Nine centralized cryptocurrency exchanges now operate proprietary Ethereum Layer 2 chains, up from one (Coinbase's Base) in mid-2023. The addition of South Korea's Upbit, which announced GIWA Chain on the OP Stack at Consensus Miami on May 5, 2026, marks the latest entrant in a structural shift: ...
"We want to own our infrastructure rather than rent block space from third parties." — Katie Haun, Founder, Haun Ventures (Bloomberg, May 4, 2026)
Nine centralized cryptocurrency exchanges now operate proprietary Ethereum Layer 2 chains, up from one (Coinbase's Base) in mid-2023. The addition of South Korea's Upbit, which announced GIWA Chain on the OP Stack at Consensus Miami on May 5, 2026, marks the latest entrant in a structural shift: exchanges are vertically integrating from order matching into block production.
The economics are straightforward. Coinbase's Base generated $78.2 million in sequencer revenue in 2025, accounting for 87.2% of all OP Chain sequencer revenue. Base's bridged total value locked (TVL) reached $13.07 billion as of May 2, 2026, with $4.49 billion in DeFi TVL. The exchange-operated L2 model converts what was previously an Ethereum gas expense into a revenue line item — sequencer fees flow directly to the exchange's balance sheet. Coinbase reports these fees under "other transaction revenue," and analysts at IG International estimate Q1 2026 Base sequencer revenue at $50 million to $80 million.
The trend now extends well beyond Coinbase. Kraken's Ink chain, OKX's X Layer, Bybit-affiliated Mantle, Gate.io's Gate Layer, HashKey's HashKey Chain, and Robinhood's forthcoming Robinhood Chain have all entered the market. Together, exchange-operated L2s account for the majority of TVL and transaction volume across the Ethereum rollup ecosystem.
As of May 6, 2026, nine major exchanges operate or are developing proprietary L2 infrastructure:
| Exchange | Chain Name | Stack | Status | TVL (approx.) | |----------|-----------|-------|--------|---------------| | Coinbase | Base | OP Stack | Mainnet (Jun 2023) | $13.07B bridged | | Kraken | Ink | OP Stack | Mainnet (Dec 2024) | ~$500M | | OKX | X Layer | Polygon CDK | Mainnet (Apr 2024) | ~$25M pre-Aave | | Bybit/BitDAO | Mantle | OP Stack (modified) | Mainnet (Jul 2023) | ~$1.82B | | Gate.io | Gate Layer | OP Stack | Mainnet (Sep 2025) | Not disclosed | | HashKey | HashKey Chain | OP Stack | Mainnet (Dec 2025) | Not disclosed | | Upbit/Dunamu | GIWA Chain | OP Stack | Testnet | N/A | | Robinhood | Robinhood Chain | Arbitrum | Testnet (Feb 2026) | N/A | | Sony (via exchange arm) | Soneium | OP Stack | Mainnet (Jan 2025) | Not disclosed |
The pattern is clear: seven of the nine chains use Optimism's OP Stack. Only OKX chose Polygon's CDK, and Robinhood chose Arbitrum. The OP Stack's dominance in exchange-operated chains has made Optimism's Superchain the de facto settlement network for centralized exchange infrastructure.
The core economic proposition of running an exchange-operated L2 is sequencer revenue. The sequencer — the entity that orders, batches, and posts transactions to Ethereum L1 — collects fees from every transaction on the network.
According to data compiled by a16z Crypto's research division, the top three L2 sequencers earned approximately $161 million in 2025: Base at $93 million, Arbitrum at $42 million, and Optimism at $26 million. For Coinbase specifically, Base's $78.2 million in network-level sequencer revenue in 2025 represented a 30x increase from the prior year, according to PANews.
The Superchain's revenue-sharing arrangement requires member chains to contribute the greater of 2.5% of sequencer revenue or 15% of sequencer profit (revenue minus L1 posting costs). Base contributed $42.4 million of the Superchain's total $48.4 million in collective sequencer revenue in H1 2025 alone — 87.2% of the total, per Messari's State of the Superchain H1 2025 report.
For exchanges, this model transforms gas fees from a cost center into a profit center. A centralized exchange that routes its users' on-chain transactions through its own L2 captures the sequencer spread — the difference between what users pay in L2 fees and what the exchange pays to post batched data to Ethereum L1. Post-Dencun (EIP-4844), L1 data posting costs dropped substantially, widening this spread.
The Optimism Superchain now comprises 34 formally governed OP Chains, with 50+ additional chains using the OP Stack independently. Per Superchain Eco data from February 2026, the Superchain accounts for 61.4% of the L2 fee market and processes approximately 13% of all crypto transactions.
In January 2026, the Optimism Foundation proposed directing 50% of incoming Superchain revenue toward monthly OP token buybacks, routing proceeds back into the Collective's treasury. This proposal, if implemented, creates a direct economic link between exchange-chain sequencer revenue and OP token value.
The concentration risk is notable. Base alone generates the vast majority of Superchain revenue. If Kraken's Ink, Upbit's GIWA, and other exchange chains scale to meaningful volume, Superchain revenue diversification improves. If they do not, the Optimism Collective's revenue base remains dependent on a single exchange operator.
Institutional adoption of the OP Stack accelerated in 2025-2026 with Upbit, HashKey, and EY-backed Nightfall deploying enterprise-grade L2s on the framework, according to Messari.
