The Race for the Holy Grail of Parallel EVM: Monad, MegaETH, and Pharos

Preface

Recently, three heavyweight parallel EVM projects have simultaneously launched their testnets: Monad on February 19, MegaETH on March 21, and Pharos on March 24. After AI Agent, the core narrative of Web3 technology seems to be shifting back to parallel EVM—the hottest theme at the beginning of 2024.

The Ethereum Virtual Machine (EVM) is the heart of Ethereum, responsible for executing smart contracts and processing transactions. The EVM acts as a computational engine that provides abstraction for computation and storage. However, it lacks scheduling capabilities; Ethereum’s execution module pulls transactions one by one from blocks, and the EVM executes them sequentially. While sequential execution ensures deterministic order and enhances security, under high load it can lead to network congestion and latency.

Parallel EVMs overcome this limitation by enabling multiple operations to execute simultaneously, significantly increasing network throughput and enhancing blockchain performance and scalability. In practice, “parallel EVM” refers to high-performance EVM-compatible blockchains that not only introduce parallel execution but also feature comprehensive upgrades across consensus, transactions, pipelining, storage, and hardware acceleration—aiming to process more transactions in less time and effectively solve traditional blockchain congestion and delays.

This article will delve into the backgrounds, architectures, and design trade-offs of the three projects: Monad, MegaETH, and Pharos.

Monad

Monad is a high-performance, EVM-compatible Layer 1 blockchain developed by Monad Labs. It aims to enhance system scalability while maintaining decentralization, addressing the low-throughput issues of existing EVM-compatible chains.

Founded in 2022 by Keone Hon, James Hunsaker, and Eunice Giarta, Monad Labs brings together talent from top-tier institutions. Keone and James are former employees of Jump Trading, a major market-making firm, while Eunice comes from a non-crypto background.

In February 2023, Monad Labs raised $19 million in seed funding led by Dragonfly. In April 2024, they secured an additional $225 million in a round led by Paradigm, bringing Monad’s valuation to $3 billion.

Monad's key advantage lies in its capacity to handle up to 10,000 transactions per second (TPS) with a 1-second block time, achieved through optimizations in four critical areas:

1. MonadBFT: A high-performance consensus mechanism based on an improved version of HotStuff. It achieves agreement on transaction ordering under partial synchrony, even in the presence of Byzantine actors. First, MonadBFT uses a two-phase BFT algorithm with optimistic responsiveness—linear communication overhead in normal cases and quadratic during timeouts. Second, it employs a hybrid signature scheme: ECDSA ensures message integrity and authenticity, while BLS signatures aggregate votes and timeout messages, solving scalability bottlenecks. Nodes maintain local transaction pools instead of a global one, reducing bandwidth usage and latency. Transactions are forwarded via RPC nodes to upcoming leader nodes. Additionally, RaptorCast—a proprietary protocol—is used for message propagation. It converts block proposals into erasure-coded chunks and broadcasts them through a two-tier tree structure, leveraging full network upload bandwidth to disseminate proposals to all validators while preserving Byzantine fault tolerance. These features enable efficient and robust consensus;

   

2. Asynchronous Execution: By decoupling consensus from execution, Monad dramatically increases execution throughput. With consensus separated, execution now occupies the entire block time rather than just a fraction. Block proposals exclude state roots but include the state root from three blocks prior, allowing nodes to detect forks. Leader nodes build blocks using a delayed view of state. To defend against DDoS attacks, nodes verify that accounts have sufficient balance to cover the highest possible debit in pending transactions. Once a node receives a proposed block—even before finality—it begins local execution without guaranteeing eventual validation or finalization. This enables significant speed improvements, allowing a single-shard blockchain to scale to millions of users;

3. Parallel Execution: Monad adopts an optimistic execution model, starting subsequent transactions before earlier ones complete. Updated states are merged sequentially. When discrepancies occur due to incorrect assumptions, Monad tracks inputs used during execution and compares them with outputs from prior transactions, re-executing if necessary. A static code analyzer predicts inter-transaction dependencies to avoid invalid parallelism. In optimal conditions, many dependencies can be predicted early; in worst-case scenarios, it falls back to sequential execution. This approach boosts efficiency and throughput while minimizing transaction failures caused by parallelism;

4. MonadDB: A custom KV database central to Monad’s architecture, designed to deliver high performance while maintaining full compatibility with Ethereum. MonadDB natively implements the Merkle Patricia Trie structure on disk and in memory and includes a custom indexing system, eliminating reliance on underlying file systems for efficient on-disk storage. It leverages asynchronous I/O to take full advantage of modern kernel support, avoiding excessive kernel threads for pending I/O requests. Additional techniques such as concurrency control, sequential writes, and data compression further optimize performance. These enhancements reduce data access latency and accelerate transaction processing, boosting overall network performance.

MegaETH

MegaETH is currently the fastest Layer 2 blockchain, developed by MegaLabs. Its defining feature is real-time blockchain performance, offering ultra-low latency and high scalability for applications requiring immediate responses.

MegaLabs was founded in early 2023. CEO Li Yilong holds a PhD in Computer Science from Stanford and previously worked at Runtime Verification Inc. CTO Yang Lei is a PhD from MIT. CBO Kong Shuyao served as Head of Global Business Development at Consensys. Growth lead Namik Muduroglu has held roles at Consensys and Hypersphere.

