Data Wars: Walrus and Irys Explained

Written by: Ponyo & compiled by: Sui Network

Key Points

Architecture: Irys is a fully functional, all-in-one Layer 1 "data chain" that provides native blob (data block) access to contracts, but requires a new set of verification nodes. Walrus is an erasure coded storage layer built on Sui, which is easier to integrate but requires cross-layer coordination.

Economic Model: Irys uses a single token, IRYS, to unify payment fees and rewards. The user experience is simple, but the risk of price fluctuations is high. Walrus divides the function into two tokens: WAL (for storage) and SUI (for gas), which can effectively isolate costs, but requires maintaining two incentive systems.

Persistence and computing power: Irys maintains 10 complete copies and streams data directly into its virtual machine; Walrus uses erasure coding with about 5 times redundancy plus hash verification, which has a lower storage cost per GB, but the protocol implementation is more complex.

Adaptability: Irys provides a "pay once, store forever" donation model, which is very suitable for saving immutable data, but the upfront cost is high. Walrus adopts a "pay as you go, automatic renewal" leasing mechanism, which is easy to control costs and can be quickly integrated with Sui.

Adoption: Although Walrus is still in its early stages, it has developed rapidly, with PB-level storage, 100+ node operators, and has been adopted by multiple NFT and game brands. In contrast, Irys is still in the pre-expansion stage, the data volume has not reached the PB level, and the node network is still growing.

Both Walrus and Irys are committed to solving the same problem: providing reliable, incentivized on-chain data storage. But their design concepts are completely different: Irys is a Layer 1 blockchain built specifically for data storage, integrating storage, execution, and consensus into a vertically integrated architecture; while Walrus is a modular storage network that relies on Sui for coordination and settlement while running an independent off-chain storage layer.

Although the Irys team portrayed it as a better "built-in" solution in the initial comparison and defined Walrus as a limited "external" system, in reality both have their own advantages and disadvantages, and different trade-offs. Based on a technical perspective, this article objectively compares Walrus and Irys in 6 dimensions, refutes one-sided assertions, and provides developers with a clear selection guide to help them decide the most appropriate path based on cost, complexity, and development experience.

1. Protocol Architecture

1.1 Irys: Vertically Integrated L1

Irys embodies the classic "self-sufficient" concept. It comes with its own consensus mechanism, staking model, and execution virtual machine (IrysVM), which are tightly integrated with its storage subsystem.

A validator node plays three roles at the same time:

• Storing user data in the form of a complete copy;

• Executing smart contract logic in IrysVM;

• Securing the network through a hybrid PoW + staking mechanism.

Since these functions coexist in the same protocol, every layer from block headers to data retrieval rules can be optimized for large-volume data processing. Smart contracts can directly reference on-chain files, and storage proofs will also follow the consensus path for sorting ordinary transactions. The advantage is the high consistency of the architecture: developers only need to face a single trust boundary, a single fee asset (IRYS), and the experience of reading data in the contract code is like native support.

But the cost is a high startup cost. A brand new layer 1 network must recruit hardware operators, build indexers, launch block explorers, harden clients, and cultivate development tools from scratch. In the early days when the verification nodes were not yet strong, block time guarantees and economic security lagged behind the old chains. Therefore, Irys' architecture chose deeper data integration at the expense of ecosystem startup speed.

1.2 Walrus: Modular Overlay

Walrus takes a completely different path. Its storage nodes run off-chain, while Sui's high-throughput L1 is responsible for handling sorting, payments, and metadata through Move smart contracts. When a user uploads a blob (data block), Walrus shards it and stores it in each node, and then records an on-chain object on Sui with the content hash, shard allocation, and lease terms. Renewals, penalties, and rewards are all executed as ordinary Sui transactions, paying gas with SUI, but using WAL tokens as the storage economic settlement unit.

Relying on Sui, Walrus immediately gains the following advantages:

• Proven Byzantine fault-tolerant consensus mechanism;

• Complete development infrastructure;

• Powerful programmability;

• A basic token economy with liquidity;

• Many existing Move developers can integrate directly without protocol migration.

But the cost is cross-layer coordination. Every lifecycle event (upload, renewal, deletion) must be coordinated between two semi-independent networks. Storage nodes must trust the finality of Sui while maintaining performance when Sui is congested; Sui verification nodes do not review whether the actual disk stores data, so they must rely on Walrus's cryptographic proof system to ensure accountability. Compared with the integrated design, this architecture inevitably has higher latency, and part of the handling fee (SUI gas) will flow to the role that does not actually store the data.

