What can you actually build on Xandeum?

Nobody uses an app because its storage is decentralized. People use apps that solve a problem. So the right question is not what Xandeum is, but what it lets you build that you could not build before. Here is the honest map: the use cases that pass that test, the receipts behind them, and the ones that do not need decentralized storage at all.

By Abdennour T Bada, Pulsar Network, operators of 12 pNodes across devnet and mainnet. We monitor the network live and publish what we see.

Why not just use AWS?

It is the right first question, and it deserves a one-breath answer. If you are hosting a website, an internal tool or a plain backup, AWS wins: cheaper, faster, more mature. Decentralized storage becomes interesting when the application itself is the trust model, because then the storage cannot belong to one company. With AWS, the company you are trusting is AWS.

That is mechanics, not ideology. A smart contract cannot open a support ticket, sign a new terms of service, or migrate a bucket when a provider changes its mind. A blockchain app that parks its data on one company's servers is decentralized in name only: the chain settles the money while a server somewhere holds everything the money is about. Every guarantee the chain makes stops at the edge of that server.

There is also a practical squeeze. On Solana, bytes a program can actually touch are scarce and expensive: rent exemption means locking roughly 6.96 SOL per megabyte as a refundable deposit that stays locked while the data exists, which is why nearly every app stores a link on-chain and the real data somewhere else. We walked through that gap in Where does data on Solana actually live? Xandeum's reason to exist is closing it: a storage layer that programs can read, write and verify at file scale, on a network no single company owns.

The three-question test

A quick filter we use before calling anything a real use case:

• Does a program need to touch the data? A smart contract or an autonomous agent reading and writing it directly, with no human in the loop.
• Must the data outlive any single company? Including the company that built the app.
• Does anyone need to prove the data is intact? Unchanged, uncensored, and still there.

Zero out of three: use a normal cloud. One: maybe. Two or three: this is exactly what a decentralized storage layer is for. The three lanes below score two or three.

1. NFTs and games that outlive their makers

The receipts here are not hypothetical. An academic study of 12,353 Ethereum NFT contracts covering 6.2 million tokens found that the off-chain assets of 25 percent of contracts were already unreachable. A 2024 study of top-selling OpenSea collections found a substantial share of metadata sitting on centralized servers, open to tampering and takedowns. And when the Foundation marketplace shut down, it committed to keep collectors' media pinned for just one more year and told users to start pinning their own files.

That is what owning an NFT often means today: you own a receipt that points at a file someone else is paying to keep alive. Games sharpen the same problem: world state, items, maps and player inventories are far too big for on-chain accounts, so they live on studio servers, and the studio's lifespan becomes the game's lifespan.

What changes with program-accessible storage: the artwork, the metadata and the game state live on the same decentralized layer the contract can read and write. The asset and the thing it represents become one object with one lifetime. Who this is for: NFT platforms, on-chain games, and anyone who sells "forever" but currently rents it by the month.

2. AI that remembers, and AI you can audit

AI agents are stateless by default: every session starts from zero. The industry is treating that as a core problem now. Anthropic launched a memory tool so agents can keep project state across sessions, and LangChain's docs call memory crucial for agents to learn and adapt. For an ordinary chatbot, a normal database solves this. For an agent holding a wallet, executing trades or acting for a DAO, it does not, because the memory itself has to be trustworthy: the mandate the agent follows, the log of what it decided, the data it acted on. That calls for storage that is persistent, program-readable and provably intact, which is precisely the shape of the on-chain agent economy.

The audit side is moving even faster. The EU AI Act now requires providers of new general-purpose models to publish summaries of their training data (older models are phased in by 2027). The C2PA content-credentials coalition passed 6,000 members and affiliates, and Canon is rolling out provenance signing for its newsroom cameras. All of it amounts to receipts for data, and receipts need somewhere tamper-evident to live. We covered the storage half of that story in AI training data provenance.

3. Real-world assets with the paperwork attached

Tokenized U.S. Treasuries alone now hold about $15 billion across 84 products. But a token without its documents is just a pointer: the deed, the invoice, the appraisal, the audit trail are what make the on-chain asset mean something, and today they mostly sit in PDFs on somebody's server, exactly the arrangement tokenization was supposed to retire.

What changes: the legal paper travels with the asset, stored on a layer the settlement contract can reference and anyone can verify, durable past any single custodian. Who this is for: RWA issuers, compliance teams, and anyone whose token is only as credible as its most fragile document link.

The quieter candidates

• Social you own: posts, media and follower graphs that belong to the user, not to a platform that can ban or fold.
• Personal data vaults: encrypted records (medical, identity, credentials) where apps request access and you hold the keys.
• DAO archives: proposals, treasury history and governance records, permanent and independently checkable.
• Data availability: rollups need cheap, verifiable places to publish data blobs. An adjacent lane with different requirements, but storage networks keep getting pulled into it.

What does not need Xandeum

Honesty is the differentiator, so: your static website does not need decentralized storage. Neither do your company backups if you already trust your vendor, nor hot video delivery, which is a CDN's job. If no program needs to touch the data and nobody needs to prove anything about it, a normal cloud is cheaper and faster, and pretending otherwise would make everything above less believable.