In June 2024, MegaLabs raised $20 million in seed funding led by Dragonfly. By December 2024, they completed a community round on Echo platform, hitting their $10 million target within three minutes. MegaETH is now valued at over $200 million.

MegaETH delivers 100k TPS and ~10ms block times, achieving millisecond-level response even under heavy loads, enabled by the following technical innovations:

1. Node Specialization: Different nodes perform distinct functions with varying hardware requirements. There are three roles: Sequencers, a single centralized node responsible for transaction ordering and execution, eliminating consensus overhead. Sequencers publish generated blocks, witness data, and state diffs to EigenDA (a data availability layer), ensuring availability. Provers fetch blocks and witness data from sequencers and perform stateless verification using specialized hardware—validating blocks asynchronously and out-of-order without storing full chain state. Full Nodes receive state diffs from sequencers to update local state and use the proof network to verify block validity, ensuring consistency and security;

   

2. Targeted Optimization: MegaETH directly addresses key bottlenecks in traditional EVM chains. For high state access latency, it introduces a new Trie structure optimized for memory and I/O efficiency, scalable to multi-TB state without added I/O cost. For serial execution, the sequencer can adopt any parallel execution strategy. For inefficient interpreters, MegaETH uses JIT compilation to eliminate interpretation overhead, delivering near-native performance for compute-intensive dApps. For high bandwidth demands in state synchronization, it implements an efficient encoding and transmission method for state diffs, enabling large-scale updates even under bandwidth constraints. Advanced compression techniques allow complex transaction state sync within limited bandwidth;

3. Mini Blocks: MegaETH issues a pre-confirmation every 10 milliseconds, known as Mini Blocks. Standard EVM blocks incur significant overhead—the header alone takes over 500 bytes, and computing three Merkle roots is time-consuming—placing a heavy burden on light clients. Mini Blocks are generated in parallel with standard EVM blocks, providing identical inclusion guarantees but drastically shortening propagation intervals. Light clients use MegaETH’s proprietary Realtime API to access transactions included in Mini Blocks but not yet in finalized EVM blocks.

   

Pharos

Pharos positions itself as a high-performance EVM-compatible Layer 1 blockchain, aiming to build the best ecosystem for RWA and payments. It boasts exceptional performance: 50,000 TPS and 2 gigagas per second (2 billion gas units).

Founded in 2024, Pharos’ leadership team carries strong institutional credentials. CEO Alex Zhang formerly served as CTO of AntChain and later as CEO of ZAN, AntChain’s Web3 brand. CTO Wishlonger was previously CSO of AntChain. CMO Laura led marketing at Solana Labs and successfully sold out the first-generation Solana phone, Saga. COO Sally worked at OKX. CCO Matthew led ecosystem development at Stellar and Ripple.

In November 2024, Pharos raised $8 million in seed funding co-led by Lightspeed Faction and Hack VC.

Pharos introduced a "Degree of Parallelization (DP)" framework, categorizing blockchain parallelization into six levels (DP0–DP5). Ethereum sits at DP0—no parallelization. From DP1 to DP5, blockchains progressively implement enhanced consensus, parallel transactions, pipelining, parallel merkelization, accelerated state access, and parallel heterogeneous computing.

Pharos adopts a DP5 full-stack parallel architecture, upgrading every layer from consensus and transactions to pipelining, storage, and hardware acceleration:

1. Scalable Consensus Protocol: A high-throughput, low-latency BFT consensus protocol designed to fully utilize network resources;

2. Dual-VM Parallel Execution: A parallel execution layer combining EVM and WASM, powered by advanced compilation techniques;

3. End-to-End Asynchronous Pipelining: Enables parallel and asynchronous processing throughout each transaction’s lifecycle and across blocks;

4. High-Performance Storage with Authenticated Data Structures (ADS): Delivers superior throughput, low-latency I/O, and cost-efficient state storage, securely scaling to billions of accounts;

5. Modular Specialized Processing Network (SPN): Seamlessly integrates new software, hardware, and geographically distributed infrastructure, supporting diverse use cases and emerging technologies.

Conclusion

Thanks to Ethereum’s long-standing user education, the EVM has become the most widely adopted environment in Web3, hosting the largest developer community and dApp ecosystem—effectively serving as the JavaScript of the Web2 world. However, Ethereum’s scalability limitations have severely constrained EVM’s growth, making parallel EVM one of the most critical technological frontiers.

Monad strikes a balance between scalability and decentralization through its parallel execution model, offering developers 10,000 TPS without sacrificing EVM compatibility. Its independent consensus grants autonomy but forgoes Ethereum’s shared security, which may deter developers prioritizing trust and security guarantees.

MegaETH excels unquestionably in latency and throughput, delivering 10-millisecond block times and 100,000 TPS—ideal for applications requiring near-instant responses like GameFi, SocialFi, and high-frequency trading. However, its centralized sequencer design raises concerns about decentralization.

Pharos offers up to 50K TPS and 2 gGas/s processing power, matching the performance of leading high-performance EVM chains like Monad and MegaETH. With its “Ant Group DNA,” Pharos focuses on institutional clients and compliance-driven RWA-Fi, positioning itself to meet future market demand for compliant, high-efficiency blockchain infrastructure.

Based on public metrics, MegaETH and Pharos outperform Monad in raw performance. However, considering Monad’s substantial funding and development resources, it remains well-positioned for breakthroughs. Ultimately, there is no clear frontrunner among Monad, MegaETH, and Pharos—the choice for developers boils down to trade-offs between performance, decentralization, and specialization.