肖峰香港演讲：隐私计算赋能，所有商业机构都将成为“代币工厂”演讲者 |肖峰，万向区块链董事长、HashKey集团董事长兼首席执行官编译|吴区块链来源：2026香港Web3节TL;DR· 三通证模型：HashKey提出的权益通证、实用通证、NFT组成的通证经济激励框架已在集团内全面落地，为2.0版本奠定基础。 · 代理经济创新路径：AI通证（作为生产资料）、区块链通证和ZK/全同态加密（FHE）隐私计算的融合，实现颠覆性的商业模式· 代币工厂模式：商业机构对数据进行代币化，通过区块链功能实现全球范围内实时小额支付，从而彻底打破线下谈判和地域限制的障碍。 · 隐私计算的商业突破：全同态加密（FHE）芯片下半年将达到每秒 1,000 笔交易（TPS）的性能。未来3到5年，公链将
能够满足最高的合规要求，推动从私有链、联盟链到公有链的全面回归。·原生数字资产与新金融体系：AI代理经济下，全新的原生数字资产将会出现。这就需要专门为机器设计的可编程货币、支付网络和资本市场系统，这将大幅降低商业成本。《通证经济学白皮书》和三通证模型各位嘉宾下午好！今天下午的主题论坛聚焦现实世界资产（RWA），拉开分会的序幕，我们将发布2026年通证经济学白皮书。回顾一下，HashKey从2023年开始发布通证经济学白皮书，在2024年版中，我们特别提出了“三代币模型”
Xiao Feng’s Hong Kong Speech: Empowered by Privacy Computing, All Commercial Institutions Will Become “Token Factories”
Speaker | Xiao Feng, Chairman of Wanxiang Blockchain, Chairman and CEO of HashKey Group
Compiled by | Wu Blockchain
Source: 2026 Hong Kong Web3 Festival
TL;DR
· Three-Token Model: The tokenomics incentive framework comprising equity tokens, utility tokens, and NFTs proposed by HashKey has been fully implemented within the group, laying the foundation for version 2.0.
· Agent Economy Innovation Path: The integration of AI Tokens (as means of production), Blockchain Tokens, and ZK/Fully Homomorphic Encryption (FHE) privacy computing enables a disruptive business model that is trustless and permissionless.
· Token Factory Model: Commercial institutions tokenize data, making it globally accessible via blockchain features with real-time micropayments, thereby completely breaking down the barriers of offline negotiations and geographical restrictions.
· Commercial Breakthroughs in Privacy Computing: Fully Homomorphic Encryption (FHE) chips will reach a performance of 1,000 transactions per second (TPS) in the second half of the year. In the next 3 to 5 years, public chains will be able to meet the highest compliance requirements, driving a comprehensive return from private and consortium chains to public chains.
· Native Digital Assets and the New Financial System: Under the AI agent economy, entirely new native digital assets will emerge. This requires programmable money, payment networks, and capital market systems designed specifically for machines, which will significantly reduce business costs.
“Tokenomics Whitepaper” and the Three-Token Model
Good afternoon, distinguished guests!
This afternoon’s thematic forum is focused on Real-World Assets (RWA), and to open the session, we will release the 2026 Tokenomics Whitepaper.
Looking back, HashKey began publishing the Tokenomics Whitepaper in 2023. In the 2024 edition, we specifically proposed the “Three-Token Model,” which refers to the economic model encompassing equity tokens, utility tokens, and Non-Fungible Tokens (NFTs). HashKey Group itself has been putting the “Three-Token Model” into practice. We have our own utility token, of course, NFTs are issued at certain events, and at the same time, HashKey Group has equity, structured towards an eventual IPO on the Hong Kong Stock Exchange.
We found that over the past decade or so, for the most fundamental base protocols, a single layer of tokens might have been sufficient. However, when it comes to applications, targeting clients, and catering to both B2B and B2C users, a single layer of tokens — such as relying solely on utility tokens — is inadequate for establishing a sound, comprehensive, and effective economic incentive mechanism. Utility tokens serve as an incentive mechanism for the community, while equity tokens provide incentives for the entrepreneurial team and shareholders. This was detailed in version 1.0 of our Tokenomics Whitepaper.
