Modular Blockchains: How Execution Splits From Settlement

Then scalability problems became impossible to ignore. When CryptoKitties congested Ethereum in 2017, it revealed a fundamental truth: asking one system to do everything well means doing nothing exceptionally. The modular blockchain thesis emerged from this realization, arguing that separation of concerns—splitting execution from settlement, consensus from data availability—could unlock performance that monolithic chains simply couldn’t achieve.

The concept has moved from academic papers to production systems. Celestia launched its mainnet in late 2023 as the first dedicated data availability layer. EigenLayer launched restaking in 2024, creating a new security model for modular infrastructure. Ethereum’s danksharding roadmap is explicitly modular. Understanding how execution splits from settlement isn’t just academic anymore—it’s becoming the foundation for how most new blockchain development happens.

When Satoshi Nakamoto designed Bitcoin, the chain had one job: maintain a shared ledger of transactions. Every node validated every transaction. Every node stored every piece of data. This was security through redundancy—expensive, but simple.

Ethereum expanded this model to include smart contracts, adding a virtual machine that every node executed. The design philosophy remained the same: security through universal participation. Every validator ran every contract. Every full node stored every state update.

This approach has genuine advantages. When execution and settlement live together, the settlement guarantee is absolute—the same validators that execute your transaction are the ones finalizing it. There’s no need to trust a separate settlement layer. The math is clean.

The problem is scalability. If every node must execute every transaction, you’re limited by the slowest node in the network. Ethereum processes roughly 15-30 transactions per second under normal conditions. Solana claims much higher throughput, but achieves it through hardware-intensive requirements that reduce the number of actual validators. The fundamental bottleneck remains: you can’t have both universal execution and massive scale on the same chain.

The modular argument isn’t that monolithic chains are broken—it’s that they’re specialized for one use case (maximum security through universal validation) while other use cases (throughput, speed, customization) require different architectures.

The modular approach separates blockchain functions into distinct layers. Consensus determines the ordering of transactions and agreement on the current state—what validators actually do when they produce blocks. Execution handles transaction processing—the actual computation that updates state, where smart contracts run. Settlement provides finality and dispute resolution; if someone claims a transaction happened, settlement confirms it irreversibly. Data availability ensures that transaction data is published and accessible. Without DA, you can’t verify the chain’s history.

In a modular design, different chains or services handle these functions independently. An execution chain might rely on a separate DA layer for data storage and a different settlement layer for finality. Each component specializes.

This isn’t theoretical. It’s how most new blockchain projects are being built in 2024 and 2025.

The execution layer is where transactions actually happen. In a modular architecture, execution chains—sometimes called rollups, appchains, or sovereign chains—focus purely on processing transactions as fast as possible.

Ethereum popularized this with rollups. Optimistic rollups like Arbitrum and Base execute transactions off-chain, then post compressed transaction data back to Ethereum for settlement. ZK rollups like zkSync and Starknet do the same but use cryptographic proofs instead of fraud proofs. Either way, execution happens separately from Ethereum’s base layer.

The execution chain makes a tradeoff: it gives up some degree of sovereignty in exchange for dramatically better performance. An Optimistic rollup can process thousands of transactions per second while still inheriting Ethereum’s security.

But there’s a subtler point that gets overlooked in most modular blockchain coverage: execution chains don’t just split from settlement for performance. They split because different applications need different execution environments.

A high-frequency trading application needs sub-second finality and minimal latency. A gaming platform needs cheap transactions with eventual finality. A financial settlement system needs absolute certainty. No single execution environment serves all these use cases well. Modular architecture lets each use case pick its own execution parameters.

Settlement is where modular architecture gets interesting—and controversial. When execution happens on a separate chain, something has to guarantee that the execution was valid.

The simplest model is rollup-style settlement: post transaction data to a more secure base layer, and let that layer resolve disputes. Ethereum serves that function for most rollups today. If there’s a dispute about an Arbitrum transaction, Ethereum’s contract resolves it.

This works, but it creates a dependency. The execution chain’s security is limited by the settlement layer’s security. If Ethereum gets compromised, every rollup built on it is compromised.

EigenLayer introduced an alternative in 2024: restaking. ETH holders can stake their tokens again to secure additional protocols, including modular execution chains and data availability layers. This creates a market for security—execution chains can rent security from Ethereum’s massive validator set rather than building their own.

