Inside Celestia’s Architecture
This module explores the technical foundations of Celestia, including its core functions, validator design, and the mechanics of Data Availability Sampling (DAS). It explains how Celestia stores data as blobs, how Blobstream allows cross-chain integrations with Ethereum, and how the network maintains decentralization and censorship resistance. The module illustrates how Celestia achieves scalability without compromising trust or accessibility.
Core Roles — What Celestia Actually Does
Celestia performs two fundamental functions within the blockchain stack: consensus and data availability. It deliberately omits smart contract execution, state storage, and settlement logic. This narrow scope is not a limitation, but a conscious architectural decision that aligns with Celestia’s modular design. By specializing in just two roles, Celestia achieves greater scalability, efficiency, and flexibility for external chains that rely on it as a foundational layer.
Consensus Without Execution
In Celestia, consensus refers to the process of ordering blocks and ensuring that validators agree on the order and inclusion of data blobs. Celestia uses a Byzantine Fault Tolerant (BFT) consensus algorithm derived from Tendermint, which is well-tested and designed to maintain network safety even in the presence of faulty or malicious nodes.
The critical difference between Celestia and a traditional Layer 1 chain like Ethereum is that Celestia validators do not interpret or execute the transactions they receive. They simply agree on the order in which the data appears and confirm that it is published. This reduces the computational burden on validators and eliminates the need for stateful computation, enabling faster and more scalable block production.
Data Availability as a Primary Function
Celestia’s second core role is data availability—ensuring that all published data (typically transaction blobs from external chains) is accessible to anyone in the network. Data availability is a fundamental requirement for any blockchain system because users and light clients must be able to verify that the data behind each block is complete and not censored or withheld.
Celestia introduces a novel solution to this challenge through Data Availability Sampling (DAS), which allows light clients to probabilistically verify that a full block of data is available by sampling small, randomly chosen parts of the data. This means that even devices with limited bandwidth and storage can independently verify that the chain is functioning honestly, significantly improving decentralization and lowering the barrier to network participation.
Stateless Infrastructure by Design
Celestia does not maintain global state or track user balances, smart contracts, or execution results. It does not validate whether a transaction is correct, nor does it provide any logic for dispute resolution or settlement finality. These responsibilities are fully delegated to the execution layers built on top of Celestia—whether they are general-purpose rollups, domain-specific chains, or experimental blockchain runtimes.
As a result, Celestia operates as a stateless infrastructure layer. It publishes and organizes data but remains agnostic to its content. This allows application developers to build highly customized chains using their preferred virtual machines (such as EVM, WASM, or SVM), consensus rules, and governance models, while still leveraging Celestia’s secure and scalable publication layer.
Enabling a Modular Ecosystem
Because Celestia decouples consensus and data availability from execution and settlement, it enables the creation of a modular blockchain ecosystem. Developers can launch execution environments (e.g., rollups or sovereign chains) without needing to build a full consensus protocol or worry about data availability. These components interact with Celestia as a base layer, relying on it for ordering and publishing their block data.
This approach removes the need for Layer 1 blockchains to be one-size-fits-all systems. Instead, it promotes horizontal scaling, where many independent chains can operate in parallel while sharing a common, lightweight data layer.
Data Availability Sampling (DAS)
In blockchain systems, data availability refers to the guarantee that all block data is accessible to every participant in the network. This is essential because, without access to the underlying data, users and validators cannot verify that a block’s transactions are valid. If even a small portion of block data is withheld, it becomes impossible to prove whether malicious activity has occurred.
In traditional monolithic blockchains, full nodes solve this problem by downloading and storing all block data. However, this approach becomes increasingly impractical as block sizes grow. As the cost of storing and transmitting data rises, fewer participants can afford to run full nodes, which undermines decentralization and increases reliance on centralized infrastructure providers.
