The market still talks about Layer-2 scaling as if execution speed is the bottleneck. That was true in 2020; it is mostly wrong now. The real constraint in crypto infrastructure is not how fast a sequencer can run an EVM fork on cloud hardware, but whether the transaction data needed to verify that execution is cheaply and credibly available to everyone else.
This is the core of the modular blockchain thesis: separate execution from data availability, let specialized networks do each job, and stop pretending one chain can optimize for compute, bandwidth, settlement, decentralization, and liquidity at the same time. It is a compelling thesis. It is also dangerously oversold when investors confuse lower fees with stronger security.
With ETH trading around $1,795 and rollups now responsible for a large share of Ethereum user activity, the practical question is no longer whether modular architecture works. It does. The harder question is who captures value when execution becomes cheap and data availability becomes a commodity market with very non-commodity security assumptions.
What is the modular blockchain thesis?
The modular blockchain thesis says blockchains should split core functions — execution, settlement, consensus, and data availability — instead of forcing one network to do everything. In practice, rollups execute transactions off-chain, post compressed data to a data availability layer, and rely on a settlement layer such as Ethereum to finalize disputes or verify proofs.
A monolithic chain like Solana or Ethereum pre-rollup attempts to bundle all functions into one validator network. That design is elegant but expensive: every full node must download, execute, and store the same data. Modular architecture removes that duplication by letting execution environments specialize while a separate base layer guarantees the data needed to verify state transitions is available.
The cleanest example is an Ethereum rollup. Arbitrum, Optimism, Base, Starknet, and zkSync run execution environments that process transactions off Ethereum mainnet. They then publish transaction data or state diffs to Ethereum, where validators do not re-execute every transaction but make the data available and enforce settlement rules through fraud proofs or validity proofs.
This distinction matters because execution is not scarce in the same way as decentralized bandwidth. A centralized sequencer can process thousands of transactions per second on ordinary server infrastructure. The expensive part is making the input data available to a global network without trusting the sequencer. That is where Ethereum blobs, Celestia, EigenDA, Avail, and NEAR DA are now competing.
How does data availability work in modular blockchains?
Data availability means the network can verify that the data required to reconstruct blockchain state was published and can be downloaded by honest participants. It is not long-term storage, and it is not the same as transaction execution; it is a guarantee that data was made available at the time security depended on it.
Rollups need DA because users must be able to independently reconstruct balances, challenge invalid state transitions, or exit the system. In an optimistic rollup, fraud proofs are meaningless if challengers cannot access transaction data. In a ZK rollup, the proof may validate a state transition, but users still need data to know their own balances and preserve censorship-resistant exits.
Ethereum historically handled this through calldata, which is permanent and expensive. EIP-4844, also known as proto-danksharding, introduced blob space: a separate fee market for temporary rollup data. Each blob is roughly 125 KiB, Ethereum targets 3 blobs per block and allows up to 6, creating about 375 KiB of target DA capacity per 12-second block, or roughly 2.7 MB per minute at target utilization.
The impact was immediate. After the March 2024 Dencun upgrade, many rollup data posting costs fell by more than 90% during normal conditions. That was not magic throughput; it was market design. Ethereum stopped forcing rollups to compete with DeFi swaps and NFT mints for the same gas resource and created a dedicated market for ephemeral data.
Celestia takes a different approach. It is a purpose-built DA network using data availability sampling, where light nodes probabilistically sample small pieces of erasure-coded block data rather than downloading the full block. The key claim is that as more light nodes participate, the system can support larger blocks without requiring every user to run a data-center-grade node.
EigenDA adds another model by using Ethereum restaking through EigenLayer. Operators commit to serving data and can be penalized under EigenLayer’s cryptoeconomic framework. This creates a flexible DA service anchored around Ethereum capital, but it also imports new correlated risk: if restaked ETH secures many services simultaneously, one failure mode can propagate through the stack.
Why does separating execution from data availability matter for traders?
Separating execution from data availability changes where fees, value capture, and failure risk accumulate. Traders should stop valuing every Layer-2 token as an execution bet and start asking who controls sequencing revenue, DA costs, bridge security, and settlement guarantees.
Execution fees are structurally compressing. Base, for example, has demonstrated that a well-run EVM rollup can onboard mass retail activity with extremely low transaction fees, especially after blobs. But cheap execution is not the same as durable token value. If users pay pennies and the rollup token has no claim on sequencing, MEV, or governance over core infrastructure, the investment case becomes political rather than economic.
The real margin sits in three places: sequencers, DA providers, and settlement layers. Sequencers order transactions and can capture priority fees and MEV. DA providers sell scarce decentralized bandwidth. Settlement layers like Ethereum monetize finality, liquidity, and trust minimization. Most application-specific rollups will struggle to capture value unless they own distribution or unique order flow.
This is why the modular trade is not simply bullish for every rollup token. It may be bullish for ETH as settlement collateral, selectively bullish for DA networks if demand exceeds supply, and bullish for infrastructure operators that can aggregate sequencing. It is less obviously bullish for generic governance tokens attached to execution environments with no fee rights.
The market’s lazy framework is to compare transactions per second. That metric is borderline useless. A chain can advertise high TPS by weakening verification, increasing hardware requirements, or hiding data assumptions behind committees. The better metric is verified throughput per unit of decentralization: how much activity can the system support while preserving cheap independent verification?
