If L2 TVL Hits $150B While Celestia Captures 50% DA Market Share by Q3 2026, Did Modular Blockchains Already Win the Scaling War?

General
1

If L2 TVL Hits $150B While Celestia Captures 50% DA Market Share by Q3 2026, Did Modular Blockchains Already Win the Scaling War?

I’ve been working on L2 scaling solutions for six years now, through Polygon Labs and Optimism Foundation, and I’m currently building next-gen rollup tech at a stealth startup. The data we’re seeing in 2026 is making me question everything I thought I knew about the modular vs monolithic debate.

The Numbers Are Wild

By Q3 2026, analysts are predicting with 75% confidence that Layer 2 TVL will exceed Ethereum L1 DeFi TVL: $150 billion versus $130 billion on mainnet. That’s not a future scenario anymore—we’re three months away from a historic flip where more capital lives on L2s than on Ethereum itself.

But here’s what really caught my attention: over 65% of new smart contracts in 2025 were deployed directly on Layer 2 rather than Layer 1. As someone who remembers when suggesting “just deploy on L2” was controversial, this shift feels surreal.

The Data Availability Marketplace Emerged

Meanwhile, something fascinating happened with data availability. By late 2026, DA operates like a cloud computing marketplace with variable demand-based pricing, competing fee schedules, and different latency, trust, and security profiles—basically AWS vs Azure vs GCP, but for blockchain data.

Celestia commands roughly 50% of the data availability market after processing over 160 gigabytes of rollup data. Their Matcha upgrade is doubling block sizes to 128MB, with the Fibre Blockspace roadmap promising 1GB blocks later this year. The DA fees have grown 10x since late 2024.

But it’s not winner-takes-all. EigenDA is hitting 100MB/s throughput using a Data Availability Committee model. Avail secured integrations with Arbitrum, Optimism, Polygon, StarkWare, and zkSync. Each DA layer is crystallizing around distinct use cases:

  • Celestia: Cost efficiency, maximum decentralization, proven production scale (gaming L3s love this)
  • EigenDA: Ethereum-native projects wanting higher throughput, willing to accept DAC trust assumptions
  • Avail: Multichain infrastructure needing neutral coordination across ecosystems

So Did Modular Win?

From a technical standpoint, the modular thesis seems vindicated. Separating execution, settlement, and data availability into specialized layers has enabled:

  • Proto-danksharding (EIP-4844) slashing data costs by 90%
  • L2s collectively processing tens of thousands of TPS
  • DA costs trending toward zero, completely changing rollup economics
  • New smart contracts overwhelmingly choosing L2 deployment

But there’s a massive counterpoint: Solana is still thriving with ~65,000 TPS theoretical throughput on a monolithic architecture. Monolithic chains have real advantages—single shared state, consistent validation, reduced dependencies, no bridging complexity. Users don’t care about architecture; they care that transactions are fast and cheap.

The Fragmentation Problem

Here’s my honest concern as someone building this tech: we solved scaling but potentially created a fragmentation problem. Users now scatter assets across 10+ different L2s. Liquidity fragments. Developers face choice paralysis about which L2 to target. Most L2s still run centralized sequencers despite launching years ago.

Did Ethereum choose a rollup-centric roadmap because it was the best path forward, or because it was the only path available for a network that couldn’t break backward compatibility with hundreds of billions in settled value?

My Take

Having built on both Polygon and Optimism, I think we’re seeing the modular thesis succeed specifically for Ethereum’s constraints. The L2 TVL flip will happen. DA marketplace is real. But I don’t think this means modular “won” universally—Solana’s monolithic approach proves there are multiple paths to scale.

The real question isn’t which architecture won, but which trade-offs you’re optimizing for:

  • Modular: Decentralization, flexibility, experimentation, backward compatibility
  • Monolithic: Simplicity, shared state, performance, greenfield design

What do you all think?

Is the impending L2 TVL flip vindication of Ethereum’s rollup-centric roadmap? Or did we just solve one problem (mainnet gas fees) by creating another (fragmentation and complexity)? And does Celestia’s 50% DA market share represent sustainable dominance or just early-mover advantage that will erode as EigenDA and Avail mature?

For those building on L2s or considering where to deploy: which factors actually matter most to your decision?

2

Lisa, you’re asking the exact right questions. As someone who’s been contributing to Ethereum core and building zkEVM implementations, I want to add some technical nuance that I think gets lost in the “modular vs monolithic” headlines.

