Modular Blockchains: DA, Settlement, Execution Separated
Executive summary
Modular blockchain architecture, the thesis that the canonical four blockchain functions (consensus, data availability, settlement, execution) should be separated into specialized layers rather than monolithic chains, transitioned from theoretical framing in 2018-2021 into the dominant production architecture by 2026. Celestia launched its DA layer in October 2023 and onboarded dozens of rollups, EigenDA from EigenLayer scaled to comparable deployment counts on the back of restaked ETH security, Avail (formerly Polygon Avail) shipped mainnet in 2024, and Ethereum's blob-based DA via EIP-4844 (Dencun, March 2024) brought DA scaling natively to Ethereum. The execution-layer ecosystem fragmented further into rollup frameworks: OP Stack (Optimism), Arbitrum Orbit (Offchain Labs), Polygon CDK, and ZK Stack (Matter Labs), each producing dozens of L2/L3 deployments. Sovereign rollups, where the rollup chooses its own settlement layer, became a Celestia-anchored pattern. By 2026, the modular vs monolithic debate has not produced a clear winner; both architectures have credible adoption with distinct trade-offs.
The four-layer modular stack
Modular blockchain architecture decomposes a blockchain into four conceptually distinct layers. Consensus determines the canonical ordering and inclusion of transactions. Data Availability (DA) ensures that the data underlying state changes is available for verification. Settlement provides finality and dispute resolution for state changes. Execution applies state transitions to produce new state from old state plus transactions. In monolithic blockchains (Bitcoin, Solana, Avalanche C-chain), all four functions occur in a single tightly coupled system. In modular architectures, these functions are explicitly separated and provided by specialized layers. The motivation: each layer has different scaling and security trade-offs, and specialization allows each to be optimized independently. DA layers can be optimized for raw bandwidth (Celestia, EigenDA, Avail). Execution layers can be optimized for throughput and customization (rollups, appchains). Settlement layers can be optimized for security and finality (Ethereum L1). The thesis predicts that this specialization produces better aggregate performance than monolithic designs at the cost of complexity. The counter-thesis (held most prominently by Solana) is that monolithic optimization with hardware scaling produces simpler systems with comparable or superior performance. By 2026, both architectural philosophies have credible adoption, and the debate is more nuanced than 2022-era dichotomies suggested.
Celestia: the dedicated DA layer
Celestia, launched on mainnet in October 2023, was the first production-scale dedicated DA layer. The architecture: Celestia validators run a Tendermint-style consensus that orders 'blob' transactions (raw data without execution semantics) and uses Data Availability Sampling (DAS) so light nodes can efficiently verify data availability without downloading entire blocks. Rollups using Celestia as their DA layer post their transaction data to Celestia and use Celestia's data-availability commitments to ensure their state is verifiable. By 2026, Celestia hosted dozens of rollups including Arbitrum Orbit chains, Manta Pacific, Eclipse (Solana-VM rollup settling to Ethereum but using Celestia DA), Movement Labs (Move-VM rollup), and many others. Celestia's TIA token underwent its launch and reward distribution through 2024, and aggregate Celestia DA blob throughput grew steadily through 2025-2026 to substantial levels. The strategic positioning: Celestia provides cheaper DA than Ethereum L1 blobs (post-EIP-4844) for many use cases, particularly those willing to accept Celestia's security model rather than Ethereum's. The trade-off is that Celestia-using rollups inherit Celestia's economic security rather than Ethereum's, and bridging back to Ethereum L1 requires additional trust assumptions or ZK-light-client infrastructure.
EigenDA, Avail, and the DA competition
EigenDA, launched as the flagship AVS on EigenLayer in 2024, took a different approach: rather than running its own consensus, EigenDA leverages restaked ETH as economic security and provides DA through a network of operator nodes. The architecture is more complex than Celestia's (operators encode and store data with KZG commitments, light clients verify via sampling), but the security model directly inherits Ethereum staking economics rather than requiring a separate token. By 2026, EigenDA was deployed by several Optimism Superchain rollups and a growing list of Layer 3s, with aggregate throughput meaningfully scaled. Avail, originally developed inside Polygon and spun out as an independent project in 2023, launched its mainnet in 2024 with a Substrate-based architecture and KZG-commitment-based DA. Avail's positioning emphasized neutrality and broad rollup-framework compatibility. By Q1 2026, the DA layer competition was four-way: Ethereum (via EIP-4844 blobs), Celestia, EigenDA, and Avail, each with overlapping but differentiated customer bases. Ethereum blobs remained the security-conscious default for high-value rollups; Celestia, EigenDA, and Avail competed on cost, throughput, and ecosystem fit. Aggregate DA blob throughput across these four layers grew several-fold between 2024 and 2026, reflecting genuine rollup-driven demand rather than speculative supply.
Rollup frameworks: OP Stack, Arbitrum Orbit, Polygon CDK, ZK Stack
On the execution-layer side, the 2023-2025 period saw the emergence of rollup frameworks: standardized stacks that allow developers to deploy their own L2 or L3 rollups with relatively little engineering work. Optimism's OP Stack, the framework underlying the Optimism Superchain, became the most-deployed: Base (Coinbase), Worldchain (Tools for Humanity), Mode, Zora, and dozens of other production chains use the OP Stack with shared messaging via the OP Superchain protocol. Arbitrum Orbit, the framework underlying the Arbitrum ecosystem, allows L3s to settle to Arbitrum One or Nova and access Arbitrum's tooling; Orbit chains include Xai, Treasure Chain, and several gaming-focused deployments. Polygon CDK, the ZK-rollup-focused framework, allows developers to deploy custom zkEVMs with optional integration via the AggLayer for cross-chain liquidity. ZK Stack, by Matter Labs (zkSync), positions similarly for ZK-based deployments. By 2026, hundreds of rollup chains had deployed across these four frameworks, with the OP Stack leading by deployment count and aggregate TVL. The strategic implication is that rollup deployment became commoditized: a project that needed dedicated execution capacity could spin up an L2 or L3 in days rather than years. The deeper question - whether all these chains have meaningful long-term traffic and revenue - is the binding test through 2027.