Celestia addresses this problem using a technique called Data Availability Sampling (DAS). DAS enables light clients—nodes that do not store the full blockchain history—to verify that all block data is available without downloading the entire dataset. This innovation is central to Celestia’s scalability and modular architecture.
How DAS Works
DAS relies on a combination of erasure coding and random sampling. When a block is created on Celestia, its data is broken into small pieces, encoded using erasure coding, and arranged into a two-dimensional data square. Erasure coding introduces redundancy, allowing the original data to be reconstructed even if some parts are missing. This is similar to how data redundancy works in distributed storage systems.
Once the data square is constructed and published, light clients do not need to download the entire square. Instead, they randomly request small portions (or “samples”) of the data. If the block producer is honest and the data is genuinely available, then a sufficiently large number of random samples will return successfully. If any portion of the block is missing or withheld, then the probability of detection becomes high as more clients perform sampling.
The mathematical properties of erasure coding and sampling theory ensure that light clients can detect unavailable or incomplete data with a high level of confidence—without needing to trust any particular validator or full node. This makes DAS a trust-minimized solution to the data availability problem.
Scaling Without Compromising Decentralization
The primary benefit of DAS is that it allows Celestia to support larger block sizes and higher data throughput without increasing the hardware requirements for network participants. Light clients can operate on consumer-grade devices, including mobile phones and embedded systems, while still verifying that block data is being published correctly.
This ability to verify data availability without downloading full blocks is what enables horizontal scalability. Instead of a single chain handling all execution, thousands of independent chains (such as rollups or sovereign chains) can publish their data to Celestia, and users can verify that data without running expensive infrastructure. This design supports a decentralized and scalable multi-chain ecosystem.
Security and Trust Assumptions
DAS does not eliminate the need for consensus or honest validators, but it does significantly reduce the trust assumptions required to verify that the system is operating correctly. Light clients do not need to trust block producers or rely on centralized APIs. They independently sample the network and can detect any attempt to censor or hide data. This supports censorship resistance and reinforces the security model of chains that use Celestia for data availability.
DAS also makes it harder for malicious validators to get away with publishing incomplete blocks. Since detection is probabilistic and decentralized, an attacker cannot predict which parts of the block will be sampled. Even if only a few clients are performing sampling, the chances of detection remain high.
Blobs and Blobstream
Blob
In Celestia, data is not structured as traditional blockchain transactions. Instead, it is stored and published as blobs, which stands for binary large objects. A blob is an opaque piece of data—Celestia does not interpret or validate its contents. Blobs are simply committed to the blockchain for ordering and availability.
This is a departure from conventional Layer 1 chains like Ethereum, where each transaction contains both data and logic that the chain must interpret and execute. Celestia avoids execution entirely. It treats blobs as payloads submitted by external chains (such as rollups), and only guarantees that these blobs are available and placed in the correct order.
Blobs offer an efficient and minimal abstraction for publishing block data. Since Celestia validators are not required to understand the blob’s contents, the network achieves greater scalability and neutrality. Execution environments built on top of Celestia can define their own formats, virtual machines, and consensus rules without constraints from the base layer.
When a rollup or sovereign chain submits data to Celestia, it packages its transaction batch into a blob. This blob is then posted to Celestia’s data availability layer and included in a block. Each blob is assigned a namespace—a unique identifier that allows clients to filter and retrieve blobs relevant to their specific chain.
The blob is split into smaller shares, erasure-coded, and arranged in a two-dimensional data square. This structure enables light clients to perform Data Availability Sampling (DAS), ensuring that the entire blob is accessible without downloading every share. Once successfully posted, the blob becomes a permanent, ordered record on Celestia, accessible to anyone running a client or verifier.
Blob Stream
Blobstream is Celestia’s mechanism for relaying its blobs to other blockchain networks. It acts as a data availability bridge, allowing Layer 2 rollups or other execution environments to use Celestia for data publishing, while still relying on a separate chain—typically Ethereum—for settlement and fraud proofs.
