IOSG: From Hot Storage to Cold Memory – Decentralized Storage in the AI Era's Storage Boom

Author:0xjacobzhao, IOSG

Recently, CXMT, known as "China's first domestic storage stock," officially listed on the ChiNext board, igniting the market with a staggering 500% surge. Despite the storage sector still experiencing ripple effects from recent corrections, AI storage continues to be frantically revalued by capital within the current wave of tech narratives. Meanwhile, decentralized storage in the Web3 space has fallen into a prolonged period of silence and despondency. Why do both share the "storage" label yet exhibit such a stark contrast in market performance? The fundamental answer lies in a complete divergence in their underlying value functions.

The revaluation of storage in the AI era is essentially a frenzy centered on "hot data efficiency," serving the ultimate maximization of computational utilization and commercial monetization. In contrast, decentralized storage steadfastly upholds the value proposition of "cold data trustworthiness," defending data fairness, censorship resistance, and the long-term memory of human civilization. The former is an efficiency system for hot data, while the latter is a trust system for cold data. The current capital market undoubtedly firmly sides with "efficiency." However, human civilization ultimately still requires an immutable memory foundation. The long-term value of trustworthy cold storage has never disappeared; it merely lies dormant in the dark side of the cycle, awaiting revaluation by the era.

Why Storage Has Re-emerged as a Focal Point in the AI Industry Chain

In the traditional IT era, storage was a "capacity business." Enterprise CIOs focused on unit capacity cost, disk reliability, disaster recovery solutions, archiving strategies, and equipment refresh cycles spanning 3–5 years. Storage was seen as an accessory following server procurement.

This current storage boom is not a traditional cyclical recovery but a revaluation of data flow capabilities driven by AI. In the era of large models, the logic of storage has qualitatively shifted from "capacity-first" to "efficiency-first," relentlessly pursuing extreme metrics like GPU feeding rates, Checkpoint write speeds, and ultra-low latency for RAG. This marks the elevation of storage's value from a "final parking spot for data" to the "high-speed conduit for data entering computation."

The evolution of resource bottlenecks in AI infrastructure is essentially a battle to fill the "shortest plank" in the barrel. Real computational utilization is not a linear sum of individual assets but a strict multiplicative effect: Real Computational Utilization = GPU × HBM × DRAM × SSD × Network × File System. Any weak link can cause the overall computational utilization rate to collapse. In the AI era, storage has transformed for the first time from a "cost center" into an "efficiency engine." This is the fundamental logic behind storage's revaluation.

AI Storage Architecture Overview: From HBM Bandwidth Organs to Data Lake Foundation

AI storage is by no means a simple stacking of individual hardware components but a tightly coupled, hierarchically scheduled complex system. Within this system, industrial value and capital focus are highly concentrated on HBM, enterprise-grade SSDs, SSD controllers, NVMe/CXL protocols, and high-performance storage systems. To clearly dissect the flow of value, we divide the AI storage architecture into four core layers from top to bottom:

• Compute-Proximate Memory Layer (Bandwidth Core): Primarily dominated by HBM, supplemented by DRAM and CXL memory pooling technologies. This layer is directly attached to the GPU/CPU package or bus, aiming to break the "memory wall." It is the first critical juncture determining whether computational power can be fully unleashed.

• High-Speed Persistent Storage Layer (IO Hub): The core logic is Enterprise-grade SSD = NAND flash + SSD controller + NVMe/PCIe data path. This layer handles high-frequency Checkpoint writes, massive training dataset loads, and RAG hot data caching. It represents the most definitive persistent storage increment in AI data centers.

• Low-Cost High-Capacity Storage Layer (Capacity Foundation): Composed of HDDs, cold storage, and data lake archiving systems. Facing exponentially growing multimodal raw data, historical logs, and compliance backups, this layer still offers an irreplaceable TCO (Total Cost of Ownership) advantage.

• AI Storage Systems & Data Software (Scheduling Brain): Includes high-performance parallel file systems, distributed object storage, vector databases, and the RAG data governance layer. What AI truly consumes is not raw hardware but the availability of data efficiently organized, indexed, and permissioned by the software stack.

