Ethereum's Layer 2 ecosystem has entered a definitive consolidation phase. Of the 73 active rollups tracked by L2BEAT as of mid-2026, three networks — Arbitrum One, Base, and Optimism — process approximately 90% of all L2 transactions and control roughly 77% of combined DeFi liquidity. The remain...
"There were way too many general-purpose layer twos, which frankly don't make sense as a product, because there's no reason to have many, many versions of the same thing." — Ben Fisch, CEO, Espresso Systems
Ethereum's Layer 2 ecosystem has entered a definitive consolidation phase. Of the 73 active rollups tracked by L2BEAT as of mid-2026, three networks — Arbitrum One, Base, and Optimism — process approximately 90% of all L2 transactions and control roughly 77% of combined DeFi liquidity. The remaining 70 rollups split the residual 23% of TVL and 10% of transaction volume.
The consolidation accelerated after EIP-4844 eliminated data-posting costs as a meaningful differentiator, exposing the real competitive variable: application ecosystem depth and distribution reach. Loopring, Ethereum's first production ZK rollup, shut down its exchange in June 2026. Zero Network closed in May. The pattern is consistent with a market that overproduced general-purpose blockspace and is now correcting toward a leaner topology.
Meanwhile, Ethereum mainnet fee revenue has collapsed. Median transaction fees fell 99.68% from Q1 2024 ($3.79) to Q1 2026 ($0.012). Daily L1 fees that once topped $30 million now hover near $500,000. The value accrual question — whether ETH benefits from L2 growth — remains unresolved.
The L2 market has bifurcated into a small cohort of winners and a long tail of low-activity chains. As of late May 2026, the TVL distribution is as follows:
| Network | TVL (USD) | L2 Market Share | Daily Transactions | |---------|-----------|-----------------|-------------------| | Arbitrum One | ~$18.0B | ~40-44% | 4.30M | | Base | ~$13.5B | ~28-30% | 12.89M | | OP Mainnet | ~$5.6B | ~12% | — | | zkSync Era | ~$4.1-4.5B | ~9% | — | | Linea | ~$3.4B | ~7% | — | | Scroll | ~$2.1B | ~4% | — | | Starknet | ~$1.5B | ~3% | — | | All others | <$1B each | <10% combined | — |
According to data compiled by BlockEden.xyz and SpotedCrypto, Arbitrum and Base together account for approximately 77% of all Layer 2 DeFi liquidity. The combined aggregate throughput of the top three networks exceeds 4,000 TPS, versus Ethereum mainnet's approximately 15 TPS.
Base leads in raw user activity, processing over 37% of all L2 transactions with 382,500 daily active users as of February 2026. Arbitrum leads in capital concentration, holding the largest institutional and DeFi deposits. The two networks serve structurally different user bases — Base skews retail and consumer via its Coinbase distribution channel; Arbitrum skews DeFi and institutional via deeper protocol integrations.
The consolidation is not theoretical. Projects are closing.
Loopring (shutdown June 28, 2026): Ethereum's earliest production ZK rollup wound down its exchange and automated market maker, taking its relayer offline. TVL collapsed from approximately $760 million at its November 2021 peak to roughly $8 million at closure. The LRC token fell from $3.75 to approximately $0.01. According to the project's post-mortem, Loopring's original design lacked a virtual machine, limiting composability, and newer zkEVM architectures rendered it technologically obsolete.
Zero Network (shutdown May 2026): The gas-free EVM-compatible L2 backed by Zerion closed after approximately 18 months of operation. Zerion stated it would refocus resources on its core wallet and API products, effectively conceding the L2 market to better-capitalized competitors.
ZKsync Lite (shutting down 2026): Matter Labs is winding down Ethereum's first ZK rollup to consolidate operations around ZKsync Era and ZK Stack chains.
Polygon zkEVM: Among the first-generation designs deprecated as newer architectures made them obsolete.
According to CryptoTimes, more than 100 crypto projects shut down in H1 2026 across all categories. According to CoinDesk, the wave of closures has drawn comparisons to the dot-com bubble burst. Mahir Kılıç, an industry analyst, noted: "This pattern of closure and consolidation isn't unique to L2s or crypto; it's pretty common in tech."
21Shares predicts that by year-end 2026, the ecosystem will consolidate around "a leaner, more resilient set of networks."
The economic picture reveals a winner-take-most structure. Base, built and operated by Coinbase, serves as the clearest case study.
Base generated approximately $75.4 million in sequencer revenue during 2025. In May 2025, it reported a profit margin of 98.3% on $5.8 million in monthly revenue. For H1 2026, sequencer revenue is estimated at $60-70 million, according to Coinbase SEC filings which classify it as a "new revenue stream."
The margin is possible because L2 sequencers collect user fees while paying minimal blob data costs to Ethereum L1. According to Yellow Research, L2 networks retain approximately $321 for every $1 paid to Ethereum for data availability. Base specifically earned over $94 million in profit while returning only $4.9 million to Ethereum in blob fees.
This asymmetry — high L2 margins, minimal L1 payment — is the central tension in Ethereum's economic model. The L2s are profitable. The question is whether the value flows back to ETH holders.
