By Why Modular Blockchains Are the Future of Web3
For years, the pitch from a new Layer 1 blockchain was simple: we're faster and cheaper than Ethereum, use us instead. That framing, one chain replacing another, has quietly stopped being how the industry's best engineers think about the problem. The actual answer turned out to be architectural, not competitive: stop asking one chain to do everything, and split the job into specialized layers instead.
What "monolithic" actually broke
A monolithic blockchain handles execution, consensus, settlement, and data availability all within a single layer, every node processes every transaction, verifies every state change, and stores everything. During the 2020 to 2021 DeFi boom, that design showed its limits directly: Ethereum routinely hit triple-digit transaction fees under load, and even high-throughput alternatives like Solana suffered outright network halts during periods of extreme activity. The problem wasn't that any particular chain was badly engineered, it was that asking one integrated system to scale every function at once runs into a hard ceiling.
The four jobs a blockchain actually does
Modular design separates a blockchain's work into four distinct functions. Execution processes transactions and updates account balances, the layer users actually interact with. Settlement provides finality, the authoritative record other layers can trust. Consensus coordinates agreement among validators about what happened and in what order. Data availability makes sure the underlying transaction data stays accessible so anyone can independently verify the chain's state. A modular system lets each of these run on infrastructure built specifically for that one job, instead of forcing all four through the same bottleneck.
- Rollups handle execution, batching transactions and computing a new state.
- A base chain like Ethereum increasingly handles settlement and data availability rather than raw execution.
- Dedicated data availability networks, Celestia and EigenDA among them, exist purely to keep transaction data cheap and accessible for rollups that plug into them.
Ethereum didn't get replaced, it got repositioned
This is the part that actually killed the "Ethereum killer" narrative: Ethereum's own roadmap leaned into modularity rather than fighting it. EIP-4844, known as Proto-Danksharding, added dedicated data blobs specifically for rollups, cutting the cost of posting rollup data back to Ethereum without increasing the load on Ethereum's own execution layer. That reinforced Ethereum's role as a settlement and data-availability layer that other chains anchor their security to, rather than a high-throughput execution chain competing transaction-for-transaction with faster rivals. By 2026, most Ethereum transaction volume happens on Layer 2 rollups sitting on top of it, not on the base layer itself, a fundamentally different shape than the "one chain wins" story from a few years ago. That settlement-layer role is also exactly what's being pitched as the backbone for a very different kind of user: see why AI agents are choosing blockchain settlement over traditional payment rails.
Sovereign rollups and the more radical version
Not every project stopped at rollups anchored to Ethereum. Sovereign rollups, built with frameworks like Rollkit in the Cosmos ecosystem, handle execution and settlement themselves and only outsource data availability externally, a more independent split than a standard rollup that leans on a base chain for both settlement and security. A newer trend called chain abstraction is starting to paper over the resulting complexity for end users: a middleware layer that hides which specific rollup or execution environment a transaction actually runs on, so using a modular app doesn't require knowing or caring about the plumbing underneath it.
The private key for every Bitcoin wallet on Earth is on this website, even Satoshi's. But even if you try for a million years, you'll never find a funded one.
Try the key collider nowThe honest trade-off nobody's fully resolved
Modular design isn't a free win. Splitting a blockchain into layers introduces complexity, fragmentation across execution environments, and a chain of trust assumptions between layers that a monolithic design simply doesn't have to manage. The fair summary from engineers working in this space is that monolithic and modular represent different bets rather than a settled argument, monolithic wagers on integration and raw single-chain performance, modular wagers on specialization paying off despite the added complexity. By 2026 the modular bet had clearly become the dominant architecture for new, ambitious infrastructure, but calling the debate fully settled would be overstating it. The same specialization logic shows up outside blockchain execution too: see how DePIN networks split raw compute out from the applications that use it, a parallel bet that dedicated infrastructure beats one-size-fits-all.