across on-premises and cloud environments is a material procurement consideration.
The strategic logic: Oracle gains a differentiated storage offering for enterprises with existing NetApp investments. NetApp gains a cloud distribution channel that does not require customers to abandon their existing operational practices. Both benefit from the migration of AI and enterprise workloads to OCI.
Editorial question: Does the fully managed, familiar-API model accelerate cloud migration for enterprises with deep on-premises storage investments, or does it primarily serve as a retention mechanism for existing NetApp customers? Connect to: Enterprise Software Watch · AI Infrastructure Special Report · Cloud & Infrastructure Intelligence
Strategic Analysis: Where the Storage Layer Is Heading
This week's developments reveal a storage and data management market being reshaped by three simultaneous pressures: AI workloads that strain capacity, data governance requirements that demand context, and supply constraints that determine cost. Vendors are responding at every layer of the stack.
| Layer | What Is Changing | This Week's Example |
|---|---|---|
| Memory & DRAM | Capacity, not bandwidth, is the constraint for AI inference | Samsung CXL memory expansion with AMD |
| Enterprise SSD | KV cache optimization and AI inference storage | Samsung next-gen SSD data storage and management |
| HDD | Capacity expansion to meet AI data growth | Toshiba ¥60B investment; 65TB drives by 2030 |
| Data platform | Semantic context and knowledge graphs for AI agents | Dell Unified Semantic Layer and Knowledge Graph |
| Cloud storage | Native integration of enterprise storage into cloud platforms | OCI NetApp Storage Service |
| Physical AI data | Autonomous driving and robotics data infrastructure | Hyundai/Kia invest in VAST Data |
The common thread. Every layer is being reshaped by the same underlying force: AI workloads have different data requirements than the applications that enterprise storage was designed for. AI needs capacity for KV caches, context for agents to understand data, governance to control what agents can access, and supply that can scale with data growth. Vendors that address those requirements — whether at the memory, SSD, HDD, data platform, or cloud storage layer — are positioning for the next phase of enterprise data infrastructure spending.
What this means for buyers. The storage decisions enterprises make today increasingly determine their AI capabilities tomorrow. Memory capacity limits model size and inference throughput. Data platform context determines whether agents can act on enterprise data. HDD supply affects the cost of storing training data and checkpoints. Cloud storage integration shapes migration flexibility. These are not independent procurement decisions — they are interlocking choices that determine the cost and capability of AI systems.
Also in Storage & Data Management
Additional developments worth tracking:
Wasabi raises $70 million at $1.8 billion valuation: Cloud storage company Wasabi has raised $70 million in new funding, valuing the company at $1.8 billion. Wasabi said it will use the funds to respond to growing demand for storage optimized for AI workloads, expanding beyond hard disk-based storage into high-performance NVMe-based storage for AI and machine learning.
Precisely launches unified data management platform: Precisely introduced the Precisely Platform, a unified data management platform built on a shared metadata and semantic foundation, designed to make enterprise data AI-ready. The platform spans data integration, data quality, data governance, location intelligence, master data management, and customer communication — addressing data across cloud environments, data lakehouses, on-premises systems, and established platforms like mainframe and SAP.
Cloudflare Basin reaches general availability: Cloudflare's data platform, now named Basin, is generally available, offering ingestion, Iceberg table management, and SQL analytics on R2 storage. The platform is designed to make it easier and more cost-effective for developers and enterprises to build applications that leverage their data.
Nasuni delivers an operational file data platform for AI: Nasuni announced general availability of AI Activate and Resilio 6.0, with Enterprise Search rolling out later this quarter. Combined with its recent acquisition of DryvIQ, these capabilities bring content intelligence, automated governance, high-performance file access and governed AI activation together in the Nasuni File Data Platform — enabling people and AI to work from the same current data without a separate copy for AI.
Everpure enhances data intelligence: Everpure (formerly Pure Storage) rolled out new data management capabilities, including a lightweight Data Intelligence File Edition, native MCP integration, and enhancements to its Universal Data Plane and Intelligent Control Plane. The new capabilities are designed to help customers improve their data estates and increase the odds that their AI projects succeed.
Oxide Computer raises $445M: Oxide Computer Company raised $445 million in Series D funding to expand manufacturing of integrated cloud systems for enterprises operating infrastructure within their own facilities. The round was led by Eclipse with participation from AMD Ventures.
What to Watch Next
- Samsung's CXL demonstration with AMD. Watch for technical details from the OCP demonstration and whether CXL memory expansion gains traction in AI server designs.
- Dell's Knowledge Graph availability timeline. Monitor whether Dell ships the Semantic Layer, Knowledge Graph and Knowledge Agents on schedule in H1 2027, and whether competitors respond with similar capabilities.
