The Road to Ethereum's Scalability: Moving Towards a New Era of 10,000 Transactions Per Second

On July 30, 2025, Ethereum will celebrate the tenth anniversary of its genesis block. After ten years of exploration, Ethereum's scaling roadmap is exploring new directions and visions.

Recently, the rise in ETH prices has brought confidence to the community, but what is even more exciting is that after several years of exploring L2 scaling, Ethereum L1 has finally found a credible path to achieve extreme scalability while maintaining maximum decentralization.

In short, starting now, the Gas limit and TPS of Ethereum are planned to increase several times each year. Validators will no longer execute each transaction repeatedly but will only verify a single zero-knowledge proof (ZK-proof) to prove that this batch of transactions has been executed correctly, thereby increasing the underlying network's TPS to tens of thousands per second.

At the same time, L2 will also synchronously scale, achieving hundreds of thousands or even millions of TPS. A new type of L2 called "native Rollup" will operate like programmable sharding, providing the same level of security as L1.

These proposals are built on ideas that Vitalik Buterin began exploring in 2017 and are supported by the core researcher of the Ethereum Foundation, Justin Drake.

Drake stated at the EthCC conference in July: "We are at a critical turning point for Ethereum scalability, and I firmly believe we are about to enter the GigaGas era of L1 — around 10,000 TPS, and the key to unlocking this era is zkEVM and real-time proof."

Drake's ultimate goal is to achieve 10 million TPS in the Ethereum ecosystem within 10 years, which means the future will inevitably be an "internet of networks" architecture: different L2s will each take on different scenarios, trade-offs, and advantages, collectively expanding the entire ecosystem to meet global demand.

Why has Ethereum L1 been unable to achieve large-scale scalability?

Although other blockchains have long attempted to use more powerful hardware and computing power to increase throughput, Ethereum has maintained an almost ideological obsession with decentralization.

From the perspective of ETH maximalists, certain "data center chains" have centralized risk points worth millions of dollars, and the government can directly conduct transaction audits on these nodes. Even chains with lower hardware specifications have cost and bandwidth requirements that are daunting, thus affecting the level of decentralization.

In contrast, Ethereum can even run on a Raspberry Pi. This low barrier design allows over 15,000-16,000 public nodes and millions of validators to participate in the network, making it almost impossible to censor transactions on Ethereum, and giving the entire network strong resilience against attacks.

Of course, the cost is extremely slow speed—current TPS is about 18-20 transactions per second, while some public chains have a TPS of about 1500 transactions per second.

To some extent, the blockchain architecture is inherently inefficient, somewhat like a spreadsheet; every time you modify a cell, all the computers around the world that have a copy must first recalculate the entire spreadsheet before they can update it, confirming that it is correct.

A co-founder of a certain ZK technology company explained: "The design of Ethereum is intended to allow anyone to keep up with the network and re-execute all transactions," which also means that the transaction volume cannot be arbitrarily expanded, because each transaction requires someone to recalculate.

It is precisely because the mainnet has limited scalability while maintaining decentralization that Ethereum had to embark on the controversial route of Layer 2 scaling in 2020.

How ZK Breaks the Blockchain Impossibility Triangle?

Ethereum founder Vitalik Buterin once proposed the concept of "Blockchain Trilemma", describing the dilemma of public chains being difficult to achieve security, scalability, and decentralization all at once.

Almost all scalability solutions can only satisfy two of the three items at the same time, inevitably sacrificing the third one.

Until now.

Zero-Knowledge Proof ( ZK-Proof ) is described as a "moon-level mathematics" technology — capable of mathematically proving that a large number of complex transactions have been correctly executed without revealing transaction details.

The process of generating a ZK proof is very complex, but verifying whether a proof is correct is both fast and lightweight.

Therefore, the future vision of Ethereum is: rather than having a bunch of underperforming Raspberry Pi nodes recalculate all transactions one by one, it is better to let validators only check the mathematical result of a very small ZK proof.

The co-founder of a certain tech company continued to explain, "Instead of having everyone redo all the transactions, it's better to simply give them a proof, telling them that these operations have already occurred, so that anyone can verify this proof without having to recalculate."

Drake even joked that in the future, the computational power required to verify ZK proofs would be so small that even a Raspberry Pi Pico(, priced at $7, would be sufficient, performing less than one-tenth of an ordinary Raspberry Pi), and there would be no need for large data centers.

zkEVM: The Roadmap to 10,000 TPS

Sophia Gold from the Ethereum Foundation recently sparked heated discussions in the community with a post on her blog: Within the next year, the L1 mainnet may integrate a zero-knowledge proof-driven Ethereum Virtual Machine (zkEVM).

It is worth noting that many practical explorations of ZK technology actually started from L2 networks, such as a certain ZK Rollup public chain incubated by Joe Lubin, one of the co-founders of Ethereum—any application that can run on Ethereum can run seamlessly on this chain.

The chain even regards itself as an extension of Ethereum and recently announced that it will burn 20% of the ETH transaction fees to support the value return to L1.

The chain leader explained that ZK technology provides an answer to the blockchain's impossible triangle: "The magic of ZK lies in the fact that we can significantly increase the Gas limit of L1, and the expansion of computational load does not make verification more complex."

He added that as the delays and costs associated with generating ZK proofs continue to decrease, we are able to handle higher throughput while keeping the hardware requirements for verification extremely low—so low that even a smartwatch can handle the verification work.

However, the community should not be overly optimistic; even if zkEVM is successfully integrated into L1 within the next year, it will not directly achieve 10,000 TPS on the first day.

Move a pawn every day, and then complete it in an instant.

Ethereum currently has five main software clients available for running the network, which means that even if one client encounters issues, the network will not directly shut down.

In the future upgrade roadmap, Ethereum plans to first release two to three modified clients that support ZK verification, allowing validators to choose to complete verification by checking zero-knowledge proofs (ZK-proofs) instead of re-executing each transaction.

Initially, only a few validators will switch to the new validation mode first, in order to identify and fix potential issues early on.

Ladislaus from the Ethereum Foundation's protocol coordination team stated, "Switching to a snarkified EVM will be a gradual process" — where "snark" refers to the use of SNARK-type zero-knowledge proofs.

Users will gradually feel the increase in the Gas limit of L1, which means that the economic activity capacity of the network is enhanced. Although the transition from L1 to ZK verification takes time, the expansion of the Gas limit is almost imminent.

Last week, the L1 Gas limit was just raised by 22%, reaching 45 million. Researcher Dankrad Feist proposed an EIP suggesting that clients automatically adjust the Gas limit three times a year. According to this plan, the Ethereum mainnet could achieve approximately 2000 TPS in four years.

Justin Drake even proposed extending this pace by two years, reaching 1 gigagas throughput by 2031, achieving approximately 10,000 TPS.