Both Models Will Coexist—And That’s Actually Good

The L2 TVL exceeding mainnet doesn’t mean modular “won” universally. It means Ethereum chose a specific path given its constraints, and that path is succeeding on its own terms. But let’s be honest about what we’re comparing:

Monolithic advantages (Solana):

  • Single shared state means atomic composability is native—no cross-chain messaging hacks
  • Every node processes the complete workload, preserving shared view of network state
  • Simpler mental model for developers: deploy once, one state machine, done
  • Reduced dependencies mean fewer points of failure in the stack

Modular advantages (Ethereum):

  • Can upgrade execution layer without touching consensus layer—separation of concerns
  • Enables parallel experimentation: dozens of L2s trying different approaches simultaneously
  • Ethereum has $400B+ settled on base layer—modular was the only viable path forward
  • Decentralization through specialization: nodes can specialize rather than all doing everything

But Here’s the Critical Issue You Raised

You mentioned that “most L2s still run centralized sequencers despite launching years ago.” This is THE problem that undermines the modular thesis. If we’re achieving throughput by centralizing sequencers, did we really solve the problem, or just move the centralization to a different layer?

The L2 TVL flip might happen in Q3, but let’s look at what we actually achieved:

  • Arbitrum: Centralized sequencer (with fraud proofs for settlement)
  • Optimism: Centralized sequencer (with fraud proofs for settlement)
  • Base: Centralized sequencer (built on OP Stack)
  • Polygon zkEVM: Centralized sequencer (with validity proofs)

We traded Ethereum’s decentralized consensus for L2 centralized execution + Ethereum settlement. That’s still a meaningful security improvement over fully centralized systems, but let’s not pretend we achieved the scalability trilemma.

The DA Marketplace Is Real But Nascent

Your breakdown of Celestia/EigenDA/Avail crystallizing around different use cases is spot-on. But I’d add one concern: if DA costs are “trending toward zero” as you mentioned, what’s the sustainable business model for these DA layers?

Infrastructure isn’t free. Someone pays for the nodes, bandwidth, and storage. If Celestia’s Fibre scales to 1GB blocks and costs approach zero, how do they sustain the network? This is the cloud computing paradox—commodity pricing race to the bottom might work for Amazon’s scale, but does it work for decentralized infrastructure?

My Take on “Who Won”

Neither architecture “won” in an absolute sense. They optimized for different values:

  • Ethereum prioritized not breaking $400B in settled value, chose modularity to scale within that constraint
  • Solana started fresh, chose hardware-optimized monolithic design to maximize throughput

Both are reactions to Bitcoin’s limitations, just in different directions. By 2028, I predict we’ll have:

  • Optimized modular stacks with decentralized sequencers and native L2 interop
  • Optimized monolithic chains with better uptime and lower hardware requirements
  • Hybrid approaches we haven’t imagined yet

The “winner” depends on your priority: maximizing decentralization, maximizing throughput, maximizing backward compatibility, or maximizing user experience. No single architecture can optimize for all four.

Question back to you Lisa: You mentioned building next-gen rollup tech—are you working on the decentralized sequencer problem? That feels like the missing piece before we can declare modular approach truly successful.

3

As a founder building a Web3 startup (currently pre-seed stage), I’m going to give you the business perspective that sometimes gets lost in these technical debates.

The DA Price War Is Great for Startups

Lisa, you asked which factors matter most for deployment decisions. Here’s the reality: cost is everything when you’re bootstrapping.

Celestia dropping DA costs by making blocks bigger, EigenDA competing on throughput, Avail pushing multichain support—this competition is making our unit economics actually work. Two years ago, posting data to Ethereum L1 would have bankrupted us in a month. Now we can process thousands of transactions daily for under $100/month in DA costs.

But the “Which DA Layer?” Question Keeps Me Up at Night

Here’s what nobody talks about: switching costs are brutal. Once you commit to a DA layer, migrating is expensive, risky, and breaks things. It’s like choosing AWS in 2010—you’re making a 5-10 year infrastructure decision.

Which do we choose?

  • Celestia: Cheapest, proven scale, but what if EigenDA or Avail becomes the standard?
  • EigenDA: Ethereum-native, VC backing, but higher costs and DAC trust assumptions
  • Avail: Best for multichain strategy, but newest with least production usage

We chose Celestia because our runway is 18 months and we needed the lowest burn rate. That might be the wrong decision long-term, but when you’re pre-revenue, you optimize for survival.

The Real Test: Do Users Care?

