Segment Category

Sub-Segments

Key Insights

By Type

• NAND Flash-Integrated Compute Engines

• Near-Storage Processing Units (NSPUs)

• Computational Storage Devices (CSDs)

• SSD Controller-Embedded Processors

Computational Storage Devices (CSDs) represent the leading segment within the By Type classification, driven by their versatile ability to natively execute a broad spectrum of workloads directly within the drive enclosure.

• CSDs are uniquely architected to support complex tasks such as database query processing, real‑time data analytics, AI inference, compression, and encryption without offloading these operations to the host CPU, thereby dramatically alleviating the memory wall bottleneck that has long constrained data‑intensive computing environments.

• The formalization of the Computational Storage Architecture standard by the Storage Networking Industry Association (SNIA) has significantly bolstered CSD adoption by establishing interoperability benchmarks, giving enterprise buyers and cloud operators greater confidence in deploying these solutions at scale across heterogeneous infrastructure environments.

• Leading industry players such as Samsung Electronics and ScaleFlux have actively commercialized CSD‑based solutions, further validating the segment's maturity and accelerating broader market acceptance across hyperscale data centers and enterprise storage ecosystems.

By Application

• Data Analytics and Business Intelligence

• AI and Machine Learning Inference

• Database Query Processing

• Data Compression and Encryption

• Others

AI and Machine Learning Inference has emerged as the dominant application segment, propelled by the exponential proliferation of AI workloads across cloud, edge, and enterprise computing environments.

• As artificial intelligence models grow increasingly complex and data‑hungry, the conventional approach of shuttling massive datasets between storage and host CPUs introduces prohibitive latency and energy overhead; NMC In‑SSD processors directly address this challenge by enabling inference computations to be performed at the storage layer, where data already resides, resulting in dramatically faster and more energy‑efficient AI pipelines.

• The convergence of large‑scale language model deployment, real‑time recommendation engines, and computer‑vision applications has created a persistent and growing demand for storage‑native compute capabilities, making AI inference a natural and strategic fit for NMC In‑SSD architectures.

• Enterprise and hyperscale cloud operators are increasingly prioritizing in‑storage AI acceleration as a means to decouple inference throughput from CPU availability, enabling more scalable and cost‑effective AI infrastructure deployments without proportional increases in server hardware investment.

By End User

• Cloud Service Providers and Hyperscalers

• Enterprise Data Centers

• Telecommunications and Network Operators

• Research and Academic Institutions

Cloud Service Providers and Hyperscalers constitute the leading end‑user segment, as these organizations operate at the frontier of data‑intensive workload management and are uniquely positioned to realize the full transformative benefits of NMC In‑SSD processor technology.

• Hyperscale cloud operators routinely manage petabyte‑scale datasets distributed across vast storage fleets, and the persistent memory‑wall problem - characterized by an ever‑widening gap between CPU processing speeds and storage I/O bandwidth - creates a compelling operational imperative to adopt in‑SSD compute solutions that minimize unnecessary data movement across storage hierarchies.

• The strategic alignment of NMC In‑SSD processors with cloud‑native workloads such as distributed database management, large‑scale log analytics, and real‑time stream processing makes these processors an ideal fit for hyperscale deployment architectures that prioritize throughput, energy efficiency, and total cost of ownership optimization.

• Cloud providers also benefit from the SNIA Computational Storage Architecture standard, which facilitates seamless integration of NMC‑enabled drives into existing infrastructure management frameworks, reducing deployment complexity and accelerating time‑to‑value for in‑storage compute investments.

By Interface Standard

• NVMe‑based NMC Processors

• PCIe‑attached Computational Storage

• SATA‑based Near‑Storage Compute

NVMe‑based NMC Processors lead this segment by virtue of the NVMe protocol's architectural superiority in supporting low‑latency, high‑parallelism communication between host systems and storage‑embedded compute engines.

• The NVMe interface is inherently designed to exploit the parallelism of NAND flash‑based storage, and its command‑set extensions provide a natural foundation for offloading computational tasks to in‑SSD processors, enabling seamless integration into modern data‑center storage fabrics without requiring disruptive changes to existing host software stacks.

• NVMe's widespread adoption across enterprise and hyperscale environments has created a large and growing installed base that is primed for incremental NMC capability upgrades, lowering the barrier to adoption and enabling storage vendors such as Kioxia, Western Digital, and Micron Technology to deliver NMC‑enhanced NVMe drives with minimal ecosystem friction.

• The ongoing evolution of NVMe standards, including emerging command‑set specifications tailored for computational storage, further cements NVMe‑based NMC processors as the preferred interface paradigm for next‑generation in‑SSD compute deployments across demanding enterprise and cloud use cases.

By Deployment Mode

• On‑Premises Data Center Deployment

• Cloud‑Native Deployment

• Edge and Distributed Deployment

Cloud‑Native Deployment stands as the dominant deployment mode, reflecting the broader industry shift toward cloud‑first infrastructure strategies and the concentration of data‑intensive workloads within hyperscale cloud environments.

• Cloud‑native deployment of NMC In‑SSD processors enables organizations to leverage computational storage capabilities as a scalable, on‑demand resource that can be provisioned and managed dynamically alongside other cloud infrastructure components, offering a level of operational flexibility that is difficult to replicate in traditional on‑premises environments.

• The cloud deployment model also facilitates rapid iteration and validation of in‑SSD compute use cases - including AI inference pipelines, real‑time analytics, and intelligent data tiering - without requiring upfront capital investment in specialized on‑premises hardware, making it particularly attractive for organizations in the early stages of NMC technology adoption.

• Edge and distributed deployment is emerging as a fast‑growing complementary mode, driven by the proliferation of latency‑sensitive applications in telecommunications, autonomous systems, and industrial IoT, where the ability to perform analytics and inference natively within edge‑deployed SSDs offers transformative advantages in bandwidth conservation and response‑time reduction.