The Next Decade for Ethereum — A Fireside Chat Between Vitalik and Legislator Qiu Dagen During the Hong Kong Web3 Carnival

Writing by: Techub News Editor

Date: April 21, 2026

Location: Hong Kong Web3 Carnival · ETH Hong Kong Community Hub

Guest Speakers:

• Vitalik Buterin (Founder of Ethereum)

• Duncan Chiu (Hong Kong Legislative Council Member, Tech Innovation Sector)

Returning to Hong Kong, Goodbye to the Chinese-speaking Community

Duncan: Today the venue is packed. Welcome back, Vitalik, to Hong Kong. The organizer gave me a task: this session should be conducted in Mandarin. For me, that’s quite a pressure because I worry many people in the audience speak Mandarin better than I do.

You’ve already shared many technical details about Ethereum at various events these days. Today, I want to keep it relaxed, talk more about your personal journey—your relationship with the Chinese-speaking community over these more than ten years, and some of your recent thoughts. We can keep the technical parts brief.

Let’s start with memories: from when you founded Ethereum to today, you’ve spent a lot of time in China, Hong Kong, and the Chinese-speaking world. How do you view this journey with the Chinese community? What are your special feelings about the builders and the community here?

Vitalik: The first time I truly heard about China’s crypto community was in 2013, before Ethereum existed, when only Bitcoin was around. I heard that there was already a large group of miners and exchanges in China, bigger than in the US, but almost no one in the English media was talking about this world, so I was very curious.

In May 2014, I visited China for the first time, going to Beijing, Shanghai, Hangzhou, Shenzhen, meeting many miners and exchanges. At that time, the ecosystem here could be said to have two biggest parts: one was miners, and the other was exchanges, both very large. I remember some companies already had hundreds of employees, while some top US companies back then had fewer than a hundred.

At the same time, there was also early thinking about smart contracts. Some teams proposed that each application could be a separate blockchain, and these chains could interact; others were writing white papers, researching how to generate decentralized randomness, which later became part of the Ethereum protocol, such as mechanisms for selecting the next block producer.

After the pandemic, many changes happened. Starting around 2022, with ZK technology, the new roadmap, and the development of Layer 2 solutions, a large number of new developers came in. Initially, many focused on L2 and DeFi, but gradually some started participating in core development, contributing to EIPs, discussing when to raise gas limits, and when to introduce ZK-EVM.

So, to summarize in one sentence, my relationship with the Chinese community started with “miners and exchanges,” and has gradually extended to participating in every layer of the tech stack—including very deep involvement in the core protocol. I think this change is very positive, but there’s still a long way to go together.

Good L2s Can’t Just “Copy Ethereum”

Duncan: From your launch of Ethereum to now, over these years, Ethereum has undergone many upgrades—you just mentioned ZK, and different phases of EVM upgrades, including one called “Shanghai Upgrade,” which also shows your affection for this land.

Recently, you’ve also shared many views on L2 solutions. You mentioned that: L2 cannot just be about scaling up, nor simply a copy of L1. Could you elaborate on how you see the development of L2 now?

Vitalik: L2 remains very important, but I believe a good L2 shouldn’t just be “another Ethereum” or a mere EVM clone.

In my view, L2 should complement L1:

• L1 provides the foundational security and trustless settlement layer;

• L2 handles those things that L1 finds difficult, such as stronger privacy, higher scalability, lower latency, and complex oracle dependencies.

If a team says “We’re building an L2,” and the result is just moving L1 logic over, using the same EVM to do the same things, then they’re not fully leveraging the design space that L2 can offer.

We’ve also seen that the most successful L2s, although initially emphasizing EVM compatibility, over time have started to develop “non-EVM” features and architectures. These are their real differentiators and advantages.

Duncan: So, simply put, L2 must have its own “purpose.” It shouldn’t be just for the sake of L2; it should be designed around specific scenarios.

Vitalik: Exactly, that’s it.

Ethereum Short-term Roadmap: Data Scaling and Compute Scaling

Duncan: You recently released a new roadmap. I even took a photo of it yesterday in the audience. It divides future development into short-term, mid-term, and long-term.

I especially want to ask about two key points in the short-term: data scaling and compute scaling. AI is very hot now, and everyone is curious about how “scaling” will be achieved. Can you be more specific about what you aim to accomplish in these two areas?

Vitalik: First, why does L1 still need to be expanded? Many people say, “Since there’s L2, why not just put most logic there?” Theoretically, that’s true—many things can be moved to L2 or higher layers. But this introduces a problem: users must trust more intermediaries.

Ethereum’s core principles—reducing trust, minimizing middlemen, enabling users to verify themselves—are hard to fully realize without L1. So, the data and computational capacity of L1 must continue to expand.

On the data side, our latest upgrade increased data capacity. But currently, only about 25% of the chain’s data is utilized, meaning the chain’s “bandwidth” is only a quarter in use. If needed, we have room to increase data support by ten times or more.

But data alone isn’t enough. If the chain only stores data without computation, developers can write many things on-chain, but interpreting this data and combining it across applications becomes very difficult.

Therefore, we are also seriously thinking about how to increase Ethereum’s compute capacity by 1000 times. Some approaches are feasible, such as using ZK proofs to verify each block’s EVM execution; but there are many challenges:

• Not all logic can be easily scaled 1000 times;

• Many application developers will need to change how they write contracts—for example, a “scalable ERC-20 token” might differ significantly from today’s standard ERC-20.

At the same time, security remains a very practical concern. Ethereum is already a complex machine. If we pursue scalability by adding more complexity without ensuring sufficient security, we risk that in two or three years, a very clever attacker—possibly aided by AI—could find all vulnerabilities.