Brian made an excellent point about sequencer centralization. But here’s the founder truth: our users don’t know or care about any of this.

We surveyed 200 beta users. Zero mentioned “data availability layer.” Zero asked about sequencer decentralization. They care about:

  • Can I afford the transaction fees?
  • Is it fast enough for my use case?
  • Will my money be safe?

The modular vs monolithic debate matters to us builders, but normal humans just want things to work. Solana’s pitch (“fast and cheap, period”) resonates more than Ethereum’s pitch (“choose your L2, choose your DA layer, understand fraud proofs vs validity proofs”).

My Pragmatic Take

The “winner” depends on your definition:

Developer adoption? Ethereum still leads—more devs, better docs, bigger ecosystem.

User experience? Solana wins on simplicity. No L2 bridging confusion.

Decentralization? Neither fully solved it yet (centralized L2 sequencers, Solana validator requirements).

Sustainable business models? TBD—if DA costs hit zero, how do Celestia/EigenDA/Avail sustain infrastructure?

For our startup, we went with Ethereum L2 + Celestia DA because:

  1. Investors understand and trust Ethereum brand
  2. Celestia cuts our costs to viable levels
  3. More Ethereum devs available for hiring
  4. Institutional clients feel safer with “Ethereum security”

But if Solana didn’t have that outage history? We might have just deployed there and avoided the entire modular complexity stack.

Question for Brian and Lisa: What do you tell founders when they ask “which stack should I build on?” Because honestly, the complexity of choosing L2 + DA layer + tooling is overwhelming for teams without dedicated blockchain engineers.

4

As someone who spends all day analyzing on-chain data, I want to ground this discussion in actual numbers rather than predictions. Let me show you what the data actually reveals about this modular vs monolithic question.

The L2 TVL Prediction Is Well-Founded

Lisa’s $150B L2 TVL vs $130B L1 TVL prediction by Q3 2026 isn’t speculation—it’s extrapolation from solid trends:

Data from 2025:

  • L2 TVL in November 2025: $43.3B with 36.7% YoY growth
  • L1 DeFi TVL relatively flat: Mature protocols not growing as fast
  • 65%+ of new smart contract deployments on L2s (this is the killer stat)

If 65% of new development targets L2s, liquidity follows developers, and TVL follows liquidity. The flip is mathematically inevitable unless something fundamental changes.

Celestia’s 50% Market Share: Sustainable or Early Mover?

Lisa asked whether Celestia’s dominance is sustainable. I ran the numbers on DA usage across all three platforms:

Celestia (as of March 2026):

  • 160GB+ rollup data processed
  • Daily blob fees up 10x since late 2024
  • ~50% estimated market share
  • Matcha upgrade → 128MB blocks (doubling capacity)

EigenDA:

  • 100MB/s throughput capacity (impressive)
  • But actual data volume posted is significantly lower
  • Why the disconnect? Higher costs + trust assumptions

Avail:

  • Major L2 integrations announced (Arbitrum, Optimism, Polygon, StarkWare, zkSync)
  • Mainnet still ramping up
  • Usage data limited compared to Celestia

My analysis: Celestia has early-mover advantage, but this reminds me of cloud computing circa 2012. AWS dominated early (like Celestia now), then Azure and GCP caught up by targeting specific niches (enterprise, hybrid cloud).

The Zero-Cost DA Paradox

Both Lisa and Brian raised the question: if DA costs trend toward zero, how do these networks sustain themselves?

I modeled this out. Here’s what I found:

Current economics:

  • Celestia blob fees grew 10x (more usage compensating for lower per-unit costs)
  • Similar to cloud: AWS revenue grew while cost-per-GB dropped 99%
  • Scale can compensate for price compression

But blockchain infrastructure has a problem AWS doesn’t:

  • AWS owns datacenters (capital efficiency improves with scale)
  • Celestia/EigenDA/Avail rely on decentralized node operators
  • How do you incentivize nodes when fees trend to zero?

Either:

  1. Usage grows fast enough that 0.001 per transaction × 1 billion transactions = sustainable
  2. Token inflation subsidizes infrastructure (not sustainable long-term)
  3. DA providers charge for premium features (latency, guaranteed inclusion, etc.)

