CLOB Wars: A New Dawn for DEX Trading

Introduction

CLOBs (Central Limit Order Books) have been at the center of discussions in recent weeks. While the term may sound unfamiliar to some, it is not a new concept. The so-called "CLOB wars" became a thing largely due to the success of Hyperliquid.

Hyperliquid, a decentralized exchange (DEX) and the first fully on-chain CLOB, rose to prominence by offering a trading experience (speed, liquidity, user experience, etc.) comparable to centralized exchanges (CEX) like Binance.

The choice to avoid private funding and instead distribute a massive airdrop of HYPE tokens to the community further fueled Hyperliquid’s success and contributed to HYPE’s strong market performance, ranking among the top performing assets this cycle, up 50.9% since the beginning of the year.

Hyperliquid’s success was a wake-up call to older DEXs, especially those still using Automated Market Makers (AMMs), showing that it’s possible to compete with ultra-fast CEXs and support market makers engaged in high-frequency trading (HFT) without sacrificing decentralization.

Thus, welcome to the CLOB wars.

---

What are CLOBs?

Introducing the concept of CLOBs

A Central Limit Order Book (CLOB) is a type of market structure commonly used in traditional finance (TradFi) and centralized exchanges, where buy and sell orders are aggregated and matched in a centralized manner.

This serves as the backbone for the advanced, fast and high performance trading experienced in CEXs. CLOBs act as a mechanism to record all buy and sell orders at various price levels and displays the highest price a buyer is willing to pay and the lowest price a seller is willing to accept.

Orders in a CLOB are typically matched based on price-time priority, meaning orders with better prices are matched first, and if multiple orders share the same price, the oldest order has priority. CLOBs support different types of orders, including market, limit, and stop orders:

• A market order is executed immediately at the best available price in the order book.
• A limit order lets a trader specify the exact price at which they are willing to buy or sell.
• A stop order is triggered when the market reaches a specific price and is often used as risk management or to enter or exit a position automatically.

For example, a trader can place a limit order saying, “I want to buy 1 BTC at $95,000,” which will only be filled if the price drops to that level. Alternatively, a market order will buy or sell BTC right away at the best current price.

The limits of DeFi

In the early days of DeFi, implementing CLOBs on-chain was nearly impossible due to blockchain limitations like high gas fees, slow transaction times, and poor scalability.

CLOBs require fast, low-cost execution and cancellations, low latency and complex matching logic. These are requirements that most layer 1 blockchains couldn't meet before, and are the reason why centralized exchanges have stood out so much in recent years.

For this reason, Hyperliquid designed a dedicated architecture specifically for high-frequency trading, rather than building on a general-purpose L1. Each action (orders and cancellations, at very high frequency) generates a large volume of on-chain data, and the Hyperliquid team was forced to design their own network to meet these requirements.

This observation is not new: it has been made since the emergence of DeFi on Ethereum. That's why a new path has been opened up for a more blockchain-native solution: Automated Market Makers (AMMs).

AMM v CLOB

Automated Market Makers (AMMs) revolutionized DeFi by enabling permissionless token swaps without relying on traditional order books, most famously the constant product formula (x × y = k) to set token prices based on their ratios in liquidity pools. Prices only shift when a trade alters the pool’s token balances, enabling simple, single-click swaps that bypass order-book complexity.

Although AMMs offer a liquidity advantage over CLOBs by enabling passive liquidity provision and reducing the need for constant order placement, they lack the advanced order types, precise price discovery, and high-frequency trading (HFT) features offered by CLOBs.

• Impermanent Loss: Liquidity providers face impermanent loss when token prices in the pool change. Meaning they would have been better off holding tokens.
• Slippage: AMMs can experience significant slippage and price volatility, leading to order executions at different prices than what is initially displayed to the user on screen. In contrast, CLOBs allow traders to set limit orders, matching at exact prices and avoiding slippage, making them better for large, precise trades.
• Front-running and MEV: Sacrificing a bit of decentralization to centralized sequencers used in CLOBs helps mitigate MEV.
• Capital Inefficiency: Despite the introduction of concentrated liquidity, most trading activity takes place within relatively narrow price ranges and a lot of capital remains idle.