What happens if data availability becomes a commodity?
If DA becomes a commodity, rollup costs fall but security differentiation becomes harder to price. The danger is that applications will chase the cheapest DA layer until a failure exposes that not all data availability guarantees are equivalent.
Ethereum blob space is expensive relative to centralized alternatives, but it has the strongest settlement integration for Ethereum rollups. Posting data to Ethereum means the same validator set that finalizes ETH also underwrites the rollup’s DA. For high-value DeFi, that integration matters. A perpetuals exchange with billions in open interest should not optimize DA the same way a gaming rollup optimizes DA.
Celestia, Avail, and EigenDA create a different design space. They allow rollups to reduce costs and launch without competing for Ethereum blob space. That is powerful for appchains, games, social networks, and high-volume consumer applications. But bridging back to Ethereum introduces additional trust assumptions: light-client verification, committee assumptions, or middleware dependencies must be understood, not waved away with the word modular.
There is also a latency tradeoff. Traders care about finality, not just fees. A rollup that executes quickly but settles slowly may feel instant in the interface while carrying hidden withdrawal and bridge risk. In stressed markets, when oracle updates, liquidations, and bridge exits matter most, the slowest or weakest component defines the risk profile.
My contrarian view: DA will not be a winner-take-all market. It will segment by risk tier. Ethereum blobs will dominate high-value financial settlement. Celestia and Avail will win cost-sensitive sovereign rollups and appchains. EigenDA will attract teams already aligned with Ethereum restaking and willing to accept restaking-specific complexity. The end state looks less like one global computer and more like cloud regions with different security service-level agreements.
Why the modular thesis is stronger than monolithic maximalism
Monolithic chain advocates argue that shared execution preserves composability and reduces bridge risk. They are right about the problem and wrong about the solution. Global synchronous composability is valuable, but it does not scale indefinitely when every validator must process every transaction in the same state machine.
Solana proves that integrated design can deliver excellent user experience when hardware, networking, and client engineering are pushed aggressively. But that model makes a specific trade: higher node requirements for lower latency and unified liquidity. Ethereum’s modular roadmap makes the opposite trade: preserve low verification costs and push execution outward to rollups.
The question is not which architecture is morally superior. The question is which failure mode the market prefers. Monolithic chains risk centralizing validation as throughput rises. Modular chains risk fragmentation, bridge complexity, and opaque trust assumptions. Anyone claiming one side has solved scaling is selling a narrative, not infrastructure analysis.
What makes modular architecture durable is that it mirrors how serious distributed systems evolve. Databases separated compute from storage. Cloud platforms separated application logic from networking, identity, and object storage. Blockchains are now separating execution from DA because the bottlenecks have different optimization functions. Compute wants speed; DA wants verifiability; settlement wants credible neutrality.
The weakest part of today’s modular stack is not the theory. It is the sequencing layer. Most major rollups still rely on centralized sequencers, often operated by the core team or a related entity. That creates censorship risk, liveness risk, and a privileged position in transaction ordering. Decentralized sequencing, shared sequencing, and based rollups are the next serious battleground.
What should builders and investors watch next?
Builders should watch blob fee volatility, DA bridge design, and proof system maturity. Investors should watch whether protocol tokens accrue revenue from sequencing, DA fees, or settlement demand rather than assuming user growth automatically becomes token value.
The first indicator is blob saturation. Ethereum’s current blob capacity is intentionally conservative. If rollup demand consistently fills the target of 3 blobs per block and pushes toward the maximum of 6, blob fees can rise sharply because the market is separate but still scarce. That would benefit Ethereum fee burn and DA alternatives, while pressuring low-margin rollups.
The second indicator is DA migration. If major rollups begin using non-Ethereum DA for meaningful financial activity, the market will need to reprice bridge and settlement assumptions. A rollup settling on Ethereum but using external DA is not equivalent to one posting blobs directly to Ethereum. The difference may be invisible to users until something breaks.
The third indicator is proof cost. ZK systems are improving quickly, with projects such as Polygon, Scroll, Starknet, and zkSync pushing prover performance and EVM compatibility. But proof generation remains operationally complex, and many systems still rely on upgrade keys, permissioned provers, or immature decentralization roadmaps. ZK does not eliminate trust; it moves trust into circuits, keys, provers, and data assumptions.
The fourth indicator is sequencing decentralization. The rollup that controls order flow controls the most valuable part of the stack. If shared sequencers such as Espresso or based rollup designs gain traction, value may shift away from individual rollup operators toward neutral ordering networks or Ethereum validators. That would be positive for neutrality and uncomfortable for rollup token economics.
The modular thesis is not that every blockchain component becomes cheaper. It is that each component gets priced separately, and the market finally sees which guarantees were being subsidized by architectural bundling.
Key Takeaway
The modular blockchain thesis is fundamentally correct: execution should be separated from data availability because compute and verifiable bandwidth are different markets. But the winning stack will not be the cheapest one; it will be the one that makes security assumptions legible under stress.
For builders, DA choice is now a product and risk decision, not a back-end detail. For investors, the smart question is not which chain has the highest TPS, but which layer captures fees when execution becomes abundant and trustworthy data availability remains scarce.