85% of Ethereum Transactions Now Happen on L2s—Did We Just Make Mainnet Obsolete?

General
1

I’ve been tracking Ethereum L2 performance metrics closely, and the numbers from Q1 2026 are absolutely staggering. Nearly 85% of Ethereum-based transactions now process via Layer 2 solutions—primarily Arbitrum, Base, and Optimism. Meanwhile, Ethereum mainnet handles just 1.1-1.2 million transactions per day compared to L2s processing 1.9 million daily transactions.

The Fee Revolution Actually Happened

Thanks to EIP-4844’s blob-based data posting (launched March 2024), L2 transaction costs have dropped to $0.001–$0.05—that’s down over 90% from 2023 levels. Even L1 fees have plummeted: as of March 12, 2026, a basic ERC-20 transfer costs just $0.01–$0.02, and a Uniswap swap runs about $0.14.

But here’s what’s keeping me up at night: if users never touch L1 anymore, what exactly is Ethereum mainnet?

The Numbers Tell a Clear Story

Current state of the ecosystem:

  • L2 TVL: $9.05B locked, $40.5B in total secured assets
  • Market leadership: Base holds 33-46% of L2 TVL, Arbitrum at 31-44%
  • Projection: By Q3 2026, L2 TVL expected to hit $150B vs. $130B on mainnet
  • Transaction split: 85% L2, 15% L1

From a systems architecture perspective, this is exactly what Vitalik’s rollup-centric roadmap envisioned: Ethereum mainnet as the global settlement and data availability layer, while L2s handle day-to-day user activity.

Settlement Layer or Cannibalization?

There are two ways to interpret these numbers:

Argument 1: This Is Working As Intended
Ethereum evolved into its proper role—a secure base layer providing settlement guarantees and data availability. L2s handle execution where speed and cost matter. Clean separation of concerns. The upcoming Glamsterdam upgrade (targeting 200M gas limit and parallel execution) further cements this model by enabling more efficient proof verification.

Argument 2: We Fragmented Too Much
We have 50+ different L2 solutions, each with isolated liquidity and user bases. Users need bridges (which introduce security risks) to move between chains. Direct L1 user activity dropping to 15% means the average person never experiences “actual Ethereum”—they use an abstraction layer built by a private company.

The ZK Maturation Factor

With zkEVM implementations reaching production maturity in 2026, we’re seeing instant finality and quick withdrawals that Optimistic rollups can’t match. This creates another layer of complexity: are we consolidating around 3-5 dominant L2s, or will specialized rollups proliferate further?

What Happens to L1 Security?

Here’s my biggest concern: If L1 fee revenue drops dramatically because everyone uses L2s, can Ethereum sustain its security budget long-term? L2s pay for data availability, but is that enough to fund validator incentives at scale?

The data availability economics are interesting: post-EIP-4844, DA represents 90% of L2 operating costs. Lower DA costs mean cheaper L2 fees and more users—but does that create a race to the bottom for L1 revenue?

Looking Ahead: Glamsterdam and Beyond

The upcoming Glamsterdam upgrade promises:

  • 100M+ gas per block (some sources say 200M)
  • Parallel transaction execution
  • Native account abstraction
  • Up to 78% fee reduction
  • Enhanced proposer/builder separation

But I keep asking myself: if we’re making L1 even more efficient at a time when 85% of users are already on L2s, who is this for?

Questions for the Community

I’m genuinely curious what others think:

  1. Is the 85% L2 / 15% L1 split healthy, or did we over-correct?
  2. Should we have 3-5 dominant L2s or embrace 50+ specialized rollups?
  3. How do we solve liquidity fragmentation without compromising on decentralization?
  4. Can L1 sustain security if fee revenue continues declining?

From a pure engineering perspective, the L2 scaling roadmap delivered exactly what it promised: cheap, fast transactions with Ethereum-grade security. But did we accidentally create a two-tier system where mainnet becomes infrastructure that regular users never see?