In practice, Blobstream allows Ethereum smart contracts to verify that a given blob was indeed published on Celestia and made available. It accomplishes this using light clients and cryptographic proofs that bridge Celestia’s block headers and namespaces into Ethereum. This enables Ethereum-based rollups to benefit from Celestia’s scalable data layer without giving up Ethereum’s security and settlement guarantees.
Several projects are already using Blobstream to separate their execution logic from data availability. For example, Manta Pacific, a Layer 2 zkEVM chain, uses Celestia as a data availability layer via Blobstream, while continuing to settle on Ethereum. This approach enables reduced fees and faster block publishing while maintaining compatibility with Ethereum’s smart contract ecosystem.
Blobstream effectively creates a hybrid model: execution runs on a high-performance rollup, data availability is outsourced to Celestia, and settlement remains anchored to Ethereum. This architecture highlights how Celestia can be used not only for sovereign rollups but also for Ethereum-based L2s looking to offload bandwidth-heavy operations.
Security and Decentralization Model
Validator-Based Security with Proof-of-Stake
Celestia operates on a proof-of-stake consensus model, using a variant of the Tendermint BFT (Byzantine Fault Tolerant) algorithm. Validators in the network stake Celestia’s native token, $TIA, to participate in the block production and consensus process. These validators are responsible for proposing and signing blocks that include blobs of data submitted by rollups and other chains.
The use of proof-of-stake aligns incentives between network participants and protocol security. Validators are rewarded for honest behavior and penalized for malicious actions such as double-signing or failing to participate in consensus. This economic mechanism ensures that validators act in the network’s best interest and provides a strong base layer for ordering and securing published data.
Unlike traditional chains, Celestia validators do not execute transactions or track global state. Their responsibilities are limited to verifying signatures, ordering data blobs, and ensuring that blocks meet the formatting and data availability requirements. This reduces their computational load, enabling the network to scale without increasing hardware demands.
Lightweight Clients and Trust-Minimized Verification
One of Celestia’s most important security features is its support for light clients—nodes that do not store the full blockchain history or execute transactions but still verify block inclusion and data availability. Using Data Availability Sampling (DAS), light clients can independently confirm that all block data is accessible without downloading it in full.
This means users and application chains do not need to trust full nodes or rely on centralized infrastructure providers to ensure network integrity. They can operate their own light clients on everyday devices, such as smartphones or consumer laptops, and still verify that Celestia is functioning honestly.
This model significantly improves decentralization. In many blockchains, full nodes have become prohibitively expensive to run, concentrating validation power in a few hands. Celestia’s lightweight verification makes network participation accessible to a wider range of users, enhancing the robustness and distribution of trust.
Censorship Resistance Through DAS
Censorship resistance is a critical property for any blockchain network. It ensures that all users can publish data and that no entity can suppress or selectively hide transactions. In Celestia, Data Availability Sampling plays a central role in protecting against censorship.
Because data blobs are split into smaller pieces and redundantly encoded using erasure coding, a malicious validator would need to withhold a significant portion of a block to successfully censor it. However, DAS makes such behavior easily detectable. Light clients request random shares of data, and if any part of the blob is missing, clients can flag the block as incomplete or unavailable.
This system creates a powerful incentive for validators to publish complete and honest data. The probabilistic nature of DAS ensures that even partial censorship is likely to be detected, especially as the number of sampling clients increases.
Low Hardware Requirements and Accessible Participation
Another dimension of decentralization is hardware accessibility. In many blockchain networks, the cost of running a validator or full node includes significant storage, bandwidth, and compute power. This raises barriers for everyday users and concentrates validation responsibilities among a small number of institutional actors.
Celestia’s minimalist design avoids these problems. Validators do not perform execution, and light clients do not require full storage of the chain. The result is a network that can be secured and verified using standard hardware, without specialized or expensive equipment. This enables broader participation across geographies and income levels, contributing to a healthier and more decentralized network.