Most importantly, CLOBs address institutional demand for sophisticated trading mechanisms on-chain, something AMMs can never provide. By restoring order-book modularity over the monolithic AMM model, CLOBs offer professional traders grade speeds comparable to CEXs.

---

Key Features of a CLOB in DeFi

Latency is the name of the game for high-frequency trading, so a decentralized CLOB must be modular, splitting functionality into distinct layers to overcome these delays. To maintain blockchain security, reduce costs, and achieve CEX-like speed, the following key components and features are necessary.

High-Performance Execution Layer

Custom L1s or L2s designed for decentralized CLOBs require ultra-fast block times (<200 ms), single-block finality for immediate settlement, and deterministic execution to prevent frontrunning.

However, rollups and many L2s can’t achieve true single-block finality because they still rely on their base L1 for ultimate settlement, which is why we distinguish between soft finality and hard finality.

• Soft finality: provisional confirmation received by the user. Transactions are accepted and irreversible under normal conditions, even though they’re not yet on the base L1. This gives rollups and order-book–based L2s their speed advantage, since users don’t wait for L1 inclusion.
• Hard finality: occurs only when the transaction is included in a block on the L1, making it cryptographically irreversible. Hard finality is the only path for specialized L1s like Hyperliquid and dYdX.

Order Book Infrastructure

A well-defined on-chain priority system: cancellation and post-only orders are executed before GTC (Good-Til-Cancel) and IOC (Immediate-or-Cancel) orders. This ensures that when a trader chooses to cancel a position, the cancellation is systematically prioritized over the placement of new orders

Sequencer and Escape Hatch Mechanism

Most CLOB DEXs rely on sequencers, centralized or semi-centralized operators that order, batch, and publish user transactions on-chain (and on L2 rollups update the L1 for final settlement). A major risk is sequencer censorship, where orders can be delayed, skipped, or user-censored. Well-designed systems should include an escape hatch that lets users bypass a malicious sequencer to securely withdraw their funds.

Zero Knowledge (zk) Infrastructure

Zero-knowledge (zk) proofs are perfect scaling and privacy solutions for off-chain matching engines. Projects like SuccinctLabs’ SP1 and RiscZero’s RISC-V exemplify this approach by enabling zk proofs to be generated for each batch of matched orders.

These proofs guarantee off-chain computation integrity and balance updates, enabling high-performance CLOBs with minimal trust assumptions and significantly lower costs.

Data Availability Layer

Blobs are a new form of cheap, high-capacity data storage provided by specialized Data Availability (DA) layers such as Celestia or EigenDA. These layers don’t execute or interpret the data; they simply guarantee that it is available for anyone to access and inspect. This removes the need to post such data directly to Ethereum, significantly reducing gas costs while preserving auditability and verifiability.

---

The CLOB landscape and key players

A mapping and comparison of CLOBs

Perhaps the best way to categorize the various projects in the sector would be by ecosystem. Key players would be analyzed by architecture, latency, liquidity, Price oracle, Security and Escape mechanism.

Specialized Layer 1

This category covers CLOBs based on a Layer 1 blockchain specifically designed for High Frequency Trading (HFT).

• Hyperliquid

Hyperliquid is undoubtedly the undisputed leader among CLOBs and a benchmark DEX for trading. But it's also a blockchain specifically designed to support the demands of such an application.

At the core of this blockchain lies the HyperBFT consensus mechanism, an optimized version of Byzantine Fault Tolerant (HotStuff/LibraBFT) algorithms. This mechanism enables the blockchain to achieve block times of less than one second, with a theoretical logical capacity of 2 million transactions per second (tps) and extremely low logical latency of about 70ms blocktime.

Hyperliquid L1 is a dedicated Layer 1 blockchain, sufficiently engineered for ultra-low-latency, high-throughput trading rather than general-purpose workloads. Oracle pricing is on chain and Hyperliquidity Provision (HLP), a vault within the platform, handles liquidity demands.