The EigenLayer approach is genuinely innovative, but it carries real tradeoffs. Restaking introduces additional smart contract risk on top of existing staking risk. It also creates economic dependencies that could amplify systemic failures. If a major restaked protocol gets exploited, the cascading effects could impact Ethereum itself.

I don’t think the industry has fully grappled with these risks yet. The restaking narrative is overwhelmingly positive in most coverage, but this is exactly the kind of area where confident claims should give way to honest uncertainty.

Most discussion of modular blockchains focuses on execution and settlement. Data availability deserves more attention than it gets.

Here’s the problem: if an execution chain posts only transaction results to the settlement layer, how do you verify those results? You need the underlying transaction data.

In a monolithic chain, every node stores everything. Verification is local—you just run the transaction yourself.

In a modular chain, data availability becomes a separate concern. The settlement layer needs to guarantee that transaction data exists and will remain available, otherwise the settlement guarantee is meaningless. If you can’t retrieve the data, you can’t verify the settlement.

Celestia addresses this with dedicated DA. It provides a separate network specifically for data availability sampling—clients can verify that data is available without downloading everything. This is technical but important: DA sampling lets light clients confirm availability probabilistically, making verification accessible even as data volumes grow.

Ethereum’s danksharding roadmap is moving toward integrated DA. The EIP-4844 upgrade in early 2024 introduced “blobs”—a more efficient way to store rollup data on Ethereum. Full danksharding will dramatically increase Ethereum’s DA capacity.

The practical implication: DA is becoming infrastructure. Most execution chains won’t run their own DA layers—they’ll rent capacity from specialized providers. This is another example of the modular pattern: specialized components that others build on top of.

Not every modular chain gives up sovereignty. Celestia’s model allows what it calls “sovereign rollups”—chains that use Celestia for DA but handle their own settlement and fork selection independently.

This is closer to the original Bitcoin model: each chain is its own authority. The tradeoff is that sovereign chains don’t inherit security from a larger network. They need their own validator sets.

The sovereign approach makes sense for projects that prioritize independence over composability. If you want absolute control over your chain’s rules—even if it means starting from scratch—sovereignty has value.

But most projects probably don’t need full sovereignty. What they need is flexible security: the ability to pick how much security to rent, from whom, and under what conditions. That’s where the modular market is heading.

Here’s where my confidence drops. Modular architecture creates specialized chains, but specialized chains need to communicate.

If execution happens on Chain A, settlement on Chain B, and data availability on Chain C, how does a user move assets between them? How do cross-chain applications work?

The industry has messaging protocols—Axelar, Wormhole, Hyperlane—but cross-chain communication remains fragile. The 2022 Wormhole hack (320k ETH stolen) and the 2023 Multichain collapse showed that bridges are the vulnerable point in modular systems.

Every security model has a weakest link. In monolithic chains, that link is the consensus mechanism. In modular chains, it’s the bridges connecting components.

Some teams are solving this with zero-knowledge proofs—proving that a transaction happened on Chain A without requiring a trusted bridge. But this technology is still early. I can’t point to a production system that has fully solved cross-chain composability in a way that I’d call mature.

This is the honest limitation: modular blockchains work well within their layers, but the spaces between layers are still problem areas.

The modular versus monolithic debate used to be theoretical. It isn’t anymore.

Solana processed an average of 65 million transactions per day at peak periods in 2024. Ethereum’s Layer 2 ecosystem—all the modular execution chains—processed similar volumes. The math is shifting: more transactions happen on modular execution layers than on monolithic base chains.

The talent flow confirms this. The strongest engineering teams in blockchain are building modular infrastructure. The funding reflects it—Celestia, EigenLayer, and various DA projects raised hundreds of millions in 2023-2024.

The architectural question is settling: the future is modular. The remaining questions are about implementation details—how settlement works, who provides DA, how cross-chain communication evolves.

The modular blockchain thesis has proven itself technically. What remains is economic and political: who controls the infrastructure layers, how security is priced, whether users accept the complexity tradeoffs.

We’re entering a period of intense experimentation. The components exist. The connections between them are still being built.

The next two years will determine whether modular blockchains deliver on their promise—or whether the complexity costs more than anyone expected. Either way, the split between execution and settlement is permanent. It’s the defining architectural shift of this generation of blockchain development.