What Is IPFS? Definition, Architecture, and Web3 Applications
Definition
The InterPlanetary File System (IPFS) is a peer-to-peer distributed protocol for storing and accessing files, websites, applications, and data. Developed by Protocol Labs and first released in 2015, IPFS replaces the traditional internet’s location-based addressing — where content is identified by the server hosting it — with content-based addressing, where content is identified by a cryptographic hash of its contents.
In practical terms, this means that when data is stored on IPFS, it receives a unique identifier (a Content Identifier, or CID) derived from the data itself. Anyone holding a copy of the data can serve it to requesters, and any requester can verify that received data is authentic by recomputing the hash. This architecture eliminates dependence on specific servers, enables distributed storage and delivery, and provides inherent data integrity verification.
How IPFS Works
Content Addressing
Traditional web addressing uses URLs — https://example.com/file.pdf — that specify a location. If that server goes offline, the URL breaks regardless of whether the file exists elsewhere. IPFS addresses data by what it contains, not where it resides. The CID QmT5NvUtoM5nWFfrQdVrFtvGfKFmG7AHE8P34isapyhCxX identifies a specific piece of data. Any IPFS node holding that data can respond to requests for that CID.
Content addressing provides several properties automatically. Deduplication — identical data stored by different users produces identical CIDs, so the network stores only one copy regardless of how many users add it. Integrity verification — any modification to the data changes the CID, so recipients can confirm that received data matches the requested CID by recomputing the hash. Immutability — a CID always refers to the same data; content at a given CID cannot be changed.
Peer-to-Peer Distribution
IPFS operates as a peer-to-peer network where participating nodes both request and serve content. When a node retrieves content by CID, it becomes a host for that content and can serve it to subsequent requesters. Popular content is automatically cached across many nodes, creating distributed delivery that improves with demand — the opposite of traditional hosting, where popular content strains the hosting server.
The Distributed Hash Table (DHT) maintains a mapping between CIDs and the nodes hosting corresponding content. When a client requests a CID, the DHT identifies nodes holding that content, and the data is transferred directly from those nodes.
Pinning and Persistence
IPFS does not guarantee data persistence by default. Nodes may garbage-collect cached content they no longer need, and data persists only as long as at least one node chooses to host it. Pinning instructs a node to retain specific content indefinitely, ensuring continued availability.
Pinning services — both centralised (Pinata, Infura) and decentralised (Filecoin) — provide persistence guarantees for content that must remain available regardless of individual node decisions. Filecoin, also developed by Protocol Labs, adds economic incentives for persistent storage atop IPFS infrastructure, compensating storage providers for reliably hosting data over specified time periods.
Web3 Applications
IPFS serves as critical infrastructure across the Web3 ecosystem.
NFT media and metadata — NFTs reference their associated media (images, videos, 3D models) and descriptive metadata through IPFS CIDs. Content addressing ensures that the media associated with an NFT cannot be modified after minting — the CID recorded in the NFT smart contract will always resolve to the original media. This is essential for NFT art and real-world asset tokenisation where data integrity underpins value.
DApp front-end hosting — Decentralised applications can host their user interfaces on IPFS, reducing dependence on centralised hosting providers. An IPFS-hosted DApp front-end remains accessible even if its development team’s servers go offline or are subject to censorship.
Decentralised websites — Entire websites can be hosted on IPFS and accessed through IPFS gateways or native IPFS-compatible browsers. Combined with blockchain-based naming systems (ENS, Unstoppable Domains), IPFS enables websites that are censorship-resistant and server-independent.
Document storage — Legal documents, governance proposals, and DAO records can be stored on IPFS with on-chain CID references, ensuring document integrity and availability without relying on centralised document management systems.
Limitations
Content availability depends on at least one node hosting the requested CID. Unpopular content not explicitly pinned may become unavailable as caching nodes garbage-collect.
Performance for initial content retrieval can be slower than CDN-backed centralised hosting, particularly for content hosted by few nodes. Subsequent requests benefit from caching, but first-access latency is a practical limitation.
Mutable content is awkward under content addressing. Because CIDs change with any content modification, applications requiring updateable content must use IPNS (InterPlanetary Name System) or similar resolution layers that map persistent names to changing CIDs — adding complexity.
Storage costs — While IPFS itself is free to use, reliable persistence requires pinning services or Filecoin storage, which carry costs comparable to (though structured differently from) centralised alternatives.
Despite these limitations, IPFS provides foundational infrastructure for a decentralised storage ecosystem that continues to mature in performance, reliability, and accessibility.
Donovan Vanderbilt is a contributing editor at ZUG WEB3, the decentralised protocol intelligence publication of The Vanderbilt Portfolio AG, Zurich. He covers Web3 fundamentals, decentralised protocols, and the infrastructure underpinning the decentralised internet.