Additionally, Hyperliquid’s implementation of GTC and IOC orders directly at the Layer 1 level meant a perpetual DEX capable of competing with CEXs in terms of execution speed, liquidity depth, and advanced trading tools. On Hyperliquid, even with an ultra-fast block time of 70 milliseconds and immediate finality (single-block), the average perceived latency for a market maker remains around 200 milliseconds (median).

This latency arises from the transaction path, where:

Total perceived latency: Propagation to the leader node (10–30 ms) + Inclusion in the next block (70 ms) + Block propagation to validators and clients (30–50 ms) + Client-side confirmation (30–50 ms) = approx. 140–200 ms

Hyperliquid’s security model is based on a sovereign Proof-of-Stake (PoS) chain. Its censorship resistance relies on the decentralization of its validator set, which which will soon accommodate a validator set increase from 21 to 24 nodes.

The platform however, does not have an "escape hatch" that allows users to retreat. The ultimate recourse in the event of validator misbehavior or failure is off-chain social coordination and governance.

• dYdX v4

The dYdX platform is a DEX that allows trading of leveraged perpetual contracts. It features an off-chain order book and matching engine, with settlements occurring on chain.

dYdX v4 supports over 175 perpetual markets with permissionless listing and classic order types such as GTC, IOC, and post-only.

Liquidity is initially bootstrapped via MegaVault, an on-chain lending mechanism where users lock USDC to earn yield and backstop market liquidity.

Its features a distributed Order Book, instead of a single sequencer, every validator node maintains an in-memory copy of the order book. New orders are gossiped across the network, and the validator selected as block proposer for each round matches orders based on its local view. Matched trades are committed on-chain with approximately 1 second hard finality (2/3+ stake approval) and sustained throughput of around 2,000 TPS.

Censorship protection relies on the proof-of-stake (PoS) consensus. If a proposer censors a transaction, users can resubmit to other validators in hopes a non-censoring validator is selected next. Misbehaving validators face on-chain slashing. Price oracles are integrated directly into the protocol (Slinky oracle). All perpetual markets share a unified USDC collateral pool, enabling up to 20× leverage with cross-margining.

With the introduction of dYdX Chain to version 5.0, new features include ability to cancel orders in batches to improve efficiency, the parallelization of signatures, which improves the speed of order execution on the chain, and creation of isolated markets

L1s and their dApps

Similar to the Specialized L1s, these dApps are built on the sovereign Layer 1 and don't have the separate soft/hard finality model of an L2 rollup. It directly inherits the finality of the L1 chain it is built on.

• Monad chain: Kuru & PerplTrade

Monad is an EVM-compatible Layer 1 blockchain that improves on Ethereum’s limitations while maintaining seamless compatibility for existing smart contracts. Its custom MonadBFT consensus, derived from HotStuff, reduces communication rounds from three to two, yielding up to 10,000 TPS and 1-second block finality.

MonadDB is a solid-state–based state store that enables high-throughput nodes with minimal RAM requirements, and network security is ensured by a decentralized PoS validator set.

Built on Monad, Kuru is a fully on-chain DEX whose vaults blend AMM-style curves with order-book granularity to solve liquidity fragmentation. Anyone can deposit into a vault and earn trading fees just like an AMM LP.

PerplTrade, another on-chain CLOB DEX, focuses on perpetual futures: it leverages Monad’s throughput for real-time order updates, low-latency execution, an on-chain clearinghouse, cross-margining, and oracle-based pricing. PerplTrade is live on Monad testnet and will launch on mainnet in Q3 2025.

• Fogo Chain: Ambient Finance & Valiant Trade

Fogo is a Layer 1 SVM-compatible blockchain that runs the Firedancer client in its purest form, delivering real-time, high-throughput performance ideal for on-chain HFT. By leveraging Firedancer’s ability to raise Solana’s compute-unit limits per block, Fogo achieves ultra-low latency and bandwidth suited for CEX-level trading.

Its multi-local consensus dynamically balances validator co-location, providing sub-50 ms block times and <1 s finality, with the security of geographically distributed validation, initially under Proof of Authority (PoA) before moving to permissioned PoS.