Exchange-operated L2s show a wide dispersion in TVL and user activity, suggesting that brand distribution — not technology — determines adoption:
Base leads by an order of magnitude. Its bridged TVL of $13.07 billion dwarfs all other exchange L2s combined. DEX volume on Base captured roughly half of all L2 DEX volume throughout 2025, per The Block's 2026 Layer 2 Outlook. Active wallets, transaction counts, and DeFi protocol deployments all track Coinbase's retail distribution advantage.
Mantle sits second among exchange-affiliated chains at approximately $1.82 billion in TVL, with over 180 dApps deployed. Its growth was driven partly by its modified OP Stack architecture and aggressive incentive programs.
Ink showed rapid early growth — TVL surged from $7 million in October 2025 to over $500 million by early 2026, per The Defiant. However, daily active users peaked at 157,000 in March 2025 and have since declined to roughly 49,000. The April 2026 Kelp DAO exploit, which left $195 million in bad debt across lending markets, contributed to a 33% TVL drawdown on Ink, according to CryptoNewsZ.
X Layer remains modest at approximately $25 million in DeFi TVL pre-Aave integration. Following Aave's deployment in late March 2026, early data showed a 20% rise in transaction volume, per CryptoNews. However, the gap between X Layer and Base is roughly 500x in TVL terms.
The data suggests a power-law distribution: the first mover with the largest retail user base captures the majority of value. Latecomers — even those backed by major exchanges — face steep adoption curves.
Upbit, operated by Dunamu Inc., is South Korea's largest cryptocurrency exchange. Its decision to partner with the Optimism Foundation on GIWA Chain (Global Infrastructure for Web3 Access), announced at Consensus Miami on May 5, 2026, carries regional significance.
GIWA Chain will be the first to operate on the Self-Managed tier of OP Enterprise, meaning Dunamu will run the sequencer directly while the Optimism Foundation provides technical support and a backup sequencer for network resilience, according to CryptoBriefing.
During its testnet phase, the network has processed nearly 100 million experimental transactions. Mainnet launch is pending completion of security audits. GIWA's team set up a dedicated booth at Consensus 2026 to court global developers, per BloomingBit.
From an economic-value perspective, Upbit's motivation mirrors Coinbase's: capture sequencer revenue, reduce dependency on third-party L1 gas costs, and create a walled-garden ecosystem where Upbit users transact on Upbit-controlled infrastructure. South Korea's crypto market, which at times exceeds U.S. spot volume on a per-capita basis, could provide GIWA with a substantial captive user base.
Robinhood's entry represents a different vector. Robinhood Chain, built on Arbitrum (not the OP Stack), launched its public testnet on February 10, 2026 and processed 4 million transactions in its first week, according to Yahoo Finance.
The chain's design purpose is explicitly tied to tokenized real-world assets — equities, ETFs, and other securities — rather than DeFi-native activity. According to Robinhood's newsroom, the chain aims to support 24/7 trading and self-custody of tokenized stocks, positioning it as infrastructure for on-chain brokerage rather than a general-purpose L2.
Robinhood Chain's choice of Arbitrum over the OP Stack is notable. It suggests that not all exchange operators view the Superchain's governance structure and revenue-sharing obligations as optimal. Arbitrum's architecture offers different trade-offs in sequencer sovereignty and ecosystem alignment.
Three structural risks warrant monitoring:
Sequencer centralization. Every exchange-operated L2 currently runs a centralized sequencer. The exchange itself orders transactions, creating the possibility of front-running, censorship, or selective transaction inclusion. While no major incidents have been documented, the architecture concentrates ordering power in the hands of entities that also operate order books — an inherent conflict of interest. Shared or decentralized sequencer solutions remain in research phases.
Regulatory exposure. Exchanges that operate L2s become infrastructure providers, not merely matching engines. If regulators classify L2 sequencing as a financial market utility or payment system, compliance obligations could escalate. The SEC's March 2026 guidance on crypto asset classification, while providing a coherent token taxonomy, did not explicitly address L2 sequencer operations.
Ecosystem fragmentation. Nine exchange L2s, each with separate liquidity pools and bridge mechanisms, fragment on-chain liquidity that could otherwise concentrate on a smaller number of networks. Cross-chain interoperability within the Superchain mitigates this for OP Stack chains, but Robinhood (Arbitrum) and OKX (Polygon CDK) remain isolated. The Kelp DAO exploit demonstrated how cross-chain bridge mechanisms — the connective tissue between these fragmented L2s — remain a primary attack surface.
The exchange-operated L2 thesis is now a market reality, not a hypothesis. In under three years, the model has progressed from Coinbase's single experiment to a nine-participant race. The economics are clear: sequencer revenue turns gas from an expense into income. The $78.2 million that Base generated in 2025 proves the model at scale.
The question is no longer whether exchanges should build L2s. It is whether more than two or three can sustain meaningful TVL and user activity. Base's dominance — holding more TVL than all other exchange L2s combined — suggests the same power-law dynamics that govern exchange market share will govern exchange-chain adoption. Upbit's GIWA and Robinhood Chain will test whether captive regional user bases and tokenized-asset specialization can carve out defensible positions.
For the Optimism Collective, exchange-operated chains represent both the greatest revenue opportunity and the greatest concentration risk. If the Superchain's revenue diversifies across Base, Ink, GIWA, and HashKey Chain, the 50% buyback proposal becomes a significant value-accrual mechanism. If Base continues to generate 87% of revenue, the Superchain's economic model rests on a single counterparty.
The infrastructure layer of crypto is quietly consolidating into a handful of vertically integrated exchange-chains. The entities that match orders now also sequence blocks. Whether this concentration improves or degrades market structure will depend on governance safeguards that, as of May 2026, do not yet exist.