Inside Celestia’s Architecture
This module explores the technical foundations of Celestia, including its core functions, validator design, and the mechanics of Data Availability Sampling (DAS). It explains how Celestia stores data as blobs, how Blobstream allows cross-chain integrations with Ethereum, and how the network maintains decentralization and censorship resistance. The module illustrates how Celestia achieves scalability without compromising trust or accessibility.
Core Roles — What Celestia Actually Does
Celestia performs two fundamental functions within the blockchain stack: consensus and data availability. It deliberately omits smart contract execution, state storage, and settlement logic. This narrow scope is not a limitation, but a conscious architectural decision that aligns with Celestia’s modular design. By specializing in just two roles, Celestia achieves greater scalability, efficiency, and flexibility for external chains that rely on it as a foundational layer.
Consensus Without Execution
In Celestia, consensus refers to the process of ordering blocks and ensuring that validators agree on the order and inclusion of data blobs. Celestia uses a Byzantine Fault Tolerant (BFT) consensus algorithm derived from Tendermint, which is well-tested and designed to maintain network safety even in the presence of faulty or malicious nodes.
The critical difference between Celestia and a traditional Layer 1 chain like Ethereum is that Celestia validators do not interpret or execute the transactions they receive. They simply agree on the order in which the data appears and confirm that it is published. This reduces the computational burden on validators and eliminates the need for stateful computation, enabling faster and more scalable block production.
Data Availability as a Primary Function
Celestia’s second core role is data availability—ensuring that all published data (typically transaction blobs from external chains) is accessible to anyone in the network. Data availability is a fundamental requirement for any blockchain system because users and light clients must be able to verify that the data behind each block is complete and not censored or withheld.
Celestia introduces a novel solution to this challenge through Data Availability Sampling (DAS), which allows light clients to probabilistically verify that a full block of data is available by sampling small, randomly chosen parts of the data. This means that even devices with limited bandwidth and storage can independently verify that the chain is functioning honestly, significantly improving decentralization and lowering the barrier to network participation.
Stateless Infrastructure by Design
Celestia does not maintain global state or track user balances, smart contracts, or execution results. It does not validate whether a transaction is correct, nor does it provide any logic for dispute resolution or settlement finality. These responsibilities are fully delegated to the execution layers built on top of Celestia—whether they are general-purpose rollups, domain-specific chains, or experimental blockchain runtimes.
As a result, Celestia operates as a stateless infrastructure layer. It publishes and organizes data but remains agnostic to its content. This allows application developers to build highly customized chains using their preferred virtual machines (such as EVM, WASM, or SVM), consensus rules, and governance models, while still leveraging Celestia’s secure and scalable publication layer.
Enabling a Modular Ecosystem
Because Celestia decouples consensus and data availability from execution and settlement, it enables the creation of a modular blockchain ecosystem. Developers can launch execution environments (e.g., rollups or sovereign chains) without needing to build a full consensus protocol or worry about data availability. These components interact with Celestia as a base layer, relying on it for ordering and publishing their block data.
This approach removes the need for Layer 1 blockchains to be one-size-fits-all systems. Instead, it promotes horizontal scaling, where many independent chains can operate in parallel while sharing a common, lightweight data layer.
Data Availability Sampling (DAS)
In blockchain systems, data availability refers to the guarantee that all block data is accessible to every participant in the network. This is essential because, without access to the underlying data, users and validators cannot verify that a block’s transactions are valid. If even a small portion of block data is withheld, it becomes impossible to prove whether malicious activity has occurred.
In traditional monolithic blockchains, full nodes solve this problem by downloading and storing all block data. However, this approach becomes increasingly impractical as block sizes grow. As the cost of storing and transmitting data rises, fewer participants can afford to run full nodes, which undermines decentralization and increases reliance on centralized infrastructure providers.