Curious to hear from folks building on L2s, running validators, or thinking about long-term protocol sustainability. Are we celebrating a scaling victory or witnessing the gradual marginalization of L1?

2

This hits home for me in a really personal way. When I first started learning Web3 development back in 2021, L1 gas fees were absolutely brutal—I remember paying $50+ for a single Uniswap transaction during the DeFi summer craze. It made the whole space feel exclusionary and honestly kind of ridiculous. How could we talk about “financial inclusion” when a simple swap cost more than some people’s daily wages?

Fast forward to 2026, and I’m building full-time on Base. The difference is night and day. My users can interact with our DeFi protocol for fractions of a cent. We’re seeing people from countries with lower average incomes actually participating in DeFi—not just reading about it, but using it daily. That wasn’t possible when every transaction cost $10-50.

So when you ask “did we make L1 obsolete?”—from a user perspective, I genuinely don’t know if that’s a bad thing if the L2s work well. The average person doesn’t care about the philosophical purity of “true Ethereum” if they can’t afford to use it.

But here’s where I share your concern: we traded one problem for another.

Instead of high fees blocking access, we now have fragmentation. My users constantly ask: “Should I bridge to Arbitrum for this, Base for that, or Optimism for something else?” The UX is confusing. Liquidity is split across chains. And every bridge interaction introduces security risks—I still remember the billions lost to bridge hacks in 2023-2025.

I’m hopeful about the upcoming improvements—native account abstraction and chain abstraction protocols could help users interact with multiple L2s without even knowing which one they’re on. But we’re not there yet.

My honest take: The 85% L2 shift is a huge win for accessibility and affordability. But we need to solve the fragmentation UX problem before we can truly call this a success. If Ethereum L1 becomes “settlement infrastructure that users never see,” that’s fine—as long as the experience on L2s is seamless and secure.

Are we over-engineering by having 50+ L2s? Probably. I think we’ll naturally consolidate to 3-5 dominant chains over time based on where liquidity and users concentrate. Market forces will solve that part.

The security budget question you raised is fascinating though—I honestly don’t know enough about validator economics to have a strong opinion there. Would love to hear from folks running nodes about whether declining L1 fees concern them.

3

@ethereum_emma I totally agree that L2s solved the fee problem—it’s genuinely impressive from an engineering standpoint. But you hit the nail on the head with fragmentation, and I want to dig deeper into why this is more than just a UX inconvenience.

The Liquidity Fragmentation Problem

From a cross-chain infrastructure perspective, we’ve created a massive coordination problem. Right now we have:

  • 50+ different L2 solutions
  • Each with isolated liquidity pools
  • Different security models and trust assumptions
  • Incompatible smart contract deployment addresses
  • Users constantly needing bridges to move value

When you fragment liquidity like this, you don’t just create user confusion—you create systemic security risks. Bridges represent the largest attack surface in crypto. We’ve seen $2B+ lost to bridge exploits between 2022-2025. Every time users need to bridge between L2s, they’re exposing their assets to another potential failure point.

The Architecture Question Nobody’s Asking

Here’s what keeps me up at night: If Ethereum L1 is just settlement infrastructure, why are we creating 50+ different L2s instead of 3-5 dominant ones?

From a systems design perspective, each additional L2 means:

  • Another bridge security model to audit
  • More fragmented liquidity
  • Higher capital inefficiency (same liquidity spread across more pools)
  • More validator/sequencer networks to secure
  • More complexity for developers deploying multi-chain

Every chain is an island until connected. The more islands we build, the more complex the bridge network becomes, and complexity is the enemy of security.

Optimism vs. Reality

The theoretical answer is “native rollup interoperability” and “chain abstraction layers”—and yes, these are coming. Shared sequencing, cross-L2 messaging protocols, and intent-based architectures could help. But we’re building these solutions after we fragmented, not before.

It’s like building 50 different incompatible train systems and then trying to retrofit universal transfer stations. Why not design for interoperability from the start?

Should We Consolidate?

I genuinely think the answer to @layer2_lisa’s question about “3-5 dominant L2s vs. 50+ specialized rollups” will be decided by liquidity concentration and security track records.