1.3 Design Summary

Irys adopts a vertically integrated monolithic architecture, while Walrus is a horizontally layered integrated modular solution. Irys has greater architectural freedom and a unified trust boundary, but it needs to overcome the ecological construction difficulties brought about by cold start. Walrus uses Sui's mature consensus system to greatly reduce the access threshold for developers in the existing ecosystem, but it must deal with the coordination complexity of the two economic domains and operator systems. There is no absolute advantage or disadvantage between the two modes, but the optimization direction is different: one pursues consistency (coherence) and the other pursues composability (composability).

When the protocol selection depends on developer familiarity, ecological attractiveness or online speed, Walrus's layered model may be more realistic. When the bottleneck lies in the deep coupling of data and computing, or customized consensus logic is required, Irys, a chain designed specifically for data, also has enough reason to bear a heavier architectural burden.

2. Token Economics and Incentive Mechanism

2.1 Irys: One token drives the entire protocol stack

Irys’ native token IRYS covers the economic model of the entire platform:

• Storage fees: Users prepay IRYS to store data;

• Execution gas: All smart contract calls are also priced in IRYS;

• Miner Rewards: Block subsidies, storage proofs, and transaction fees are all paid in IRYS.

Since miners are responsible for both data storage and contract execution, computing income can make up for the lack of storage income. In theory, when DeFi activity on Irys is strong, computing income will offset data storage, thereby achieving services close to cost price; if contract traffic is low, the subsidy mechanism will be adjusted in the opposite direction. This cross-subsidy mechanism helps balance miner income and align incentives for various roles in the protocol. For developers, a unified asset means fewer custody processes and a more streamlined user experience, especially for scenarios where users are not expected to be exposed to multiple tokens.

But the disadvantage is the risk linkage of a single asset: once the price of IRYS falls, the rewards for computing and storage will decrease simultaneously, and miners will face a double squeeze. The economic security of the protocol and data persistence are therefore tied to the same price fluctuation curve.

2.2 Walrus: Dual Token Economic Model

Walrus splits functional responsibilities into two tokens:

• $WAL: The economic unit of the storage layer. Users pay for leasing space with WAL, and node operators receive WAL rewards by staking and storing data fragments, and the rewards are also linked to their delegated staking weight.

• $SUI: The gas token used to coordinate transactions on the chain. Any transaction such as uploading, renewal, and punishment on Sui consumes SUI and is rewarded to Sui verification nodes, not Walrus storage nodes.

This separation keeps the storage economy clear: the value of WAL is only affected by data storage needs and lease periods, and will not be disturbed by DEX transactions or NFT crazes on Sui. At the same time, Walrus can also inherit Sui's liquidity, cross-chain bridges, and fiat currency entry-most Sui builders already hold SUI, so the marginal cost of introducing WAL is low.

However, the dual-token model also has the problem of incentive split. Walrus nodes cannot participate in SUI's fee income, so the price of WAL must be sufficient to independently support hardware, bandwidth, and return expectations. If the price of WAL stagnates and SUI gas soars, the user's usage cost will increase, but the storage party will have no direct benefit. On the contrary, the outbreak of DeFi on Sui has driven the income of verification nodes, but it has nothing to do with Walrus nodes. Therefore, to maintain long-term balance, it is necessary to actively optimize the economic model: storage prices need to float flexibly according to hardware costs, demand cycles, and WAL market depth.

2.3 Design Summary

In short, Irys provides a unified and concise user experience, but centralizes risk; Walrus draws a clear boundary at the token level, bringing more sophisticated economic accounting, but needs to deal with two market systems and fee diversion issues. Builders should weigh whether they prefer a seamless experience or a separate management of economic risks to match their own product planning and funding strategies.

3. Data persistence and redundancy strategy

3.1 Walrus: Using erasure codes to achieve lightweight and high reliability

Walrus divides each data block (blob) into k data shards and adds m redundant check shards (using the RedStuff encoding algorithm). This technology is similar to RAID or Reed-Solomon encoding, but is optimized for decentralized and high-node-change environments. The original file can be reconstructed by taking any k of the k + m shards, which brings two advantages:

• High space efficiency: Under typical parameters (about 5 times expansion), the required storage space is reduced by half compared to the traditional 10 times replica replication scheme. In simple terms, to store 1GB of data on Walrus, the overall network capacity requires about 5GB (shards stored in multiple nodes), while a traditional full-copy system may require 10GB to achieve similar security.

• Strong on-demand repair capability: Walrus's encoding method not only saves space, but also bandwidth. When a node loses connection, the network only rebuilds the missing shards instead of the entire file, greatly reducing bandwidth overhead. This self-healing mechanism only needs to download data approximately equal to the size of the lost shard (i.e. O(blob_size/number of shards)), while traditional replica systems usually require O(blob_size) data.

The allocation of each shard and node will exist in the form of an object on Sui. Walrus rotates the pledge committee every epoch, challenges the node availability through cryptographic proofs, and automatically recodes when the node loss exceeds the safety threshold. Although this mechanism is complex (involving two networks, multiple shards, and frequent verification), it can achieve the highest durability with minimal capacity.