Ambient Finance, a perpetual CLOB DEX, is building on Fogo to exploit. Enshrined Pyth oracle inside the consensus layer for real-time, millisecond-resolved price feeds. It also has frequent batch auctions to neutralize MEV and protect users from toxic order flow.

Additionally, Valiant Trade is developing its own perpetual DEX on Fogo

• Sei: Monaco

Sei is an EVM-compatible Layer 1 blockchain that currently delivers ~400 ms block finality and 12,500 TPS on mainnet, with its upcoming “Gigagas” upgrade targeting over 200,000 TPS. Its “Twin Turbo Consensus” suite of optimizations drives these ultra-fast finality times.

On Sei, Monaco is developing a fully on-chain CLOB tailored for high-frequency trading by leveraging the chain’s Parallelization Engine to maximize throughput. Other CLOB projects in the ecosystem include Citrex Markets and Oxium, each building order-book–based DEXs that capitalize on Sei’s low latency and high performance.

Rollups

These are Layer 2s or network extensions built on Ethereum or Solana, used to create a CLOB DEX.

• Bullet

Bullet (formerly Zeta) is a high-performance trading protocol built on Solana, offering spot trading, derivatives, and borrow/lend functionality. It leverages the Sovereign Labs SDK and integrates a RISC-V–powered zkVM for zero-knowledge execution, ensuring robust privacy and on-chain verification.

Bullet employs application-specific sequencing: its sequencer enforces predefined ordering rules, prioritizing maker orders and post-only types, to minimize toxic MEV. Trade settlement and data availability both occur on Solana, benefiting from its high throughput and secure PoS network of over 1,000 validators.

Price feeds are sourced from the Pyth Lazer off-chain oracle. Native to Solana, Bullet taps into the chain’s deep liquidity and can handle up to 7,840 exchange orders per second. Users receive a “soft” confirmation within 2–3 ms, as the sequencer batches thousands of transactions every ~3 seconds before submitting them to Solana for hard finality.

• Byreal

Byreal is a decentralized exchange developed by Bybit, aiming to combine centralized exchange liquidity with on-chain transparency. It features a smart routing engine that aggregates fragmented liquidity and supports gasless swaps with sub-200ms quote latency through a hybrid model combining CLMM (Concentrated Liquidity) and RFQ (Request for Quote) mechanisms.

The platform introduces a token launch system using a Smart Price Ladder, allowing users to place bids across multiple price tiers instead of a single fixed price (measures to reduce bot dominance during token launches). Byreal also offers Revive Vaults, bbSOL-native yield vaults designed to support capital rotation strategies, with integrated yield flows across the Solana ecosystem and beyond.

• Ethereal

Ethereal is a decentralized exchange that supports perpetuals, spot trading, borrowing/lending, real-world asset (RWA) products, and yield strategies. With <20 ms latency and approx. 1 million orders per second, it promises to be fast and secure. It is powered by USDe and Ethena for deep liquidity. Deposits of USDe into Ethereal use the OFT (Omnichain Fungible Token) standard via LayerZero, enabling seamless one-click deposits from any supported chain.

Ethereal is deployed as an EVM appchain with the following architecture: Settlement is via Converge and execution uses the Arbitrum environment with DA provided by Celestia. The platform is currently live on testnet.

• Drift Protocol v2

This is a Solana-native perpetual DEX featuring a hybrid on-chain/off-chain design for optimal performance and transparency. It features an onchain order book and settlement with an off-chain matching mechanism while leveraging JIT Auctions & Backstop AMM for deep liquidity.

The project features cross-margining, multi-asset collateral with up to 20× leverage, and lending and borrowing within the platform. Governance & Staking use $DRIFT tokens to stake, vote, and earn protocol rewards.

• Starknet: Paradex

Paradex is a perpetual DEX built on Starknet, a ZK-Ethereum rollup. It batches trades off-chain, computes a single STARK proof, and submits that proof to Ethereum, achieving much higher throughput, faster processing, and lower fees while relying on Ethereum’s settlement security.

Unlike most rollups that publish full transaction data for replay on L1, Starknet uses state diffs: it only posts the change in state before and after each batch, rather than raw inputs. This compact representation, combined with the STARK proof’s integrity guarantee, lets Starknet sustain around 7,000 TPS and onchain verification for a finality of around ~200 ms.

All without saturating Ethereum or requiring sequencer throttling. ZkSync similarly supports state diffs, so future DEXs like GRVT could adopt this model to avoid bandwidth constraints.

• Lighter

Lighter is a decentralized trading platform optimized for perpetual futures trading, built on its own zk-rollup (zkLighter) and secured by Ethereum. It supports full CLOB trading options such as market, limit, stop order, etc., and implements zk-SNARK proof for all trades and liquidations. Currently in private beta with a point program for early users

The protocol uses Merkle-based trees for efficient and verifiable storage of orders, accounts, and transactions. Instead of blobs, it uses an Order Book Tree, a unique and specialized data structure. This tree serves as a dynamic storage for active orders, supporting various order operations with constant circuit size. Latency and throughput are about 10k TPS and <5 ms latency for finality.

It also uses an on chain oracle to fetch index prices. No AMMs and liquidity comes from market makers and social trading vaults operated by experienced traders. To further improve security, they are exploring methods like timelock encryption. In a case of sequencer failure or censorship, users can reliably generate Merkle proofs to confirm their asset ownership and position details and withdraw against the last valid state root.

CLOBs on Blobs

These are projects commonly on Ethereum L2s (Layer 3) that combine the off-chain performance of a CLOB with the scalable infrastructure of a modular DA layer (Celestia), or blobs. This camp includes a formidable list of technically sophisticated teams, most made verifiable and trustless through ZK proofs.

• Rise chain: NitroDEX

RISE is an Ethereum Layer 2 blockchain optimized for speed. Its unique shredded-block architecture processes micro-batches (“shreds”) in parallel, enabling approximately 5 ms latency and throughput of around 50,000 TPS.

NitroDEX on RISE is developing a fully on-chain sequencer CLOB, leveraging Celestia for its data-availability layer. Although still in testnet and with limited public details, building on RISE positions NitroDEX as a key player.

• MegaETH: GTE and Valhalla

MegaETH is a high-performance Layer-2 solution engineered for real-time trading, delivering 1–10 ms latency and up to 100,000 TPS with consistent millisecond response times. It leverages EigenDA for data availability, ensuring scalable and cost-efficient off-chain data storage.

Several DEX dApps are already building on MegaETH. GTE offers both spot and leveraged trading alongside a token launchpad, with liquidity provision backed by an AMM pool and an escape mechanism that lets users fall back to MegaETH or to Ethereum’s L1 for secure withdrawals.

Valhalla provides additional perpetual and derivatives trading capability, similarly taking advantage of MegaETH’s low-latency, high-throughput infrastructure. It introduces a parallelized sequencer colocated with the MegaETH sequencer so that they can get the benefits of MegaETH’s latency while still getting app-specific sequencing and gasless trading.

• Aevoxyz

Aevo is an Optimism (OP)-Stack rollup tailored for derivatives trading, offering off-chain matching at around 5,000 TPS with on-chain custody and final settlement secured by state roots posted to Ethereum. It supports sophisticated features such as spreads and portfolio margining, across more than 150 trading pairs. Aevo’s mainnet is live, featuring growing volumes and a native $AEVO token for governance and trading incentives.

• derivexyz

This is another feature-rich, on-chain CLOB built on Derive Chain, an OP Stack. It employs a centralized matching engine for ultra-low-latency order execution, while all trades, margin updates, and liquidations settle on-chain to ensure transparency and security.

The platform supports perpetuals, vanilla options (calls and puts), and spot trading under a unified cross-margin framework, accepting multiple collateral types (USDC, WETH, WBTC). Subaccounts allow traders to isolate risk per strategy, enabling sophisticated portfolio management without cross-position exposure.

Multichain

• Hibachi

Hibachi is a high-performance, privacy-preserving perpetual DEX built around an off-chain CLOB architecture. At its core, Hibachi executes orders off-chain very fast  while using zkproofs via Succinct to verify every balance update, order fill, and system state. These proofs are posted to Celestia as encrypted data blobs with final validation on Ethereum.

