From Hot Storage to Cold Memory: Decentralized Storage Under the AI-Era Storage Boom
从热存储到冷记忆：AI时代存储热潮下的去中心化存储
Summary
Opening with CXMT's ChiNext debut, this report dissects how the AI-era storage repricing is fundamentally about hot-data efficiency, then contrasts decentralized storage's trust-maximizing cold-memory value function, mapping Filecoin and Arweave's opportunities and hard realities.
Why it matters
Storage is a critical repriced node in the AI value chain, yet the market misreads decentralized storage's long silence — the two sides maximize fundamentally different value functions, and understanding this divergence is the prerequisite for judging the Web3 storage track.
Key ideas
The AI storage repricing is about hot-data efficiency: HBM, enterprise SSDs and the storage software stack form a tightly coupled layered system.
Decentralized storage is cold memory that maximizes trust, following a different value function from efficiency-maximizing hot storage.
Filecoin and Arweave's struggles stem from supply-demand incentive mismatch and missing enterprise capabilities; future opportunities lie in data provenance, public archives, and trusted notarization.
Versions
Chinese and English versions of this report.
Published or syndicated by
Related research
2025.12 · Intelligence Infra