Introducing the Blockworks MCPLearn more

Get a demoExplore data

What are the differences between parallel EVMs and Arbitrum Stylus?

Offchain Labs’ tech lead Rachel Bousfield breaks down some misconceptions on the effectiveness of parallel EVMs

article-image

MarinaNy/Shutterstock modified by Blockworks

share

Parallelizing the Ethereum Virtual Machine (EVM) has been a topic of interest for many in the cryptocurrency space in recent weeks.

Parallelization increases transaction throughput and improves blockchain scalability by executing multiple transactions simultaneously rather than in the sequential order in which they arrive. 

However, Rachel Bousfield, tech lead at Offchain Labs, told Blockworks in an interview that people often overstate the value of parallelism.

“The kinds of returns and fee reductions that people expect don’t really play out in practice,” Bousfield said. “A lot of the times when people talk about fees being cut due to parallelism, it’s often cut due to other things that are around it.”

Read more: Parallelized EVMs are gaining popularity, but they won’t scale blockchains alone

Bousfield explains that different blockchains have different hardware requirements. In the case of Ethereum, the blockchain is designed to ensure that running a node is relatively inexpensive and that low-end computers are able to meaningfully contribute to the Ethereum network in the form of running applications or running validators.

This differs from other blockchains, which may enable parallelism, but the cost demands for running a node become much higher. 

Ryan Watkins is the co-founder of Syncracy Capital, on Solana — a network that does enable parallelism. According to a post by Watkins, it is estimated that the cost to run a node is five times more expensive to run than Ethereum nodes. He said that the network currently has an estimated 40% of the number of nodes that Ethereum has.

“If Ethereum wanted to, they could dramatically increase the demands and costs of running an Ethereum validator, and you would see performance improve, there’d be more capacity, and people’s fees would be lower, but is that really the kind of scaling that Ethereum wants in its future? It’s not clear to me that that would be an advisable decision,” Bousfield said.

Read more: Scaling Ethereum’s virtual machine is a ‘solvable problem,’ says Monad Labs’ Galler

Additionally, Bousfield notes that parallelism enables throughput to improve efficiency when there are multiple users wanting to do different kinds of things in crypto at the same time. 

“The problem is that in real life, the actual demand we see on these blockchain networks is when people want to do very similar things to each other. When there is an airdrop, everyone wants to mint it at the same time. When there is a price discrepancy between DEXs, everybody wants to rush in and get MEV arbitrage out of it,” Bousfield said.

She notes that the technical term to describe this type of activity is called “contention,” adding that gas prices are often the highest when multiple people hope to do the same thing.

In fact, a recent study by Polygon Labs shows that parallelism is applicable to around 55% of the transactions in most blocks on its network. 

“This means that if parallelism was perfect, had a million cores, and ran really, really fast to the point where everything parallelized was executed instantaneously, you could, at best, double Polygon’s capacity with that,” she explained. 

With that said, Bousfield notes that parallelism itself is not a bad thing, but it is not the silver bullet that many are anticipating.

On tackling the problem around transaction speeds and ways to increase throughput, Bousfield notes that Arbitrum Stylus achieves this by making it simpler for hardware to read and interpret data.

In a traditional EVM, when the hardware receives data, it must check that it is accurate, enable branches and simulate it in memory — steps that often take a great deal of time. In contrast, Bousfield notes that Stylus is designed to speak the language of the central processing unit (CPU).

“By removing that layer of interpretation, Stylus is able to get a 10-100x speed on all compute workloads,” she said. “I think that strategies like this are where the big gains are going to be.”


Get the news in your inbox. Explore Blockworks newsletters:

Tags

Decoding crypto and the markets. Daily, with Byron Gilliam.

Upcoming Events

Marina Bay Sands Singapore

October 7, 2026

DAS Asia is a a single-day summit at Marina Bay Sands Singapore featuring conversations between the builders, investors, and global leaders are shaping the trajectory of the digital asset ecosystem in Asia & North America.

Hilton Park Lane, London

November 10-11, 2026

DAS London is a two-day summit at the Hilton Park Lane in London featuring conversations between the builders, allocators, and policy makers who are shaping the trajectory of the digital asset ecosystem in the UK, Europe, and North America.

Javits North

March 30-April 1, 2027

DAS NYC is a three-day summit at Javits North bringing together the builders, allocators, and policymakers shaping the future of digital assets across institutional finance and global markets.

recent research

Sponsored Article Template - DoubleZero White.png

Research

Every distributed system inherits the physical limitations, or advantages, of the network it runs on. Physics-bound systems can optimize their execution layers, compress their state, and parallelize their runtime, and still lose time they cannot recover to the speed of light through fiber and the routing decisions of carriers that are not optimized for consensus. The public internet routes traffic by least cost, not best performance. For a system where the order and timing of messages determine who earns and who loses, that is a structural tax and a measurable networking design flaw for traders.

The daily brief, in your inbox

Markets, protocols, and policy, read by 200K+ professionals.

Blockworks Inc.133 W 19th St., New York, NY 10011

Solutions

InvestorsExchangesEnterprisesOnchain Businesses

© 2026 Blockworks Inc.

TermsPrivacy PolicyPrivacy CenterNews Archive