Celestia addresses this problem using a technique called Data Availability Sampling (DAS). DAS enables light clients—nodes that do not store the full blockchain history—to verify that all block data is available without downloading the entire dataset. This innovation is central to Celestia’s scalability and modular architecture.
How DAS Works
DAS relies on a combination of erasure coding and random sampling. When a block is created on Celestia, its data is broken into small pieces, encoded using erasure coding, and arranged into a two-dimensional data square. Erasure coding introduces redundancy, allowing the original data to be reconstructed even if some parts are missing. This is similar to how data redundancy works in distributed storage systems.
Once the data square is constructed and published, light clients do not need to download the entire square. Instead, they randomly request small portions (or “samples”) of the data. If the block producer is honest and the data is genuinely available, then a sufficiently large number of random samples will return successfully. If any portion of the block is missing or withheld, then the probability of detection becomes high as more clients perform sampling.
The mathematical properties of erasure coding and sampling theory ensure that light clients can detect unavailable or incomplete data with a high level of confidence—without needing to trust any particular validator or full node. This makes DAS a trust-minimized solution to the data availability problem.
Scaling Without Compromising Decentralization
The primary benefit of DAS is that it allows Celestia to support larger block sizes and higher data throughput without increasing the hardware requirements for network participants. Light clients can operate on consumer-grade devices, including mobile phones and embedded systems, while still verifying that block data is being published correctly.
This ability to verify data availability without downloading full blocks is what enables horizontal scalability. Instead of a single chain handling all execution, thousands of independent chains (such as rollups or sovereign chains) can publish their data to Celestia, and users can verify that data without running expensive infrastructure. This design supports a decentralized and scalable multi-chain ecosystem.
Security and Trust Assumptions
DAS does not eliminate the need for consensus or honest validators, but it does significantly reduce the trust assumptions required to verify that the system is operating correctly. Light clients do not need to trust block producers or rely on centralized APIs. They independently sample the network and can detect any attempt to censor or hide data. This supports censorship resistance and reinforces the security model of chains that use Celestia for data availability.
DAS also makes it harder for malicious validators to get away with publishing incomplete blocks. Since detection is probabilistic and decentralized, an attacker cannot predict which parts of the block will be sampled. Even if only a few clients are performing sampling, the chances of detection remain high.
Blobs and Blobstream
Blob
In Celestia, data is not structured as traditional blockchain transactions. Instead, it is stored and published as blobs, which stands for binary large objects. A blob is an opaque piece of data—Celestia does not interpret or validate its contents. Blobs are simply committed to the blockchain for ordering and availability.
This is a departure from conventional Layer 1 chains like Ethereum, where each transaction contains both data and logic that the chain must interpret and execute. Celestia avoids execution entirely. It treats blobs as payloads submitted by external chains (such as rollups), and only guarantees that these blobs are available and placed in the correct order.
Blobs offer an efficient and minimal abstraction for publishing block data. Since Celestia validators are not required to understand the blob’s contents, the network achieves greater scalability and neutrality. Execution environments built on top of Celestia can define their own formats, virtual machines, and consensus rules without constraints from the base layer.
When a rollup or sovereign chain submits data to Celestia, it packages its transaction batch into a blob. This blob is then posted to Celestia’s data availability layer and included in a block. Each blob is assigned a namespace—a unique identifier that allows clients to filter and retrieve blobs relevant to their specific chain.
The blob is split into smaller shares, erasure-coded, and arranged in a two-dimensional data square. This structure enables light clients to perform Data Availability Sampling (DAS), ensuring that the entire blob is accessible without downloading every share. Once successfully posted, the blob becomes a permanent, ordered record on Celestia, accessible to anyone running a client or verifier.
Blob Stream
Blobstream is Celestia’s mechanism for relaying its blobs to other blockchain networks. It acts as a data availability bridge, allowing Layer 2 rollups or other execution environments to use Celestia for data publishing, while still relying on a separate chain—typically Ethereum—for settlement and fraud proofs.