- Toshiba's HDD capacity ramp. Track whether Toshiba achieves its goal of doubling capacity by fiscal 2027, and whether that shifts pricing dynamics in the HDD market.
- VAST Data's physical AI deployments. Watch whether Hyundai and Kia's investment leads to specific autonomous driving and robotics deployments using VAST's platform.
- OCI NetApp Storage Service GA. Monitor the general availability timeline and early customer adoption, particularly in regulated industries.
- Wasabi's NVMe expansion. Track whether Wasabi's move into high-performance AI storage gains traction against incumbents like Pure Storage and NetApp.
- Data platform consolidation. Watch whether the semantic layer and knowledge graph capabilities Dell is building become standard features across enterprise data platforms, or remain differentiators.
The storage and data layer is becoming the bottleneck — and the battlefield — of enterprise AI. Memory capacity limits inference throughput. Data context determines whether agents can act. HDD supply affects the cost of training data. And the vendors that address those constraints are positioning for the next phase of enterprise infrastructure spending.
For storage hardware vendors: Samsung's OCP presentation and Toshiba's HDD expansion represent two sides of the same coin. Samsung is addressing the memory and SSD layer with CXL and KV cache optimization. Toshiba is addressing the HDD layer with capacity expansion. Both are responding to the same demand signal: AI workloads need more storage, more capacity, and more intelligence built into the storage layer itself.
For data platform vendors: Dell's semantic layer and knowledge graph are a recognition that the bottleneck for enterprise AI is not compute — it is context. AI agents that do not understand business terminology, data relationships, or access controls cannot act on enterprise data safely. Vendors that build that context layer into their platforms gain a structural advantage as agentic AI adoption grows.
For enterprise buyers: Storage decisions are increasingly AI decisions. Memory capacity determines what models can run. Data platform capabilities determine what agents can do. HDD supply shapes the cost of storing training data. These are interlocking choices that should be evaluated together, not as separate procurement categories.
For the market: The storage and data management market is becoming more strategically important as AI adoption grows. The companies that control the memory, SSD, HDD, and data platform layers are increasingly in positions of leverage over the AI systems that depend on them. That is a shift in the balance of power within the AI stack — away from applications and toward the infrastructure that makes them possible.
The CODEW Research Flywheel
Storage & Data Management Watch tracks the hardware, software, and platforms that enterprises use to store, manage, govern, and activate data — and how AI is reshaping those layers. It sits alongside the other Watch series and feeds into the broader CODEW research architecture:
| Product | Core Question |
|---|---|
| Storage & Data Management Watch | Enterprise storage, data infrastructure, memory, HDD/SSD, data governance and AI data platforms |
| Semiconductor Watch | Chip design, memory, packaging, and the semiconductor supply chain |
| Hardware Watch | PCs, GPUs, storage devices and the computing hardware built on chips |
| AI Watch | AI models, agents, platforms and the competitive AI landscape |
| Enterprise Software Watch | Enterprise applications, SaaS, workflow platforms and vendor strategy |
| AI Infrastructure Special Report | Data centers, compute, power, cooling and the physical infrastructure of AI |
Storage & Data Management Watch → Memory & DRAM → Enterprise SSD → HDD → Data platforms → Data governance → Cloud storage → AI data infrastructure → Semiconductor Watch / Hardware Watch / AI Watch / Enterprise Software Watch
The CODEW Stat
718 zettabytes — the projected volume of global data generation by 2030, according to IDC, roughly four times the 2024 level. Approximately 60% of that is expected to be stored on HDDs. It is a number that explains Toshiba's ¥60 billion expansion, Samsung's emphasis on storage alongside memory, and the entire competitive dynamic in the HDD market. AI does not just consume compute — it generates data at a scale that the storage industry has never previously had to accommodate. The companies that can store, manage and activate that data at reasonable cost will hold positions of increasing strategic importance as the AI economy grows.
Sources: Reporting sourced from Yonhap News Agency, Times of India, CNA (Central News Agency Taiwan), TMCnet, HostingJournalist, DigitalToday, 163.com, HPCwire, Databricks, Börse München, The Hindu Businessline, and The CODEW Storage Pulse, covering storage and data management developments from October 1–14, 2026. Company product announcements are attributed to official channels; independent reporting is used for market context and financial data. All factual claims regarding product releases, partnerships, financial results, capacity expansions, and corporate strategy are drawn from company disclosures and contemporaneous reporting. Editorial analysis is clearly distinguished from reported facts throughout. Product availability, pricing, and financial figures should be confirmed against primary sources before publication. Forward-looking capacity and revenue projections are company guidance and are labeled as such. This series does not provide investment advice.
Reviewed by Erwin Castro
on
Wednesday, October 14, 2026
Rating:
No comments: