The "2026 Hong Kong Web3 Carnival," jointly hosted by Wanxiang Blockchain Lab and HashKey Group, focused on cutting-edge forums on tokenization, agent economy, and privacy computing in the afternoon session after two days of intensive tracking. As the opening guest for the afternoon session, Wanxiang Blockchain Chairman and HashKey Group Chairman and CEO Xiao Feng delivered a speech titled "Innovations in Agent Economy Models: AI Tokens, Blockchain Tokens and the Integration of Privacy Computing," which evolved from the "Three Token Model" to the "Token Factory" and the ultimate form of agent economy. He systematically reviewed the gains and losses in the development of blockchain and consortium chains over the past decade, deeply analyzed the potential intelligent reconstruction of business models and financial service forms due to AI agent economy, zero-knowledge proofs, and fully homomorphic encryption, and provided a rich framework for building digital assets and new financial infrastructure in a technology environment focused on gaining trust and authorization.

Below is the edited transcript of Wanxiang Blockchain Chairman and HashKey Group Chairman and CEO Xiao Feng's speech at the "2026 Hong Kong Web3 Carnival" RWA Forum:

Xiao Feng: Distinguished guests, good afternoon!

I believe everyone is already feeling quite tired, given yesterday's "bombardment" and this morning's activities.

This afternoon's forum focuses on RWA, and we will first announce the "Token Economics White Paper for 2026." Looking back, HashKey began releasing the "Token Economics White Paper" in 2023, and the 2024 version specifically highlighted the "Three Token Model," which includes equity tokens, functional tokens, and NFTs (non-fungible tokens), representing the economic models of the three types of tokens. HashKey Group itself practices the "Three Token Model." Thus, we have our own functional tokens, and we also offer NFTs during certain events, while HashKey Group possesses equity, and our equity is listed as a structure on the Hong Kong Stock Exchange IPO.

We have found that, over the past decade, a basic foundational protocol may only require a single layer of tokens, but when targeting applications, customers, and B-end/C-end users, just having one layer of tokens, such as only functional tokens, may not establish a sound, complete, and effective economic incentive mechanism.

Functional tokens are an incentive mechanism aimed at the community, while equity tokens serve as an incentive mechanism for entrepreneurial teams and shareholders. This forms the 1.0 version of our "Token Economics White Paper."

Now entering the 3.0 version, we delve into how the agent economy brought about by AI Agents will integrate with Crypto and Blockchain.

Therefore, my opening speech today is titled: Innovations in Agent Economy Models, exploring how AI Tokens, Blockchain Tokens, and privacy computing technologies like ZK fully homomorphic encryption might contribute to innovative models in the agent economy.

Looking back at the characteristics of blockchain technology from a business innovation perspective involves two main points: first, the blockchain network is a trustless network, meaning no one needs to KYC (Know Your Customer) or sign contracts in advance, as it is an open network where trust and permission are not prerequisites. This is the primary technical characteristic of native blockchain commercial activities. However, simply having this technical feature is obviously insufficient; another characteristic of blockchain is that it is open and transparent. It is hard to imagine financial institutions, such as banks or others that have high demands for privacy protection and compliance, wanting to put their entire business processes on the blockchain, as they would face a huge problem related to "data transparency." "Data transparency" means that banks and financial institutions, including those dealing with highly sensitive medical data, cannot run directly on a digital native blockchain.

However, to date, AI has indeed brought immense economic vitality, especially after AI transitioned from large models to AI Agents, with discussions about how the AI agent economy could unlock more than ten times the commercial value in the future. Yet, the challenge remains in data transparency, and under such transparency, the AI agent economy clearly has significant flaws, which must be addressed with privacy computing technologies.

Reflecting on the development of blockchain technology, since the launch of Bitcoin's blockchain in 2009, its nearly 16 years of progress has proven the enormous commercial and economic value of blockchain technology. However, blockchain technologies represented by Bitcoin certainly possess the characteristic of public transparency, leading traditional banks and government regulatory bodies to propose the concept of "consortium chains/permissioned chains" in 2015. Consortium and permissioned chains emerged due to the inability of compliance-related businesses to operate on public ledgers because public chains are openly transparent.

However, while the emergence of permissioned chain technology, such as consortium chains, alleviates privacy concerns to a certain extent—since only permitted individuals can become blockchain nodes and share information within allowed boundaries—it still has notable limitations.

Over the past decade, with the idea of consortium chains being proposed, two significant consortium chain organizations have emerged: one is R3, consisting of global banks, and the other is IBM's consortium chain organization called Hyperledger. Ultimately, it has become apparent that over the past ten years, they have not produced any commercially viable applications. Thus, a view has emerged that consortium chains might not be blockchain, which was indeed true in the context of that time.

However, with the rise of tokenization and the tokenization of traditional financial assets, consortium chains have begun to re-emerge. We can almost understand that major global banks have set up their own internal permissioned chains. However, these internal permissioned chains may just represent single nodes, with the bank confirming the operation of the blockchain solely for itself—this is referred to as a "private chain."

Why is there a return?

A globally renowned bank offers blockchain tokenization services for its clients, such as deposit tokenization, without needing to resolve trust issues or requiring a third party as a blockchain node to provide trust endorsement. Since you have already trusted the bank and are its client, you only need to use tokenization within the bank’s account system to complete remittances from New York to Hong Kong in two minutes. If you do not tokenize your deposits, it might take two days to complete such a remittance, hence the revival of private chains.

However, the revival of private chains also presents a challenge: when clients from two banks conduct cross-bank foreign exchanges, other blockchains are still needed beyond the private chain. Thus, the exploration of consortium chains continues. We know that SWIFT, along with nine major global banks, is discussing how to use blockchain technology and deposit tokenization tools to resolve cross-bank, cross-border fund exchanges. Currently, discussions are ongoing, but it is clear that consortium chain technology is making a comeback, starting with the return of single-node private chains.

When it comes to cross-bank transactions, a significant issue still remains: how much internal bank data can you allow your partners to see? A new technology has emerged: privacy computing, zero-knowledge proofs, and fully homomorphic encryption, which can perform calculations while upholding privacy protection. The results computed can completely match those when using plain text. New technologies have existed for a while.

I remember during the Ethereum Devcon held in Shanghai in 2016, many speakers mentioned privacy computing technologies like zero-knowledge proofs and formal verification. Yet, to date, there has not been widespread application due to performance shortcomings, where costs, time, efficiency, and effectiveness have not supported commercial application.

However, according to my understanding, chips for fully homomorphic encryption may be released in the second half of this year, achieving a performance of approximately 1,000 transactions per second. This will clearly meet some commercial application scenarios, as some situations do not require real-time results. You may wait for 10 minutes, half an hour, or an hour. Under fully homomorphic encryption, the combination of blockchain tokens, AI tokens, zero-knowledge proofs, and privacy computing technologies will likely enable truly disruptive innovations in the business models of the agent economy. A complete disruptive innovation can only emerge when these three technologies are integrated.

Moreover, if we envision that zero-knowledge proof and fully homomorphic encryption algorithms are commercially viable in terms of efficiency, performance, and cost, then blockchain technology might undergo another return. At that point, there may genuinely be no need for private chains or consortium chains; all encrypted data could go onto public chains. Even running privacy protection on public chains may be sufficient to meet the highest global compliance requirements, which is something we may see happen in the next three to five years.

Yesterday, Ethereum founder Vitalik also spoke on this stage about Ethereum's developmental path over the next five years. He mentioned that Ethereum does not need to compete for who has the fastest chain, as long as it maintains decentralization and security, with no need for any improvements in performance. The "impossible triangle" of blockchain—decentralization, security, and performance—means that decentralization and security issues are to be resolved, leaving performance improvements to others through hardware acceleration, algorithm optimization, and Layer 2 or Layer 3 networks based on different application scenarios.

Let me provide an example to illustrate how these three technologies—AI Tokens, Blockchain Tokens, and zero-knowledge proofs and fully homomorphic encryption—create new business models.

For instance, consider a hospital, whose medical data is highly valuable but also subject to strict privacy protection requirements. According to the current Token economics model, think of a future where all commercial institutions become "Token Factories." If these medical data are supported by sufficiently advanced zero-knowledge proofs or fully homomorphic encryption privacy computing technologies, any hospital can convert its medical data into Tokens, becoming a factory for tokens. Anyone can utilize its data to derive specific technical features without gaining access to any individual privacy-protected data.

The confluence of these three technologies might disrupt traditional business models and represent the ultimate form of the agent economy. If only AI Tokens and privacy computing technologies exist, this business model also makes sense. A hospital could innovate without the support of blockchain Tokens, but its commercial field cannot expand globally.

We know that all digital technologies essentially provide global services; digital products and services have always been global. If there is no blockchain support, even if a hospital uses homomorphic encryption for its data, anyone wanting to utilize hospital data must find that hospital offline, negotiate, sign an agreement, and make payments through a bank, which is not the approach of a Token Factory.