3.2 Irys: Conservative but robust multi-copy mechanism

Irys deliberately chose a more primitive and direct durability method: each 16TB data partition is fully stored by 10 pledge miners. The protocol prevents duplicate counting of the same hard disk by introducing the "salt value" (Matrix Packing technology) of specific miners. The system will continuously read and verify the node hard disk through "proof-of-useful-work" to ensure that every byte is real, otherwise the miner will be punished and the pledged assets will be deducted.

In actual operation, whether the data is available depends on: whether at least one of the 10 miners responds to the query? If a miner fails to verify, the system will immediately start re-replication to maintain the standard of 10 copies. The cost of this strategy is up to 10 times the data storage redundancy, but the logic is simple and clear, and all states are concentrated on one chain.

3.3 Design Summary

Walrus focuses on: coping with the problem of frequent node replacement through efficient encoding strategies and Sui's object model, thereby ensuring data persistence without increasing costs. Irys believes that: as hardware costs drop rapidly, more direct and heavier multi-copy mechanisms are more reliable and worry-free in actual engineering.

If you need to store PB-level archived data and can accept higher protocol complexity, Walrus's erasure code has more advantages in terms of per-byte economy. If you value simplicity of operation and maintenance (one chain, one proof, sufficient redundancy) and think that hardware expenditure is negligible relative to the speed of product delivery, Irys's 10-copy mechanism can provide durability with minimal thinking.

4. Programmable data and on-chain computing

4.1 Irys: Smart contracts with native support for data

Since storage, consensus mechanism and Irys Virtual Machine (IrysVM) share the same ledger, contracts can call the read_blob(id, offset, length) method as easily as reading their own state. During block execution, miners stream requested data fragments directly into the virtual machine, perform deterministic checks, and continue processing the results in the same transaction. No oracles, no user parameters, no off-chain transfers.

This programmable data structure enables the following use cases:

• Media NFTs: Metadata, high-resolution images, and royalty logic are all on-chain and enforced at the byte level.

• On-chain AI: Perform inference tasks directly on model weights stored in partitions.

• Big data analytics: Contracts can scan large data sets such as logs and genetic files without external bridges.

While gas costs increase with the number of bytes read, the user experience is still an IRYS-denominated transaction.

4.2 Walrus: "Verify first, then calculate" mode

Since Walrus cannot directly stream large files into the Move virtual machine, it adopts the "hash commitment + witness" design mode:

• When a user stores a blob, Walrus will record its content hash on Sui;

• After that, any caller can submit the corresponding data fragment and a lightweight proof that proves the correctness of the fragment (such as a Merkle path or a complete hash);

• The Sui contract will recalculate the hash and compare it with the Walrus metadata. If the verification is successful, the data is trusted and the subsequent logic is executed.

Advantages:

• Ready to use without any modification to the L1 protocol;

• Sui verification nodes do not need to perceive GB-level big data content.

Limitations:

• Manual data acquisition required: the caller must pull data from the Walrus gateway or node and package limited-length data fragments in the transaction (limited by Sui's transaction size);

• Shard processing overhead: For large data processing tasks, multiple micro-transactions, or off-chain preprocessing + on-chain verification are required;

• Double gas cost: users need to pay SUI gas (for verifying transactions) and WAL (indirectly pay for underlying storage fees).

4.3 Design Summary

If your application requires the contract to process several MB of data per block (such as on-chain AI, immersive media dApps, verifiable scientific computing processes, etc.), the embedded data API provided by Irys is more attractive.

If your scenario focuses more on data integrity proof, small media display, or heavy calculations occur off-chain and only the results need to be verified on-chain, Walrus can already do it.

So, this choice is not about "whether it can be achieved", but at which layer do you want to put the complexity: the protocol bottom layer (Irys) or the middleware application layer (Walrus)?

5. Storage period and permanence

5.1 Walrus: Pay-as-you-go leasing model

Walrus adopts a fixed-period leasing model. When uploading data, users use $WAL to pay for a fixed period of storage (charged at 14 days per epoch, and the longest one-time purchase is about 2 years). After the lease expires, if there is no renewal, the node can choose to delete the data. The application can write an automatic renewal script through the Sui smart contract to turn "leasing" into de facto "permanent storage", but the responsibility for renewal always belongs to the uploader.

The advantages are that users do not have to prepay for capacity that may be abandoned, and pricing can track real-time hardware costs. In addition, by setting data lease expiration times, the network can garbage collect data that is no longer paid for, preventing the accumulation of "permanent garbage". The disadvantages are: missed renewals or running out of funds will cause data to disappear; long-running dApps must run their own "keep-alive" robots.