As an ecosystem extension, decentralized storage does not directly engage in the millisecond-level race for AI hot data. Instead, it anchors itself in the niches of public dataset notarization, AI training data provenance, and long-term cold memory archiving, establishing its unique position as the "trustworthy cold layer."

HBM: The "Bandwidth Organ" Closest to Compute in the AI Storage Chain

High Bandwidth Memory (HBM) is not traditional storage but a high-bandwidth memory layer proximate to the GPU. Its core mission is not to preserve data but to continuously "feed" data to the compute units with extremely high bandwidth. HBM is the closest component to compute power in the AI storage chain, with the highest certainty. It directly determines whether GPUs can be "fed adequately" and is currently the most critical supply chain bottleneck.

The core architecture of HBM is "3D DRAM stacking + 2.5D advanced packaging": Through TSV (Through-Silicon Via) vertical stacking and CoWoS (Chip-on-Wafer-on-Substrate) heterogeneous integration, the distance between memory and compute is compressed to the extreme, achieving a generational leap in bandwidth. Its industrial barriers extend beyond DRAM design, encompassing a systems engineering challenge involving DRAM process technology, TSV, ultra-thin stacking, packaging, thermal management, testing, and customer qualification. A flaw in any single step can render an entire HBM stack unusable.

Currently, only three global giants—SK hynix, Samsung, and Micron—can mass-produce HBM stably, constructing a triple moat of top-tier DRAM process, packaging capability, and certification by customers like NVIDIA/AMD.

DRAM & CXL:System Memory Foundation & Memory Pooling Engine

While HBM tackles GPU-proximate extreme bandwidth, DRAM solidifies the server's system memory foundation, and CXL aims to break physical boundaries, reconfiguring the organization of memory resources within data centers.

• DRAM: Primarily handles CPU-side caching, data preprocessing, intermediate state buffering, and system operation. It is the most fundamental system memory layer for servers. The global DRAM market is highly concentrated among the three giants: SK hynix, Samsung, and Micron; ChangXin Memory Technologies (CXMT) is a key variable for DRAM domestic substitution in China.

• CXL (Compute Express Link): A next-generation cache-coherent interconnect protocol for data centers. It aims to break the limitations of traditional DIMM slots, local memory capacity, and server memory silos, driving memory architecture toward expansion, pooling, and sharing. CXL is currently still in the early stages of moving from platform support to scaled deployment, with high long-term architectural value. Core companies include Astera Labs and Montage Technology.

Enterprise-grade SSD: The Data Hub Built on NAND, Controllers & NVMe

Enterprise-grade SSDs are the most crucial high-throughput persistent storage increment in AI data centers. With extremely high throughput, ultra-low latency, and stable QoS, they continuously "feed" data to GPUs, spanning the entire lifecycle from training data loading and Checkpoint writes to RAG retrieval, inference caching, and log streaming.

Within the AI storage architecture, an SSD is not an isolated hardware component but a highly coupled system, which can be summarized by the industrial formula: Enterprise-grade SSD = NAND Flash + SSD Controller + NVMe/PCIe Data Path. These three layers represent independent industrial chain segments:

• NAND Flash (Raw Material Layer): Determines storage density and unit cost. Representative companies: Samsung, SK hynix (Solidigm), Micron, Kioxia, Western Digital, YMTC (Yangtze Memory Technologies Co.).

• SSD Controller (Performance Enabler Layer): Determines performance unlocking, error correction, QoS stability, and wear leveling. Representative companies: Phison, Silicon Motion, Marvell, Maxio.

• NVMe/PCIe (Data Path Layer): Determines data transfer efficiency from storage to compute. Combined with technologies like GPUDirect Storage, it reduces CPU memory bounce buffers and CPU involvement, significantly alleviating I/O bottlenecks. Representative companies: Broadcom, Marvell, Astera Labs.

HDD / Cold Storage / Archive: The Low-Cost Foundation for AI Data Lakes

AI will not eliminate HDDs. As multimodal large models crave video and image data, and enterprise compliance logs/historical datasets expand exponentially, the demand for low-cost cold data storage is also surging simultaneously. In the AI storage architecture, SSDs and HDDs collaborate based on a tiered division of business value: SSDs handle hot data and high throughput; HDDs handle low cost and long-term preservation. Representative companies include Seagate, Western Digital, Toshiba.