EIP-4844, implemented in March 2024, introduced blob transactions that cut L2 data-posting costs by 90-99%. Transaction costs for end users dropped from $0.50-$5.00 to $0.001-$0.05. As of mid-2026, Base averages approximately $0.02 per transaction (the lowest among major L2s), Arbitrum One and OP Mainnet around $0.09, and zkSync Era around $0.07. For blob cost specifically, Linea reports the lowest at $0.012 per transaction.
The benefits to users are clear. The cost to Ethereum L1 is equally clear.
Ethereum mainnet median transaction fees fell 99.68% from $3.79 in Q1 2024 to $0.012 in Q1 2026, according to a peer-reviewed analysis published on arXiv. Total daily L1 fees, which once exceeded $30 million, now hover near $500,000. This fee compression directly undermines the "ultrasound money" narrative — ETH's burn rate has slowed proportionally, increasing net issuance.
According to Yahoo Finance and Phemex Research, Ethereum's market dominance has fallen to 10.4%, a three-year low. The debate over whether L2 growth constitutes cannibalization or maturation remains unresolved. Sygnum Bank has argued that cannibalization fears are "overblown," framing Ethereum as an evolving global settlement layer. Critics counter that a settlement layer that collects $500,000 per day while its L2 clients collect $321 for every $1 remitted is a subsidy arrangement, not a sustainable business model.
Optimistic rollups dominate with approximately 80% of total L2 TVL. ZK rollups hold the remaining 20%, a share that has remained stable despite significant technical maturation.
Within the ZK category, zkSync Era leads at $4.1-4.5 billion TVL, followed by Linea ($3.4 billion), Scroll ($2.1 billion), and Starknet ($1.5 billion). According to multiple ecosystem reports, the historical weakness of ZK rollups — harder EVM compatibility and expensive proving — has largely closed in 2026. Prover hardware has improved, zkEVMs have matured, and proving networks like Succinct's Open Prover Layer have commoditized ZK computation into a marketplace.
The remaining gap is in ecosystem depth. Arbitrum and Base benefit from years of developer and protocol accumulation. ZK networks, despite technical superiority in finality time and proof economics, have not yet overcome the liquidity and application moats built by their optimistic competitors.
One notable shift: according to market data, dominant ZK rollups collectively settle more stablecoin volume than every optimistic rollup combined. This suggests ZK networks may be capturing higher-value, payment-focused activity even as optimistic rollups lead in aggregate DeFi metrics.
The OP Stack and Arbitrum Orbit frameworks have lowered rollup deployment costs to near-zero. Combined application-specific rollup deployments on these two frameworks exceeded 100 as of Q1 2026. Enterprise adopters include Kraken (INK), Uniswap (UniChain), Sony (Soneium), and Robinhood (Arbitrum Orbit L2, which processed 4 million transactions during its first testnet week in February 2026).
Rollup-as-a-Service (RaaS) platforms like QuickNode now enable enterprise-grade rollup deployment in minutes. This commoditization of blockspace production has created a structural oversupply problem.
According to BlockEden.xyz, most new L2 and L3 launches in 2025 became "ghost towns within weeks of their token generation events." The ease of deployment has decoupled chain supply from demand. The result: dozens of technically functional chains with negligible usage, draining liquidity from the ecosystem via fragmentation.
The economic logic is circular. Low deployment costs encourage new chain launches. New chains fragment liquidity across more venues. Fragmented liquidity reduces the value proposition of each individual chain. Users and capital consolidate onto the largest networks, reinforcing the concentration.
Ethereum's Glamsterdam hard fork, targeting late August or Q3 2026, represents the mainnet's most significant upgrade since The Merge. Three core EIPs define the upgrade:
The upgrade does not directly address the L2 fee revenue cannibalization problem. Its primary impact is on mainnet execution capacity, which may enable Ethereum to recapture some activity that migrated to L2s purely for cost reasons. Whether 10,000 TPS on mainnet materially changes the competitive dynamic with L2s that already offer sub-cent transactions remains to be seen.
The Ethereum L2 ecosystem in August 2026 resembles a mature market undergoing natural selection. The overproduction of general-purpose blockspace — fueled by near-zero deployment costs and speculative token launches — has created a correction that favors networks with genuine distribution advantages (Base via Coinbase), deep protocol ecosystems (Arbitrum via DeFi integrations), and institutional credibility (Optimism via the Superchain narrative).
The economic question is not whether L2s work — they demonstrably do, processing millions of daily transactions at sub-cent costs. The question is who captures the value. At present, L2 operators capture sequencer revenue at 98%+ margins while Ethereum L1 collects diminishing blob fees. This is the inverse of the platform economics that most ETH holders anticipated.
The Glamsterdam upgrade may improve mainnet competitiveness, but it does not address the structural subsidy flowing from L1 to L2. Until Ethereum finds a mechanism to recapture proportional value from L2 activity — whether through blob fee market maturation, shared sequencing mandates, or other mechanisms — the L2 consolidation story is simultaneously a success story for Ethereum's scaling roadmap and an unresolved challenge for ETH's economic model.