Market forces will push users toward:

  1. Chains with the deepest liquidity (lower slippage, better UX)
  2. Chains with proven security and uptime
  3. Chains where the protocols they care about are deployed

We’re already seeing this with Base and Arbitrum dominating TVL. I expect natural consolidation to 3-5 major general-purpose L2s, plus maybe 5-10 specialized rollups for specific use cases (gaming, privacy, etc.).

The L1 Revenue Question

On the validator economics point: If L1 only earns from L2 data availability posting, and data costs keep dropping post-EIP-4844, we need to do the math on whether that sustains Ethereum’s security budget long-term.

L2s are supposed to pay for settlement and DA. But if we optimize DA costs too much (which we should for user affordability), does that create a tragedy of the commons where L1 security degrades because there’s insufficient fee revenue?

I don’t have the answer, but it’s worth modeling out different scenarios.

My Take

:bridge_at_night: Bridges are the circulatory system of Web3—but right now we have too many incompatible blood types. We need to either:

  1. Consolidate to fewer L2s with deep liquidity, OR
  2. Build bulletproof cross-L2 interoperability (and we’re not there yet)

The 85% L2 shift is a win for throughput and fees. But we’re not done—interoperability and security standardization are the next frontiers. Otherwise we risk fragmenting the ecosystem to the point where “Ethereum” becomes just a brand name for 50 different chains that barely talk to each other.

4

Coming at this from a DeFi practitioner angle—I build yield optimization strategies for a living, so I watch TVL flows and liquidity depth obsessively. The data tells a pretty clear story about where we’re headed.

Follow the Money: Where TVL Is Actually Going

In Q1 2026, we’ve seen a massive shift in where capital is deployed:

  • Base leads with 33-46% of L2 DeFi TVL (depending on how you measure)
  • Arbitrum holds steady at 31-44%
  • The remaining 20-30% is split across Optimism, zkSync, Starknet, and dozens of smaller L2s

But here’s what the raw numbers don’t show: liquidity fragmentation is killing capital efficiency.

When the same DeFi protocol deploys on 4 different L2s, you end up with:

  • Smaller liquidity pools on each chain
  • Higher slippage for larger trades
  • Worse yields for LPs (less trading volume per pool)
  • Arbitrage bots draining value trying to keep prices in sync

As someone who runs automated yield strategies, this is a nightmare. I need to maintain liquidity positions across multiple chains, bridge funds constantly (paying fees and taking bridge risk), and deal with inconsistent pricing.

The Harsh Reality: Fragmentation Hurts Users

@bridge_builder_ben is absolutely right about the systemic risks. From a purely economic standpoint, fragmentation means:

  1. Capital inefficiency: $100M split across 5 chains is way less useful than $100M on one chain
  2. Price impact: Smaller pools = higher slippage = worse execution for users
  3. Yield compression: LPs earn less because volume is fragmented
  4. Bridge costs: Every cross-L2 move costs fees and time

The protocols deploying on 10+ L2s aren’t doing it because it’s optimal—they’re doing it because that’s where users are. But this creates a coordination failure where everyone spreads thin trying to be everywhere.

L1’s Role Today: High-Value Settlement Only

In my day-to-day work, L1 Ethereum has exactly one use case now: large value transfers and bridge settlements.

When I’m moving $500K+ between protocols or doing treasury rebalancing, I use L1 because:

  • Security matters more than speed at that scale
  • Paying $10-20 in gas is negligible on a $500K move
  • L1 has the deepest liquidity for major assets (WETH, stablecoins)

But for anything under $10K? L2s all day. The fee savings are just too big to ignore.

Is this sustainable? Honestly, I think so. L1 evolved into exactly what it should be—the secure settlement layer for high-value transactions and L2 batch posting. Users don’t need to care about L1 if L2s work seamlessly.