Duncan: It sounds like you’re always balancing “scaling” and “security” very carefully.

Vitalik: Yes, I can share a painful story. A few years ago in Shanghai, we were preparing to upgrade the network. A few hours before the upgrade, I was sleeping when I was suddenly awakened and told that a new attack vector had been discovered, requiring immediate action.

We locked ourselves in a room for three or four hours, figuring out how the attack worked and how to patch it. By around 8 a.m., we released a client patch, and the meeting started as usual. At that moment, everyone thought “we won.”

But two days later, a second wave of attacks appeared; five days later, a third. Later, there were even more incidents. Over about a month, the Ethereum chain was barely usable.

We still don’t know who the attacker was, but what’s certain is that they tested nearly all DoS attack surfaces across the main clients. That period was very painful, but it also taught us that if a system is complex enough, someone will test all boundary conditions—and that “someone” could very well be AI in the future.

That’s why we now design with extra caution, employing more formal verification and security tools to make the client and protocol “provably secure.”

AI and Quantum Computing: Threats Are Controllable, But Must Be Reconstructed in Advance

Duncan: You’ve also spoken multiple times about AI and quantum computing’s impact on blockchain security. Media often say things like “Who’s more vulnerable—Bitcoin or Ethereum?” Taking this opportunity, can you systematically share your views on AI and quantum as new variables?

Vitalik: I don’t think they will “destroy Ethereum,” but if we do nothing, the problems will become very serious.

I like to use an analogy: imagine a country that has never experienced rain, and people don’t even know what “rain” is. As a result, houses are designed without rain protection. When it finally starts raining heavily, about 5% of the houses will leak or collapse.

Scientists now tell you: in five or ten years, it will start raining, and they have some ideas on how to build rainproof houses. Theoretically, we know what to do, but the real challenge is: should we start renovating every house, school, and office building now? That’s a long and difficult social engineering process.

Quantum security is similar. We know which cryptography is quantum-resistant. I even wrote a hash-based quantum-safe signature scheme smart contract on GitHub back in 2017, but it’s quite hidden. The problem is efficiency:

• Current elliptic curve signatures are 64 bytes;

• Quantum-resistant signatures might be 2300 bytes.

If we do nothing and just replace all signature algorithms with quantum-resistant ones, the gas cost per transaction could jump from about 20k to 200k, reducing transaction capacity by roughly ten times.

So, we need smarter architectures, like signature aggregation: having block producers combine many large signatures into one proof using Stark, proving “all signatures are valid.” The chain only needs to store this proof, not all original signatures. Even with thousands of transactions, the raw signature data might be 3-4 MB, but the final on-chain proof would be only about 200 KB.

Theoretically, we know the route, but moving from “theory” to “safe operation on mainnet” involves a big leap. The next few years are about making these solutions real.

AI follows a similar logic. Our strategy is to heavily use formal verification, ensuring that key ZK code and client implementations meet strict security standards. L1 is relatively “centralized,” with a single protocol that’s prioritized and closely monitored, making upgrades easier. But at the application layer, fragmentation is high—each DApp has its dependencies and off-chain components, with inconsistent security models.

I believe that in the next three to five years, Ethereum’s ecosystem will need to invest heavily to ensure AI is used to help us find vulnerabilities and improve security, rather than being exploited by attackers to automate attack vectors.

Duncan: If all these quantum security solutions are added, will it increase gas costs? People might be very concerned about that.

Vitalik: If it’s just “directly replacing signature algorithms,” then yes, gas will go up, and transaction capacity will decrease. That’s why we need those aggregation and Stark proof methods I mentioned—to spread out the costs. Theoretically, these solutions can keep transaction costs within an acceptable range, but it requires a lot of engineering and protocol work.

A Message to Hong Kong and the Chinese-speaking Community: Don’t Just Make “Reprints of the Previous Ethereum”

Duncan: Today marks the official opening of ETH Hong Kong Community Hub. Your return has been warmly welcomed by the Chinese community here, and many are looking forward to your message.

Standing at this moment, what would you like to say to the builders here, to Hong Kong, and to the entire Chinese-speaking world building on Ethereum? What are your expectations for this Hub?

Vitalik: I think the past two years have been a great opportunity for the Ethereum ecosystem to “rethink many things.”

On one hand, ZK has transformed from a “very academic” concept into a technology that can be used in production environments; on the other hand, AI has reduced coding costs by about ten times. Now, many people who previously couldn’t code are using AI to build HTML pages or even simple smart contracts.

This means we have more tools to face a more complex world. Everything is accelerating, and demands are becoming more intricate. So I encourage everyone:

• Don’t just start from “what the current Ethereum ecosystem looks like,” but return to the core principles that first drew you to Ethereum;

• Ask yourself: what does the world really need now? Among these questions, which require Ethereum, and which can be addressed with AI, ZK, hardware security, and other technologies?

In places like Shenzhen, I see many open-source hardware projects and open-source AI initiatives. I think these are very promising directions: for example, using secure hardware to improve node and wallet security, or leveraging open-source AI to make on-chain interactions smarter and safer.

My biggest hope for this Community Hub is that it becomes a place where everyone can “start from zero” and rethink. Not just repeat the Ethereum stories from three or five years ago, but use new tools and technologies to create something entirely different from what we did before.

Duncan: Return to the original intention. Every good project starts from real problems and genuine user pain points, not just for the sake of technology.

Time’s almost up. Once again, thank you, Vitalik, for sharing, and thank you for your hopes for Hong Kong and the Chinese community.