AI Storage Software Stack: The Scheduling Hub for Data Availability

What AI truly consumes has never been bare drives, but "data services" meticulously organized by software stacks. This architecture transforms underlying hardware into knowledge assets directly usable by upper-layer AI, specifically divided into four layers:

• High-Performance Storage Systems (Feeding System): Core focus is on concurrent throughput and low latency. Parallel file systems address the "data starvation" problem in GPU clusters, ensuring rapid flow for training and inference. Representative companies: VAST Data, WEKA, Pure Storage.

• Object Storage (Raw Data Lake): Core revolves around Object, Key, and Metadata management, handling massive unstructured data. It does not pursue ultimate low latency but builds a capacity foundation with low cost and cloud-native characteristics. Representative company: AWS S3.

• Vector Database (Semantic Indexing Layer): Vector databases are responsible for storing, indexing, and retrieving vectors generated by embedding models, enabling AI to precisely locate relevant content from vast knowledge bases. Representative companies: Pinecone, Milvus.

• RAG Data Layer (Knowledge Invocation Layer): Goes beyond mere retrieval, encompassing data chunking, cleaning, permission control, and citation provenance, ensuring enterprise data can be called by large models safely, accurately, and traceably. Representative company: Databricks.

From AI Hot Storage to Decentralized Cold Memory: Efficiency Maximization vs. Trust Maximization

AI storage is an extreme efficiency-driven system. Its value function focuses on maximizing computational output. HBM bandwidth determines whether GPUs can be fed adequately; SSD throughput determines the efficiency of dataset and Checkpoint read/writes; low latency is crucial for real-time RAG and inference experiences. These metrics ultimately converge into GPU utilization and cost per Token, directly determining the commercial profitability of AI applications. The ultimate goal of AI storage is not preservation but acceleration, serving productivity.

The value function of decentralized storage is entirely different. It questions whether data will still exist in a decade, whether it has been tampered with, and if it can resist centralized censorship. Through cryptographic proofs and distributed networks, it constructs an open-access, permanently preserved public data foundation. Its ultimate goal is to defend the absolute authenticity and sovereign independence of data, serving fairness, censorship resistance, and civilizational memory.

AI storage is the "hot storage" fueling future productivity; decentralized storage is the "cold memory" preserving undeletable historical records for human civilization. The former serves efficiency, pursuing extreme speed; the latter serves trustworthiness, defending silent memory. The former determines how fast models run; the latter determines whether memories will be deleted. Currently, market mechanisms reward efficiency in productivity, placing AI storage at the forefront while decentralized storage seems to be experiencing a silent phase of valuation collapse and narrative drainage.

The Vision and Reality of Decentralized Storage

There are numerous decentralized storage projects, but in terms of industry mindshare and ecosystem depth, the core representatives remain Filecoin and Arweave. Although both fall under "decentralized storage," their underlying architectural philosophies are almost two completely different paths—the former uses market-based contracts to approximate the elasticity of AWS, while the latter uses a one-time social contract to approximate the permanence of a library.

• Filecoin: Has constructed the most complete verifiable economic system through PoRep and PoSt. It should no longer compete head-on with AWS in the consumer cloud storage market. Instead, it should pivot towards AI data provenance, public dataset hosting, and compliance archiving, providing a verifiable chain for model auditing and copyright proof. Its necessary path is to encapsulate itself into an S3-compatible API supporting fiat payments, upgrading from a "cheap storage market" to "verifiable computation infrastructure."

• Arweave: With the narrative of "pay once, store forever," its Blockweave and SPoRA mechanisms incentivize miners to preserve and provide quick access to as much data as possible, especially scarce historical data. Its optimal position is as the foundation for humanity's public memory—preserving human rights records, war crime evidence, cultural canons, archiving legal and financial history, and providing permanent, accessible long-term memory for AI Agents. Arweave's value lies not in speed but in its ability to carry civilizational memory across cycles.