Market Consolidation Is Already Happening

@ethereum_emma mentioned natural consolidation to 3-5 dominant L2s, and I completely agree. We’re already seeing this play out:

  • Base is winning the “normie DeFi” market (thanks to Coinbase’s distribution)
  • Arbitrum is the go-to for serious DeFi protocols and power users
  • Optimism is carving out governance and public goods funding narratives

The other 47 L2s? Most will fade into irrelevance as liquidity concentrates where network effects are strongest.

This is just basic market dynamics. Liquidity attracts more liquidity. Developers deploy where users are. Users go where liquidity is deep. It’s a flywheel that favors the top 3-5 chains.

The Security Budget Question Is Real

@layer2_lisa raised a crucial point about L1 fee revenue. Let me throw some rough numbers out:

  • If 85% of activity moves to L2s, L1 fee revenue drops ~85%
  • L2s pay for DA (data availability) posting, but those costs keep falling thanks to EIP-4844
  • Validators still need incentives to secure the network

The math has to work. Either:

  1. L2 DA fees need to be sufficient to fund L1 security, OR
  2. We rely more heavily on ETH issuance to pay validators (inflationary), OR
  3. L1 fee revenue needs to come from high-value settlement activity

I don’t know which path Ethereum takes, but it’s a legit concern. If L1 becomes “just settlement infrastructure,” it better be profitable enough to sustain top-tier security.

My Bottom Line

From a pure capital efficiency and user experience standpoint: We need consolidation, not more fragmentation.

The 85% L2 shift is great for affordability and throughput. But we over-corrected by launching 50+ incompatible chains. The market will naturally consolidate to 3-5 dominant general-purpose L2s, and that’s a good thing.

L1 doesn’t need to be where users transact daily—it needs to be the rock-solid security layer that everything else builds on. That’s a feature, not a bug.

But @bridge_builder_ben’s point about interoperability is the real test: Can we make cross-L2 movement seamless and secure? If not, we’ll just end up with 3-5 walled gardens instead of one expensive mainnet. That’s not progress—it’s just moving the problem around.

5

This discussion has been fascinating, and I want to add a cryptographic perspective on why the L1-as-settlement-layer model is not just acceptable—it’s actually architecturally elegant from a ZK rollup standpoint.

The Beauty of Separation: Execution vs. Verification

From a zero-knowledge proof systems perspective, what we’re witnessing is the proper separation of computation and verification:

  • L2s handle execution: They process transactions, generate state transitions, and create ZK proofs of correctness
  • L1 handles verification: It validates proofs and ensures data availability without re-executing transactions

This is exactly how ZK systems are supposed to work. L1 doesn’t need to know what happened in every transaction—it just needs cryptographic proof that state transitions were valid. This is massively more efficient than having L1 execute everything itself.

ZK Rollups vs. Optimistic: The Finality Advantage

@layer2_lisa mentioned zkEVMs reaching production maturity in 2026, and this is where things get really interesting.

Optimistic rollups (Arbitrum, Optimism, Base):

  • Assume validity unless challenged
  • 7-day withdrawal delays for security
  • Lower computational overhead for proving

ZK rollups (zkSync, Starknet, Polygon zkEVM):

  • Prove validity cryptographically
  • Instant finality once proof is verified on L1
  • Higher computational cost but faster settlement

For users and applications that care about finality—institutional settlements, high-value DeFi operations, cross-chain transfers—ZK rollups are architecturally superior. You get L1 security with L2 speed, and you don’t wait 7 days to withdraw.

As ZK proving technology improves (which it is, rapidly), the computational overhead gap is narrowing. We’re reaching a point where ZK rollups can match Optimistic rollups on cost while offering superior security and finality guarantees.

Privacy as a Bonus

One thing nobody’s mentioned yet: ZK rollups can offer privacy that L1 never could.

With proper ZK circuit design, you can process transactions where:

  • Users prove they have sufficient balance without revealing exact amounts
  • Smart contract state is partially private
  • Compliance checks happen without exposing sensitive data

This opens up use cases—private DeFi, confidential voting, enterprise applications—that simply aren’t possible on transparent L1. If we’re evolving the Ethereum ecosystem, why not add privacy to the feature set?