What should the Token Factory approach be? It should leverage the trustless, permissionless characteristics of blockchain technology to convert all that hospital's data into tokens, into AI Tokens, and open it up to global users without the need for permission or trust. Anyone with a need can access the data at any time, similar to how we use the Bitcoin or Ethereum networks, 24/7, without signing agreements or undergoing KYC onboarding. Calling data will consume Tokens, thus paying the hospital, transforming it into a Token Factory. The combination of these three elements will represent the final outcome and expression of the agent economy.

If any individual's privacy data—such as years of medical check-up data—can be accessed in a permissionless, trustless manner after encryption, they can send requests online to global insurance companies. My data is here, encrypted; you have your actuarial models and can come to me, using your actuarial models under homomorphic encryption to calculate my data and provide me with a personalized cost-effective insurance plan.

Thus, the means of financial services may be entirely transformed under these three integrated technologies. The way financial services work in the agent economy changes fundamentally—there will no longer be insurance brokers or intermediaries. You will not belong to any financial institution; you are not its client, yet you also belong to all financial institutions as your data is there. You can search the entire network for the "optimal solution" in a trustless, permissionless manner.

In particular, during this period, I have been exploring information about privacy computing, zero-knowledge proofs, and fully homomorphic encryption technology. Based on this information, combined with the trustless, permissionless commercial characteristics of blockchain technology, and the AI Tokens brought by AI Agents, I conclude the combination and integration of these three.

I want to emphasize that there is now a misunderstanding; AI Tokens are referred to as the currency unit of AI intelligent economies. This is not the case; AI Tokens are the production materials of the agent economy, not its currency unit. From power, through chip computational power, large models, algorithms, to applications, this is the five-layer structure of Token economics proposed by Nvidia's Jensen Huang. This five-layer structure describes the production process of agent outcomes, depicting how to generate products or services for agents. However, to obtain, purchase, or enjoy products or services, you must pay with currency, and this currency must be digital currency because the currency used by AI Agents cannot be traditional currency. It must be programmable, subdivided currency capable of real-time settlement.

Because when an AI Agent calls another's API, for example, when an AI Agent calls a hospital's data Token, it cannot go through a banking account to say, "I will pay you now, and tomorrow we can check what services you provide," I must confirm that the money arrives instantly. I might only consume a few cents, a dime, a dollar with each call, regardless of the amount; the cost of the current banking payment system cannot support tiny payments for an API data call, hence digital currency is the "blood" of the intelligent economy.

Moreover, the agent economy will also form another digital asset: in the agent economy, digital currency and digital assets must not only be transformed and tokenized but also whether it is central bank CBDCs, bank deposit tokenization, or stablecoins issued by commercial institutions—all essentially do one thing: tokenize currency/funds. The reason for tokenizing currency or funds is that non-tokenized currency cannot be used by machines since the currency used by machines must be machine-readable and understandable. An AI Agent can only use currency that is programmable. Currently, US dollars, Hong Kong dollars, and Chinese yuan are not programmable; they must be tokenized before they can be programmed, identified, and used by machines.

To summarize.

The AI agent economy is a trustless business model. Trustless and permissionless interactions will significantly reduce commercial costs. As we are aware, a large system is operational in society to build trust in commerce. How extensive is this system? So extensive that even prisons exist to ensure commercial trust, as violating trust and violating the law can lead to imprisonment. Accountants, lawyers, courts, and police are involved; an entire societal structure exists to ensure that participants in commerce do not breach trust or face costs associated with breaching trust. But with a trustless, permissionless business, the costs of maintaining that vast system are no longer necessary.

A new asset class will emerge from this new commercial model. Within the trustless, permissionless framework of the AI agent economy, the new asset class is referred to as "native digital assets." Therefore, crypto encompasses native digital assets like Bitcoin and Ethereum, as well as twin digital assets, such as various tokenized financial assets that already exist in reality, with the Digital Twin merely represented on chain. AI will also generate native digital assets and twin digital assets. AI's twin digital assets can be encapsulated through tokenization; native digital assets originating from AI will surely emerge in the coming years, building entirely new business models, which will also require a brand new financial service system suitable for machine usage, not solely human-centered financial services.

So far, our well-functioning financial service system is designed for people. In the future, a new financial service system and even capital market system will undoubtedly emerge built for AI, machines, and AI-native digital assets.

This concludes my presentation today. Thank you all!

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。