In practice, Blobstream allows Ethereum smart contracts to verify that a given blob was indeed published on Celestia and made available. It accomplishes this using light clients and cryptographic proofs that bridge Celestia’s block headers and namespaces into Ethereum. This enables Ethereum-based rollups to benefit from Celestia’s scalable data layer without giving up Ethereum’s security and settlement guarantees.
Several projects are already using Blobstream to separate their execution logic from data availability. For example, Manta Pacific, a Layer 2 zkEVM chain, uses Celestia as a data availability layer via Blobstream, while continuing to settle on Ethereum. This approach enables reduced fees and faster block publishing while maintaining compatibility with Ethereum’s smart contract ecosystem.
Blobstream effectively creates a hybrid model: execution runs on a high-performance rollup, data availability is outsourced to Celestia, and settlement remains anchored to Ethereum. This architecture highlights how Celestia can be used not only for sovereign rollups but also for Ethereum-based L2s looking to offload bandwidth-heavy operations.
Security and Decentralization Model
Validator-Based Security with Proof-of-Stake
Celestia operates on a proof-of-stake consensus model, using a variant of the Tendermint BFT (Byzantine Fault Tolerant) algorithm. Validators in the network stake Celestia’s native token, $TIA, to participate in the block production and consensus process. These validators are responsible for proposing and signing blocks that include blobs of data submitted by rollups and other chains.
The use of proof-of-stake aligns incentives between network participants and protocol security. Validators are rewarded for honest behavior and penalized for malicious actions such as double-signing or failing to participate in consensus. This economic mechanism ensures that validators act in the network’s best interest and provides a strong base layer for ordering and securing published data.
Unlike traditional chains, Celestia validators do not execute transactions or track global state. Their responsibilities are limited to verifying signatures, ordering data blobs, and ensuring that blocks meet the formatting and data availability requirements. This reduces their computational load, enabling the network to scale without increasing hardware demands.
Lightweight Clients and Trust-Minimized Verification
One of Celestia’s most important security features is its support for light clients—nodes that do not store the full blockchain history or execute transactions but still verify block inclusion and data availability. Using Data Availability Sampling (DAS), light clients can independently confirm that all block data is accessible without downloading it in full.
This means users and application chains do not need to trust full nodes or rely on centralized infrastructure providers to ensure network integrity. They can operate their own light clients on everyday devices, such as smartphones or consumer laptops, and still verify that Celestia is functioning honestly.
This model significantly improves decentralization. In many blockchains, full nodes have become prohibitively expensive to run, concentrating validation power in a few hands. Celestia’s lightweight verification makes network participation accessible to a wider range of users, enhancing the robustness and distribution of trust.
Censorship Resistance Through DAS
Censorship resistance is a critical property for any blockchain network. It ensures that all users can publish data and that no entity can suppress or selectively hide transactions. In Celestia, Data Availability Sampling plays a central role in protecting against censorship.
Because data blobs are split into smaller pieces and redundantly encoded using erasure coding, a malicious validator would need to withhold a significant portion of a block to successfully censor it. However, DAS makes such behavior easily detectable. Light clients request random shares of data, and if any part of the blob is missing, clients can flag the block as incomplete or unavailable.
This system creates a powerful incentive for validators to publish complete and honest data. The probabilistic nature of DAS ensures that even partial censorship is likely to be detected, especially as the number of sampling clients increases.
Low Hardware Requirements and Accessible Participation
Another dimension of decentralization is hardware accessibility. In many blockchain networks, the cost of running a validator or full node includes significant storage, bandwidth, and compute power. This raises barriers for everyday users and concentrates validation responsibilities among a small number of institutional actors.
Celestia’s minimalist design avoids these problems. Validators do not perform execution, and light clients do not require full storage of the chain. The result is a network that can be secured and verified using standard hardware, without specialized or expensive equipment. This enables broader participation across geographies and income levels, contributing to a healthier and more decentralized network.