If you have ever tried to explain the modular blockchain critique to someone, then you already know the look. Their eyes do not glaze over immediately. They give you about twelve seconds of polite nodding before they decide the cheeseboard is more interesting. And honestly? They might be onto something.
Here is the thing about this grand architectural debate reshaping crypto in 2026: both sides are right, both sides are wrong, and everyone is pretending otherwise at very expensive conferences.
The Celestia hoodie crowd says the future is specialized layers working in harmony. The Solana crowd points to raw throughput numbers and shrugs like someone who just won a foot race against a committee. Neither group wants to acknowledge that the other has a point, because acknowledging nuance in crypto is considered a personal weakness.
So let us have a proper look at what is actually going on, minus the venture capital poetry.
What does “modular” even mean?
Think of a traditional blockchain as one of those old Swiss Army knives, the kind with 47 tools folded into it. Useful? Sure. But you would not perform surgery with the little scissors, and nobody has ever successfully opened a wine bottle with the fish scaler. Yet there it is, all crammed together, because once upon a time someone thought doing everything in one place was the pinnacle of design.
A modular blockchain looks at that Swiss Army knife and says, “What if the scissors had its own job, the knife had its own job, and the corkscrew went to corkscrew school?” Each function, whether that is executing transactions, reaching consensus, settling disputes, or making data available, gets handed to a specialized layer that does only that one thing and presumably does it very well.
Celestia, the poster child of this movement, handles consensus and data availability exclusively, leaving execution to rollups that rent its infrastructure. Ethereum, after its Fusaka upgrade was activated in December 2025, has been racing down the same philosophical corridor, using PeerDAS technology to let nodes verify blob data through sampling rather than downloading everything wholesale.
The Fusaka upgrade introduced PeerDAS, a new networking protocol that allows nodes to verify blob data availability through sampling rather than downloading entire blobs. That is a meaningful shift. It is also the kind of sentence that makes dinner party guests reach for the cheeseboard.
The genuinely impressive part
Here is where credit is due, grudgingly or otherwise. Modular systems have moved from whitepaper fantasy into something resembling actual infrastructure. Celestia has processed over 160 GB of rollup data and supports dozens of rollups.
Ethereum’s rollup ecosystem now runs Arbitrum at around 800 transactions per second, Base at roughly 1,200, and Optimism at about 600. Those are not theoretical numbers from a pitch deck. Those are real transactions from real users who probably did not stop to appreciate the elegant architecture underneath.
Celestia’s V8 Hibiscus upgrade, now live on the Mocha testnet, introduces single-signature cross-chain transfers and ZK-verified messaging, letting networks built on Celestia securely communicate and verify state with each other using zero-knowledge proofs. That is the modular stack starting to solve its own biggest problem: fragmentation. Whether it solves it fast enough is a different argument entirely.
Fusaka feels less like an upgrade and more like a line in the sand. Before it, modular scaling was a roadmap. After it, it becomes the operating system Ethereum is actually running on. Celestia separately published Vision 2.0 in early January 2026, targeting up to 1 terabit per second of blockspace throughput, enough capacity to bring every market on-chain, from crypto exchanges to AI agent payments and real-time data markets.
Both of those announcements, arriving within weeks of each other, tell you something: the modular blockchain critique that the whole thing is theoretical is becoming harder to defend.
Now for the uncomfortable part
Modular architectures move complexity around. They do not make it disappear. This is the modular blockchain critique that supporters wave away like an annoying relative at a family gathering, but it is the one that keeps proving itself correct.
Bridge security, the most historically exploited surface in blockchain, is central to modular architectures and carries concentrated risk. Operational immaturity is real, and ecosystem fragmentation, as different layers adopt incompatible standards, makes cross-layer composition increasingly difficult. None of those problems is solved by having a cooler architecture diagram.