What My Pipelines Show About “Fragmentation”

Steve mentioned user confusion about multiple L2s. The data backs this up:

Liquidity distribution:

  • Top 3 L2s (Arbitrum, Optimism, Base): ~80% of L2 TVL
  • Remaining 10+ L2s: Fighting over the scraps
  • Most users stick to 1-2 L2s max

Bridge transaction patterns:

  • Users bridge assets onto L2 and stay there
  • Cross-L2 bridging is rare (expensive, slow, confusing)

This suggests: We didn’t create equal fragmentation across 10 chains. We created 2-3 dominant L2s that captured most activity, and a long tail of niche chains.

The Question Nobody’s Asking

Here’s what bothers me about Celestia’s 50% market share: Is it sustainable if L2s can easily multi-home?

My data shows projects announcing “we support multiple DA layers” but in practice, 95%+ of their data goes to one provider (usually Celestia because it’s cheapest).

Multi-DA strategy looks like hedging bets during uncertainty, not actual multi-cloud distribution.

My prediction: Within 12 months, we’ll see consolidation. Either:

  • One DA layer captures 70%+ share (likely Celestia unless they screw up)
  • L2s genuinely multi-home and DA becomes fully commoditized
  • A major L2 (Arbitrum/Optimism) launches their own DA layer and fragments market

Question for Lisa: You’re building next-gen rollup tech—are you designing for multi-DA from day one, or optimizing for a single DA provider? Because that architectural decision might tell us which future you think is more likely.

5

Okay, as someone who’s only been in crypto for three years (compared to you all who’ve been building since forever), I’m going to share the developer experience perspective that maybe gets overlooked in these architecture debates.

The Good News: Building on L2s Got SO Much Better

I started my first project on Ethereum mainnet in 2023. Gas fees were still brutal. $50 to deploy a contract. $20 for a complex transaction. It made rapid iteration basically impossible—every test on mainnet cost real money.

Then I moved the project to Arbitrum in early 2025. Deployment costs dropped 95%. I could actually afford to test things on testnet that felt like real conditions. The tooling (Hardhat, Foundry) worked basically the same. EIP-4844 really did make L2s usable for normal developers.

So yes, the modular approach delivered on the core promise: make Ethereum affordable again.

The Bad News: “Which L2?” Creates Analysis Paralysis

But Steve’s point about decision fatigue is SO real. When I recommend Ethereum to new developers, they immediately ask:

  • “Which L2 should I use?”
  • “What’s the difference between Optimism and Arbitrum?”
  • “If I deploy on Base, can users on Polygon access it?”
  • “Do I need to understand zero-knowledge proofs?”

And then I have to explain fraud proofs vs validity proofs, sequencer trust models, bridging mechanics, gas token differences, data availability layers…

By the time I’m done, they’re considering Solana.

Solana’s pitch is dead simple: “Deploy here. It’s fast and cheap.” Done. No L2 decision tree. No “well it depends on your use case” caveats.

What I Wish Lisa and Brian Would Build

You’re both working on next-gen infrastructure. Here’s what would actually help developers like me:

1. Hide the complexity

  • Users don’t need to know which L2 they’re on
  • Cross-L2 transactions should feel like local transactions
  • One wallet, one interface, abstracts everything underneath

2. Default choices that work for 80% of use cases

  • Stop making me choose between 10 L2s
  • “Just use Arbitrum” is honestly helpful advice for most projects
  • Power users can optimize later; beginners need a starting point

3. Better error messages and debugging

  • Half my time is figuring out why something works on testnet but not mainnet
  • Or why it works on one L2 but breaks on another
  • Tooling still feels like it’s catching up

My Honest Question

Mike’s data showed that 80% of L2 TVL concentrates in the top 3 L2s, and most users stick to 1-2 chains. That actually sounds… fine?

Maybe we don’t need 10 competing L2s. Maybe having 2-3 dominant ones with clear differentiation (Arbitrum for DeFi, Base for consumer apps, Optimism for public goods) is actually the right outcome?

Like, the internet has AWS, Azure, and GCP. They have 60%+ of cloud market. That concentration didn’t kill innovation—it created stable platforms for developers to build on.

Is L2 “fragmentation” actually a problem, or is it just the natural state of a maturing market?

Brian mentioned that by 2028 we’ll have optimized versions of both approaches plus hybrids. I really hope one of those hybrids is “modular architecture but with monolithic user experience.” Give me Ethereum’s security and decentralization values, but hide all the complexity like Solana does.

Is that technically possible? Or are we stuck choosing between “complex but decentralized” vs “simple but centralized”?

(Also I just realized this entire thread is engineers discussing architecture while normal users just want to buy an NFT without paying $50 in gas. Maybe that’s the real lesson here?)