Ultra-fast off-chain matching for CEX-level speed, zkproof settlement to verify all updates onchain, encrypted trade data to protect user privacy and permissionless withdrawals with onchain guarantees. Latency is about 6 ms. Hibachi sources price feeds off-chain from PythNetwork and StorkOracle

If censorship is detected, users can submit a forced transaction, which is guaranteed to be processed. All that’s needed to exit is access to the latest encrypted blob, which can be decrypted on Arbitrum or Base and used to recover funds, independent of Hibachi’s operator.

Censorship resistance is further strengthened by enabling anyone to register as a sequencer, secured via depositing a bond into the sequencer, a radical departure from existing rollups that rely on a single trusted sequencer.

• Vertex Protocol

Vertex is a multichain, high-performance perpetual DEX whose Vertex Edge architecture unifies an off-chain order book with on-chain settlement and custody. Liquidity is aggregated across eight chains—Arbitrum, Base, Sei, Blast, Mantle, Sonic, Abstract, Berachain, and Avalanche—to maximize depth and capital efficiency.

Off-chain sequencer matches orders in real time (CEX-style latency of 5–15 ms) and onchain clearing house handles margining, risk management, and settlement.

Vertex Integrated AMM provides baseline liquidity and supports both spot and perpetual markets with cross-margining. Governance is powered by the $VRTX token, distributed through airdrops and liquidity-mining incentives.

---

A New Dawn for DEXs

A fair question might be, “If decentralized trading is  striving to replicate what centralized exchanges already excel at, why bother?” The answer is simple: self-custody and permissionless trading. These are core principles that define DeFi: empowering users with full control over their assets and the freedom to trade without intermediaries or gatekeepers.

This is widely understood in crypto, and I believe we’re on the cusp of a paradigm shift: most trading activity will soon migrate fully on-chain. We’re already seeing early signs. Bybit, for instance, has expanded into TradFi markets (gold, stocks, and forex) to stay relevant. Coinbase has also announced plans to begin offering perpetual-style crypto futures to its users. Yet the most striking development is Byreal, a DEX incubated by Bybit that delivers CEX-grade speed with DeFi transparency. As traders demand more control and greater transparency, other CEXs are likely to follow suit.

Another reason I’m convinced DEXs will dominate is the growing interest from industry leaders. Binance founder, CZ, is reportedly exploring investments in on-chain dark-pool DEXs and perpetual platforms to protect large trades from front-running, an issue underscored when James Wynn lost over $100 million on Hyperliquid.

Although that loss sounds alarming, it actually demonstrates the presence of serious “taker flow” (high-stakes, aggressive retail and speculative traders). Such activity attracts institutional market makers, who provide the continuous liquidity backbone. Keep an eye on emerging dark-pool DEXs and on-chain perpetuals; they may well define the next frontier in decentralized trading.

---

Conclusion

General-purpose chains attempting to balance decentralization and performance face a fundamental constraint: they can never truly offer colocation or endgame-level latency.

The reason lies in how consensus sits along the path to execution. As these networks expand their global validator sets to remain decentralized, they inevitably introduce latency and variability, factors that undermine the low-latency execution needed for high-performance trading.

This is why CLOBs on Blobs architecture is so compelling. It marks a radical shift away from the monolithic “world computer” model  toward an unbundled, modular stack where execution, settlement, and data availability are delegated to specialized, interoperable layers, each hyper-optimized for a single function.

Ethereum is well-positioned here with its rollup-centric roadmap. Meanwhile, Solana, through integrations like Bulk Trade may offer an alternative path, approaching the speed of Layer 1 platforms like Hyperliquid while maintaining a broader validator set and greater decentralization.

Ultimately, the “CLOB Wars” won’t be decided by technology alone. The winners will be the platforms that synthesize the best of both worlds: the speed, liquidity, and UX of centralized exchanges, combined with the self-custody, transparency, and composability that only decentralized systems can deliver.

The race is on, innovation is relentless, and in the end, the real winners are the users, who stand to gain from a more open, performant, and decentralized financial future.