The Celestia situation is particularly instructive as a case study in the gap between building something and people actually using it. Eclipse accounts for over 83 GB of the 160 GB of data published to Celestia, meaning more than half of all usage comes from a single project, and the remaining rollups have collectively published just under 77 GB across roughly two and a half years of mainnet existence. One heavy user and a lot of fee collectors are not the distributed, thriving ecosystem the roadmap promised.
On the competitive pressure front, Ethereum’s Fusaka upgrade has effectively narrowed the cost gap that made Celestia attractive in the first place. Rollups that chose Celestia at least partially for its price are reconsidering, now that Ethereum-native data availability is becoming cheap enough to remove the financial incentive to build on a separate layer. When your main value proposition is cost, and your competition cuts its prices, the conversation gets awkward quickly.
Solana watches all of this and smirks
While Ethereum and Celestia have been building elaborate multi-layer architectures and explaining them to confused investors, Solana has been running a single, fast, mostly reliable chain and processing around 2,500 transactions per second in real conditions today.
The Solana camp’s argument, delivered with the confidence of someone who passed the exam while others were still debating which pencil to use, is that users do not care about architecture. They care about speed, cost, and not having to bridge assets through three different protocols to buy a JPEG.
This is not a strawman. It is a legitimate critique of modular design. The UX cost of fragmentation is real. When a user has to understand which rollup they are on, which bridge they need, and why their transaction is sitting in a fraud-proof window, the elegance of the underlying architecture offers them zero comfort.
That said, Solana’s critics point out that a monolithic chain’s hardware requirements eventually push toward centralization. Running a high-performance Solana validator demands significantly more resources than operating a light node on a modular system designed with consumer-grade participation in mind. Both sides are describing real trade-offs. Neither side is lying. Both sides are omitting inconvenient facts, which is slightly different.
What is actually happening by 2026?
The honest version of this story is a hybrid future, not a winner. The most intelligent voices on both sides have stopped treating this as a winner-take-all contest and shifted the question to what each architecture is best for. Modular configurations excel at customization and application-specific deployment. Monolithic chains excel at immediate composability and simpler mental models for users.
Venture capital investment in crypto infrastructure grew 44 percent year over year in 2025, reaching $7.9 billion, with a significant portion targeting modular solutions, data availability networks, rollup frameworks, and Layer 2 developer tooling. Money, whatever else it is, is a reasonably honest signal. The money is going into modular infrastructure.
At the same time, Ethereum’s upcoming Glamsterdam upgrade, expected in mid-2026, will bundle approximately 25 Ethereum Improvement Proposals and aims to further reduce block times and boost throughput. The modular stack keeps compounding its improvements. The criticism that it is all theory is weakening with every upgrade cycle.
Real-world asset tokenization, the sector everyone said would never move fast, has largely landed on modular infrastructure because the cost profiles and customization match what regulated institutions need. Gaming, DeFi, and AI compute markets are building on application-specific rollups at a pace that would have seemed implausible three years ago. The modular blockchain critique that nobody is actually building on it is looking less accurate by the quarter.
The verdict nobody wants to give
Here is the conclusion that makes everyone slightly unhappy, which is usually a sign it is close to correct. Modular blockchain design is genuinely useful, actually working in production, and increasingly central to how serious infrastructure gets built. The modular blockchain critique that it moves complexity without removing it is also genuinely valid, and the people raising it are not just Solana maximalists with an agenda.
The future probably looks like Ethereum and its rollups handling most of the value, Celestia and EigenDA competing on data availability for specialized use cases, monolithic chains holding the performance-and-simplicity corner of the market, and a generation of interoperability protocols desperately trying to make all of it feel seamless to the person who just wants to swap tokens without reading a whitepaper.
Modular blockchain architecture is not magic. It is engineering. Good engineering, increasingly tested engineering, but engineering that comes with trade-offs, like everything else in this industry. Anyone telling you otherwise is probably trying to sell you a token.
The cheese board, for what it is worth, is just cheese. No layers required.
