

Irys Blog

2025-01-01
Technical
Irys vs Walrus: Built in vs Built on

At a glance, Irys and Walrus look similar—decentralized data, execution, economic incentives, and tokens. But the resemblance is skin-deep. Under the hood, their protocols, assumptions, and tradeoffs are wildly different.
Let's dive in.
Protocol
Irys doesn't just store data. It is a full blockchain.
Where other protocols are pieced together from separate systems, Irys is one seamless stack.
Decentralized Storage Networks (DSNs) like Walrus—and IPFS before it—focus on decentralizing data storage. But they rely on external systems to handle computation, reward distribution, and other operations.
Irys is different. It's a self-contained datachain—a blockchain optimized for storing and computing on data.
As a complete blockchain, Irys includes its own miners, native token, and integrated layers for storage, execution, and security. Miners do it all. They store data, execute smart contracts, and secure the network using Irys' storage-based proof-of-useful-work.
DSNs don't have that luxury. They depend on external infrastructure & systems for key pieces of their stack. Pinata, a centralized company, charges monthly for IPFS pinning (making sure the data is available via IPFS). Walrus uses Sui's CoinStandard for its $WAL token and smart contracts for reward logic. These dependencies impose constraints. And with it, complexity.
Walrus is a real improvement over earlier DSNs. It replaced centralized reward distribution with smart contracts. It stitched together an impressive system using Sui as a base layer. It works. But it's still a hybrid solution with multiple dependencies.
Walrus storage nodes are not part of the Sui blockchain. They use a separate token. They don't participate in execution, consensus, or security. That separation matters.
Irys, by comparison, is vertically integrated. It doesn't need to navigate constraints or depend on other tokens because the protocol owns the full stack.
That's the difference. Irys is built in. Walrus is built on top.
Irys Storage & Execution flowchart
Tokenomics
Another key difference between Walrus and Irys is their tokenomics. Walrus uses a dual token strategy. Storage is paid for with WAL and execution with SUI. Because the bookkeeping for each "blob" stored by Walrus is managed in Sui smart contracts, paying for storage on Walrus requires both $WAL and $SUI.
This is different form Irys where both storage and execution happen within the same protocol and are provided by the same nodes using the same token. This tight integration of storage and execution allows Irys to offer all of its utility, storage + computation, using a single token, simplifying pricing and developer experience when building on Irys.
A byproduct of this design decision is that Irys miners can offer storage at near cost while earning fees from computing smart contract execution and programmable data. No matter what type of operation being performed (storage or computation) the value accrues to Irys' native token and funds mining operations. For Walrus, the protocol has to pay all its storage nodes (and staked token holders wanting yield) with profit margins on storage. That means, Walrus has to charge a premium on data storage to pay as yield to token holders and storage node operators. Any fees from execution are paid in SUI and go to the Sui network. This means that storage on Walrus will always come at a premium, where storage on Irys can be priced at the physical cost of the storage.
Data Durability
Walrus and Irys take fundamentally different approaches to data storage and the fault tolerance of their data. Walrus uses erasure coding ("Red Stuff"), expanding data 4–5x and slicing it into fragments ("slivers") spread across multiple storage nodes. This enables data recovery even if 2/3 of nodes are offline. However, it introduces significant bookkeeping overhead: each sliver and its assignment must be tracked in a Sui smart contract, incurring execution cost and complexity. Encoding also only happens when data is uploaded—Walrus doesn't reward storage nodes for providing empty capacity.
Irys instead stores 10 full replicas of each data segment. Miners pre-fill drives with cryptographic randomness tied to their mining address, enabling them to persist data at disk speed with no re-encoding. Ten staked miners are assigned to each 16TB partition in the storage layer, and failing a storage challenge results in slashing and reassignment. This shifts complexity from tracking slivers to simple replica assignment—only 10 addresses per 16TB of storage need to be maintained.
Walrus offers strong fault tolerance in environments with low trust but adds coordination and bookkeeping overhead. Irys' ability to encode capacity before data arrives and its storage based mining mechanism allow for simpler logistics and better scalability. While Irys stores 10x the data, it minimizes protocol-layer complexity, making it efficient for high-throughput, low overhead storage. This combined with the ability to offer storage at near cost, enables Irys to provide twice as much data replication as Walrus but at a substantially lower cost.
Walrus Decentralized Storage Network flowchart
Programmable Data
Programmable Data is the ability to reference storage layer data from smart contracts running in the execution layer and perform computation on it.
Walrus supports a form of programmable data, but it's important to note that Walrus storage nodes do not execute smart contracts—Sui validators do.
Even so, Walrus has found an ingenious way to enable this functionality. The cryptographic fingerprints of all Walrus blobs are stored in a bookkeeping contract on Sui. When a developer wants to use Walrus blob data inside a Sui smart contract, they include the relevant portion of the blob in the transaction.
The contract then checks the provided data against the stored cryptographic fingerprint. If the check passes, the contract can proceed with computation—no need for a trusted oracle or third party.
This is a clever workaround, but it adds several constraints. Developers must manually retrieve and submit blob data, and the amount of data is limited by Sui's transaction size. There is also the overhead of cross-contract calls to validate the fingerprint.
Irys, by contrast, handles Programmable Data natively. The caller only needs to specify the offsets of the storage layer data they want to reference. Miners then coordinate to retrieve the data directly from the storage layer without requiring the user to supply it. This creates a smoother developer experience with fewer moving parts and lower overhead costs.
Storage Term
The final difference is storage duration—or as we call it, the storage term. This is where Irys truly creates long-term value.
Irys supports both short-term and permanent onchain storage. Its sustainable endowment model funds rewards for storing data over the next 200+ years. As storage costs fall—say, more than 1% per year—those savings extend the term. Advances in storage tech directly increase how long data can be preserved.
Walrus takes a different path. It defines storage time in 14-day epochs, with a maximum prepayment of 53 epochs—just under two years. Extensions are possible but not protocol-native. Users are expected to re-upload data or trigger SUI-based renewals. The mechanism is vaguely defined and left to developers to implement.
That's a critical divergence. On Irys, the protocol guarantees storage payouts over time. On Walrus, that responsibility shifts to individual developers. Each app must create its own system to ensure long-term availability. This leads to inconsistent guarantees.
With Irys, permanent means permanent. The same level of durability and assurance applies across all apps. That's the power of protocol-level permanence.
Conclusion
At first glance, Irys and Walrus might appear to offer similar value—decentralized data, onchain logic, and token-based incentives. But that similarity is surface-level. Underneath, they represent fundamentally different approaches to protocol design.
Irys is a vertically integrated datachain. It handles storage, execution, and security with a single protocol, token, and mining mechanism. This integration simplifies developer experience, reduces overhead, and creates a unified economic system. It also unlocks features like programmable data and permanent storage that are native to the protocol itself.
Walrus, like other DSNs, distributes these responsibilities across multiple systems. While it has made impressive strides working within SUI's framework, it's ultimately limited by its dependencies.
The result is clear: Irys isn't just storing data, it's redefining what decentralized infrastructure can be. For builders who care about cost-efficiency, long-term durability, and simplified architecture unconstrained by multiple dependencies, Irys creates a clear path forward.
One chain. One token. One protocol. Irys unlocks the value of data.