The dilemma for decentralized storage projects like Filecoin and Arweave does not lie in an erroneous value proposition, but in a long-term misalignment between productization, retrieval experience, real demand, and token incentives. This reveals the vast chasm between geek ideals and mainstream commercial adoption:

• Supply-Demand Incentive Misalignment: Early networks like Filecoin rapidly expanded capacity via token rewards but failed to build a sufficiently strong paying demand side, leading to vast capacity but low utilization and poor payment conversion. The system rewarded "I can store" rather than "need me to store."

• Lack of Enterprise-Grade Service Capabilities: AWS's moat is not hard drives, but a "data operating system" comprised of APIs, SLAs, permission management, compliance auditing, and technical support. Enterprises buy "peace of mind," not experimental infrastructure requiring them to manage keys and node selection.

• Retrieval Experience Shortcomings: "Storing it" does not equal "retrieving it stably and with low latency." Dispersed nodes, complex topology, and a lack of unified SLAs make it difficult to handle AI hot data workflows; better suited for trusted cold archiving and data provenance.

• Privacy & Compliance Inadequacies: Enterprise private data cannot simply be written to a public, permanent network; the right to erasure is inherently at odds with permanent immutability. Decentralized storage is better suited for public data and long-term archives, not indiscriminately hosting core private data.

• Token Economics Amplifying Cycles: Financialization during bull markets masked insufficient demand; declining miner ROI during bear markets exposed commercial viability shortcomings. Tokens can cold-start supply but cannot automatically create demand and sustainable revenue.

Other decentralized storage projects often focus on specific ecosystems or niche segments: Storj/Sia have weaker cross-cycle industry mindshare and Web3 narrative influence compared to Filecoin/Arweave; BNB Greenfield/Walrus are tied to specific public chain ecosystems like BNB or SUI; Celestia/EigenDA belong to the Data Availability (DA) layer, serving Rollup transaction confirmation rather than long-term archiving; Projects like 0G, with mixed AI/DA narratives, attempt to integrate storage, data availability, computation, and AI agent settlement into an AI-native modular infrastructure suite, but their real demand, developer adoption, and commercialization loops remain to be validated.

Future Opportunities for Decentralized Storage: The Long-Term Pendulum Between Efficiency and Trust

During periods of technological红利 (bonanza) explosion, capital chases efficiency frantically, assigning extremely high premiums to assets like GPUs and HBM. Decentralized storage, which advocates for "trust and fairness," is naturally marginalized. However, history's pendulum will not rest forever on the side of efficiency. Events such as unreasonable platform bans and content takedowns, AI copyright lawsuits forcing data provenance proof, data sovereignty disputes triggered by geopolitical conflicts, the disappearance of public archives due to data monopolies, and regulatory pressure for auditing model training data compliance could all potentially catalyze a revaluation of "trustworthy storage." The future opportunities for decentralized storage may still manifest unique value in the following directions:

• AI Data Provenance: Combining cryptographic proofs to construct "data lineage proof," addressing regulatory and audit pressures.

• Public Datasets & Civilizational Archives: Anchoring censored archives and cultural heritage, building an irreplaceable, undeletable memory layer.

• Trusted Archiving & Compliance Notarization: Implementing trusted self-attestation through hash notarization, providing high-grade digital notarization services.

• Integration with ZK/TEE/DID Technologies: Mitigating privacy tensions, upgrading from a mere "storage protocol" to "trusted data infrastructure."

• Invisible Product Strategy: Providing S3-compatible APIs and fiat billing, allowing users to directly purchase "trusted archiving" services.

AI storage and decentralized storage—one pursues ultimate efficiency, providing the fuel for our rush into the future; the other defends silent memory, safeguarding our right to look back at the past. Currently, the market unreservedly rewards efficiency, making decentralized storage appear silent and even collapsed. However, as the AI era further amplifies issues like data monopolies, copyright disputes, and the fragility of historical memory, decentralized storage may experience a value revaluation as a "trustworthy cold layer." Those memories that cannot be easily erased by platforms, corporations, or any single power might transform from idealistic romance, from fringe belief, into essential infrastructure.