The Data Availability Economics

@defi_diana and @bridge_builder_ben both raised concerns about L1 security economics. Let me add some technical context.

Post-EIP-4844, data availability represents 90% of L2 operating costs. This is actually good news for long-term sustainability:

  1. L2s must pay for DA: Every rollup posts transaction data to L1 (either as calldata or blobs)
  2. More L2 activity = more DA revenue for L1: As L2 usage grows, so does L1’s DA revenue stream
  3. Glamsterdam upgrade: The upcoming 200M gas limit and parallel execution make L1 more efficient at processing proof verifications and DA

The concern about “race to the bottom” is valid—but remember that L1’s security doesn’t just come from fees, it comes from staking. Validators earn both fees and staking rewards. As long as Ethereum remains the settlement layer for trillions of dollars in L2 economic activity, there’s strong incentive to secure it.

Why L1 Isn’t Obsolete—It’s Evolved

Think about it this way: L1 is the global notary for all L2 activity.

Every major L2 transaction ultimately gets finalized on L1 through:

  • Proof verification (ZK rollups)
  • Fraud proof mechanisms (Optimistic rollups)
  • Data availability commitments

L1 doesn’t need to see every transaction—it just needs to ensure that everything settling on it is valid and available. This is more important, not less, as L2 activity grows.

From a security standpoint, L1’s role as the trust anchor becomes more critical when billions of users interact with dozens of L2s. If L1 security degrades, the entire L2 ecosystem collapses. That’s why validator incentives and fee economics matter so much.

Fragmentation: A Real Problem with Technical Solutions

I agree with @bridge_builder_ben that fragmentation is problematic, but I’m optimistic about technical solutions:

  1. Shared sequencing: Multiple L2s can share a decentralized sequencer network, enabling atomic cross-L2 transactions
  2. ZK-based bridges: Prove state validity cryptographically instead of relying on validator committees
  3. Chain abstraction layers: Users interact with a unified interface that routes transactions across L2s transparently

These aren’t sci-fi concepts—they’re being built right now. By late 2026 / early 2027, I expect cross-L2 interoperability to feel significantly smoother than it does today.

The Security Budget Question: Math We Need to Do

@layer2_lisa asked a critical question: Can L1 sustain its security budget if fee revenue drops?

Let’s think through the scenarios:

Scenario 1: DA fees alone sustain security

  • L2s pay for blob space and proof verification
  • As L2 activity grows, total DA fees could exceed historical L1 transaction fees
  • This assumes L2 adoption continues accelerating

Scenario 2: Staking rewards carry more weight

  • ETH issuance (inflation) funds validator rewards
  • Fee revenue becomes secondary to staking yield
  • Holders accept minor inflation in exchange for security

Scenario 3: High-value settlement generates sufficient fees

  • L1 becomes the layer for institutional settlements, large DeFi moves, and L2 batch posting
  • Fewer transactions but higher average value
  • Fee revenue from settlement activity funds security

I suspect we’ll see a mix of all three. But it’s absolutely worth modeling different scenarios to ensure Ethereum’s long-term sustainability.

My Take: We Didn’t Make L1 Obsolete—We Made It Better

The 85% L2 transaction share isn’t a failure of L1—it’s L1 evolving into its optimal role.

  • Users get cheap, fast transactions on L2s
  • Developers deploy where users are (currently fragmenting, but will consolidate)
  • L1 provides global settlement, data availability, and security guarantees
  • ZK rollups offer instant finality and privacy that wasn’t possible before

Did we fragment too much? Yes, probably. Market forces will consolidate this to 3-5 dominant general-purpose L2s plus specialized rollups.

Is L1 irrelevant? Absolutely not. It’s the trust anchor for the entire ecosystem—more critical than ever, even if users don’t directly interact with it daily.

The real question: Can we solve cross-L2 interoperability and security standardization before fragmentation becomes a permanent UX problem? That’s the frontier we need to focus on in 2026-2027.