Sovereign rollups and the Celestia model
Sovereign rollups, a concept pioneered by the Celestia team, are rollups that use a DA layer for data availability but do not settle state on a separate L1. Instead, the rollup itself is the canonical state, and any disputes are resolved by the rollup's own social consensus (typically by hard-forking, similar to how L1 chains resolve disputes). The model contrasts with smart-contract rollups, which post state to an L1 (typically Ethereum) and inherit the L1's settlement guarantees. The sovereign-rollup case: by avoiding L1 settlement, the rollup avoids dependency on any specific L1's economics or governance and gains flexibility in upgrade paths. The trade-off: bridge security to other chains is weaker because the rollup lacks an objective state oracle. By 2026, sovereign rollups remained a smaller architectural niche than smart-contract rollups but had production deployments, particularly in Cosmos-zone-adjacent contexts and in some appchains seeking deep customization. Movement Labs, Dymension, and several other sovereign-rollup-friendly platforms grew through 2024-2025. The Cosmos ecosystem more broadly (with its Inter-Blockchain Communication protocol) maintained an alternative modular philosophy where each appchain has its own consensus and uses IBC for interoperability rather than rollup-style settlement. The Cosmos vs Ethereum-rollup architectural debate persisted into 2026 without clear resolution.
Recent milestones (2024 to 2026)
Major milestones in the modular-blockchain stretch: Celestia mainnet launch (October 2023) and ecosystem growth through 2024-2026 including dozens of rollup integrations; EigenDA mainnet launch and AVS rollout (2024) with significant rollup-customer adoption; Avail mainnet (2024) and steady ecosystem growth; the Dencun hardfork (March 2024) shipping EIP-4844 blobs and the resulting collapse in L2 DA costs; Optimism Superchain expansion via Base, Worldchain, and dozens of OP Stack chains; Arbitrum Orbit deployment growth; Polygon CDK launch and AggLayer development; ZK Stack maturation with multiple Hyperchains; Arbitrum Stylus enabling Rust-based smart contracts; Eclipse mainnet launch as a Solana-VM rollup using Celestia DA and Ethereum settlement; Movement Labs Move-VM rollup mainnet; the broader appchain proliferation across Cosmos and EVM ecosystems; the rise of intent-based interoperability (Across, LayerZero, Wormhole, Hyperlane, Skip) connecting fragmented modular systems. By Q1 2026, aggregate L2 and rollup TVL crossed $50B, with execution distributed across hundreds of chains. The modular architecture had clearly produced more chains; whether it had produced proportionally more economic activity was the open question.
Risks, monolithic comparison, and outlook through 2027
Modular architectures face three structural critiques. First, fragmentation: more chains mean more bridges, more liquidity silos, and more user UX complexity. Cross-chain swaps and asset transfers remain a friction point despite intent-based interop progress. Second, security inheritance: a rollup using a third-party DA layer inherits that DA layer's security; a chain bug in EigenDA or Celestia could affect dozens of rollups simultaneously. Ethereum-blob-based DA is safer in this respect but more expensive. Third, sustained economic activity: hundreds of L2 and L3 deployments exist, but only a small fraction (Base, Arbitrum One, Optimism, Linea, Polygon zkEVM, Starknet) have meaningful sustained activity. Many rollup chains exist as branding exercises rather than economic primitives. The monolithic counter-argument, pressed most prominently by Solana and the Sui/Aptos Move-VM chains, is that hardware-scaled monolithic chains achieve simpler UX, deeper liquidity, and competitive throughput without modular complexity. Solana's 2024-2025 growth (Firedancer client, Token Extensions, network reliability improvements) demonstrated that monolithic scaling is viable, and Solana's TPS and active-user metrics rivaled the entire Ethereum modular ecosystem in some periods. Looking forward to 2027, the realistic trajectory is convergence rather than victory: Ethereum modular ecosystems will continue to dominate institutional and DeFi-heavy use cases; Solana and similar monolithic chains will continue to dominate consumer, payment, and high-throughput retail use cases; the rollup-framework consolidation will reduce the number of practically active L2/L3 chains from hundreds toward dozens; and intent-based interoperability will reduce the user-facing fragmentation. The architectural philosophy debate will fade as both approaches find their natural niches.
Watch points
- Celestia, EigenDA, Avail, and Ethereum blob throughput shares of total rollup DA
- Rollup-framework deployment count vs sustained-activity ratio across OP Stack, Arbitrum Orbit, Polygon CDK, ZK Stack
- Cross-chain interop volume via Across, LayerZero, Wormhole, Hyperlane
- Solana and other monolithic-chain growth versus aggregate Ethereum-modular activity
TL;DR
Modular blockchain architecture (DA via Celestia/EigenDA/Avail/Ethereum-blobs, execution via OP Stack/Arbitrum Orbit/Polygon CDK/ZK Stack, settlement on Ethereum L1) became the dominant Ethereum-ecosystem pattern by 2026 with $50B+ aggregate L2 TVL, though monolithic chains (Solana) demonstrated competitive performance with simpler UX through 2027.
Get DeFi Intel research in your inbox
Weekly long-form coverage of papers, incidents, jurisdictions, chains, tokens and the people building them. Free tier covers headlines; Pro adds the analyst-grade breakdowns.