5.2 Irys: Permanent storage guaranteed by the protocol layer

Irys provides a "permanent storage" option similar to Arweave. Users only need to pay $IRYS once, which can fund miners' storage services for hundreds of years in the future in the form of an on-chain fund (endowment) (assuming storage costs continue to fall, it can cover about 200 years). After completing the transaction, the responsibility for storage renewal is transferred to the protocol itself, and users no longer need to manage it.

The result is a "save once, use forever" user experience, which is very suitable for: NFTs, digital archives, and datasets that need to be tamper-proof (such as AI models). But its disadvantage is that the initial cost is high, and this model is highly dependent on the price health of $IRYS in the next few decades, which is not suitable for frequently updated data or temporary files.

5.3 Design Summary

If you want to control the data lifecycle and pay for actual use, please choose Walrus; if you need unshakable long-term data persistence and are willing to pay a premium for it, please choose Irys.

6. Network maturity and usage

6.1 Walrus: Production-grade scale

The Walrus mainnet has only been online for 7 epochs, but it has 103 storage operators and 121 storage nodes running, with a total of 1.01 billion WALs pledged. The network currently stores 14.5 million blobs (data blocks), triggers 31.5 million blob events, has an average object size of 2.16MB, and has a total storage volume of 1.11PB (about 26% of its 4.16PB physical capacity). The upload throughput is about 1.75KB/s, and the shard map covers 1,000 parallel shards.

The economic level also shows strong momentum:

• The market value is about US$600 million, and the FDV (fully diluted valuation) is US$2.23 billion;

• Storage price: about 55K Frost per MB (equivalent to about 0.055 WAL);

• Write price: about 20K Frost per MB

• The current subsidy ratio is as high as 80% to accelerate early growth

Several high-traffic brands have adopted Walrus, including Pudgy Penguins, Unchained and Claynosaurs, all of which have built asset pipelines or data archiving backends on it. The network currently has 105,000 accounts, 67 projects are being integrated, and it has supported PB-level data transmission for real-world scenarios such as NFTs and games.

6.2 Irys: Still in the early stages

According to the Irys public data panel (as of June 2025):

• Contract execution TPS ≈ 13.9, storage TPS ≈ 0

• Total storage data volume ≈ 199GB (officially claimed to have 280TB of space)

• Number of data transactions: 53.7 million (of which 13 million in June)

• Number of active addresses: 1.64 million

• Storage cost: $2.50 / TB / month (temporary storage), or $2.50 / GB (permanent storage)

• Miner system "coming soon" (uPoW mining mechanism not yet enabled)

Programmable data call fee is $0.02 per chunk, but actual data writes are still very limited because the permanent storage fund is not yet in place. Contract execution throughput is currently good, but batch storage capacity is still basically zero, reflecting its current focus on virtual machine functions and developer tools rather than data carrying capacity.

6.3 What the number represents

Walrus has reached PB-level scale, can generate revenue, and has been rigorously tested by consumer NFT brands. Irys is still in the early bootstrap stage, rich in features, but requires miners to join and meet data volume requirements.

For customers evaluating production readiness, Walrus currently performs as follows:

• Higher real usage: over 14 million blobs uploaded, PB-level data storage;

• Wider operating scale: over 100 operators, 1,000 shards, over $100 million in stake;

• Stronger ecological appeal: top Web3 projects are already integrating and using it;

• Clearer pricing system: WAL/Frost charges are clear and transparent, and the on-chain subsidy mechanism is visible.

Although Irys's integrated vision may play an advantage in the future (such as miners going online, permanent storage funds landing, and TPS improvement), Walrus has a more practical leading advantage in terms of quantifiable throughput, capacity, and customer usage at this stage.

7. Looking to the future

Walrus and Irys represent the two ends of the on-chain storage design spectrum:

• Irys integrates storage, execution, and economic models into an IRYS token and a dedicated L1 blockchain for data, providing developers with a frictionless on-chain big data access experience and a built-in "permanent storage" protocol-level commitment. Accordingly, development teams need to migrate to a young ecosystem and accept higher hardware resource consumption.

• Walrus builds an erasure-coded data storage layer on top of Sui, reusing mature consensus mechanisms, liquidity infrastructure, and development toolchains to achieve a very cost-effective storage cost per byte. However, its modular architecture also brings additional coordination complexity, dual-token experience, and continued attention to "lease renewal".

Which one to choose is not a question of "right or wrong", but depends on the bottleneck you care about most:

• If you need deep data and computing combination capabilities, or a "permanent preservation" commitment at the protocol level, then Irys's integrated design will be more suitable.

• If you value capital efficiency, the ability to quickly go online on Sui, or highly customized control over the data life cycle, Walrus' modular solution is a more pragmatic choice.

In the future, the two will likely coexist in parallel as the on-chain data economy continues to expand, serving different types of developers and application scenarios.