Content Storage: Fully On-Chain (calldata) vs Decentralized Storage (IPFS/Arweave)

Data lives on the base layer: Content is embedded in transaction calldata (e.g., Ethereum, Arbitrum, Optimism) or directly in state. This guarantees immutability and availability as long as the chain exists. It enables native, permissionless composability for smart contracts, allowing protocols like Art Blocks or Autoglyphs to render directly from chain data.

Expensive at scale: Storing 1MB on Ethereum L1 calldata can cost $100+ during peak congestion. While L2s like Arbitrum reduce costs, they remain prohibitive for large assets (e.g., videos, high-res images). Costs are tied directly to gas price volatility, making long-term budgeting difficult.

Pay once, store forever: Protocols like Arweave offer permanent storage for a one-time, upfront fee (~$5-10 per GB). IPFS (with pinning services like Pinata, Filecoin) provides scalable, content-addressed storage at a fraction of on-chain cost. Ideal for NFT metadata, dApp frontends, and large datasets.

Smart contracts reference off-chain pointers (e.g., IPFS CID, Arweave TX ID). This introduces liveness dependencies on external networks and pinning services. If pinning lapses (IPFS) or the storage network fails, assets become unavailable, breaking dApps. Requires careful oracle or gateway design for reliable access.

• Maximalist NFTs & Generative Art: Where the art is the contract (e.g., Chain Runners, Loot).
• Critical Protocol Logic: Small, immutable configuration data or proofs that must be 100% available for contract execution.
• Ultimate Censorship Resistance: Data cannot be decoupled from the chain's security.

• NFT Metadata & Media: The standard model (ERC-721, ERC-1155) for cost-effective collections like Bored Ape Yacht Club.
• Static dApp Frontends: Hosting websites on IPFS via Fleek or Arweave via Bundlr.
• Archival Data & Large Files: Historical logs, datasets, or video content where on-chain cost is untenable.

Absolute data availability: Content stored in Ethereum calldata or EIP-4844 blobs is guaranteed by the network's consensus. This is critical for NFT metadata (e.g., CryptoPunks), on-chain games (e.g., Dark Forest), and protocol-critical logic where data must be inseparable from the chain state.

Zero-trust access within the EVM: Smart contracts can read and verify data directly without external oracles or bridges. Essential for decentralized autonomous organizations (DAOs) storing proposals, on-chain randomness (e.g., Chainlink VRF), and minimal-trust applications that require proofs against historical state.

High marginal cost for large data: Ethereum blob storage costs ~$0.10 - $1.00 per 128KB, making it prohibitive for media files. Throughput is constrained by block/block space, unsuitable for high-frequency, large-scale data like video or extensive game assets. Use for critical, small payloads only.

Orders of magnitude cheaper for bulk data: Storing 1GB on Arweave costs a one-time fee of ~$10-20, versus thousands on-chain. Ideal for NFT collections (e.g., Bored Ape Yacht Club using IPFS), dApp frontends, and archival data where cost-per-byte is a primary constraint.

Designed for arbitrary file types: Natively handles images, video, and large datasets. Protocols like IPFS with Filecoin for incentivized persistence or Arweave for permanent storage enable use cases like decentralized social media (Lens Protocol), music NFTs, and research datasets.

Persistence depends on external incentives: IPFS pins can lapse; Arweave relies on endowment model. Requires a bridge (like a smart contract storing a CID) to the blockchain, introducing a trust assumption. Not suitable for data that must be live-verified in a smart contract without an oracle.

Permanent, verifiable persistence: Data is stored directly in transaction calldata (e.g., Ethereum, Arbitrum) or contract state (e.g., Solana). This ensures cryptographic permanence tied to the chain's consensus. This matters for high-value digital artifacts (e.g., Art Blocks NFTs) and core protocol logic where data availability must equal settlement layer security.

Prohibitively expensive for large files: Storing 1MB on Ethereum calldata can cost $100+ during high congestion. This scales linearly with data size. This matters for applications with rich media (video, high-res images) or frequent data updates, making it economically unfeasible for most non-financial data.

Orders of magnitude cheaper: Storing 1GB on Arweave costs a one-time fee of ~$8. IPFS via pinning services (Pinata, Filebase) is subscription-based (~$20/month for 100GB). This matters for scaling to millions of users (NFT collections, social dApps) and storing large datasets (game assets, research data) without L1 gas constraints.

Persistence models vary: Arweave uses endowment for permanent storage (200+ year target). IPFS is a peer-to-peer network; content disappears unless pinned by a persistent node (via Pinata, Infura, or a dedicated provider). This matters for long-term assurance; you trade absolute blockchain guarantee for cost savings, introducing a provider dependency for IPFS.