Moving Towards Version 3.0: Innovation in the Agent Economy Model
Now entering version 3.0, we are focusing on how the agent economy brought about by AI Agents integrates with Crypto and Blockchain. Therefore, the title of my opening speech today is: Innovation in the Agent Economy Model — The Potential Disruptive Innovation in the Agent Economy Model Brought by the Convergence of AI Tokens, Blockchain Tokens, and Privacy Computing Technologies like ZK and FHE.
Commercial Characteristics of Blockchain Technology and Privacy Challenges
Looking back at blockchain technology from the perspective of commercial innovation, its main characteristics lie in two aspects: First, the blockchain network is trustless. That means anyone can join without prior KYC or signing a contract; it is an open network available to anyone. Being trustless and permissionless are the primary technical characteristics of native blockchain commercial activities. However, having only this technical characteristic is obviously not enough. Another characteristic of blockchain is its openness and transparency. It’s very difficult to imagine banks or other financial institutions that have extremely high requirements for privacy protection and compliance putting their entire business processes entirely on-chain. Doing so would lead to a massive problem: “data streaking” (data exposure). Because of this “data streaking,” it is highly unlikely that banks and other financial institutions — including medical data that requires stringent privacy protection — could operate directly on a digitally native public blockchain.
However, up to today, AI has indeed brought tremendous economic vitality. Especially as AI evolves from Large Language Models (LLMs) to AI Agents, everyone is now talking about how the AI agent economy can unlock more than 10 times the commercial value in the future. But the problem encountered is data transparency. Under data transparency, the AI agent economy obviously has a huge flaw, and to solve this flaw, we must rely on privacy computing technologies.
The Evolution of Blockchain: From Public Chains to Consortium Chains and the Eventual Return
Reviewing the development process of blockchain technology, since the launch of the Bitcoin blockchain in 2009, nearly 16 years of development have proven the immense commercial and economic value of blockchain technology. However, blockchain technology represented by Bitcoin indeed possesses the technical characteristic of being open and transparent. Therefore, around 2015, traditional banks and government regulatory agencies around the world began proposing the concept of “consortium chains/permissioned chains.” The very reason for consortium chains and permissioned chains is that the open and transparent data on public chains makes it impossible for many compliant businesses to run on distributed ledgers or public chains.
However, while the emergence of permissioned chain technologies like consortium chains provided a certain degree of mitigation for privacy protection — because only permissioned individuals can become nodes on a consortium chain, and they can only access each other’s information and data within permissioned scopes — it also has major flaws. Over the past decade, along with the proposal of the consortium chain concept, we witnessed two very large consortium chain organizations: R3 among global banks and IBM’s Hyperledger. Ultimately, after ten years, we found that they hadn’t produced any commercially viable applications. This led to a view at the time: perhaps a consortium chain is not a true blockchain, which was indeed true in that context.
But now, with the advent of tokenization, and specifically the tokenization of traditional financial assets, consortium chains are making a comeback. We can almost ascertain that major global banks are already running their own internal permissioned chains. It’s just that the permissioned chain running internally at a bank might be a single-node setup, serving only to confirm blockchain operations within the bank itself. This type of single-node permissioned chain is known as a “private chain.”
Why can private chains make a comeback? When a world-renowned bank provides blockchain-based tokenization services for its clients, such as deposit tokenization, it doesn’t need to solve the trust problem. It doesn’t require a third party to act as a blockchain node to endorse trust because the clients already trust the bank and are already its clients. It’s simply using tokenization within the bank’s account system to complete a remittance from New York to Hong Kong, reducing the cross-border remittance time to just 2 minutes. Without deposit tokenization, such a remittance might take 2 days. Therefore, private chains are the first to revive. But the revival of private chains also has problems. When clients of two different banks engage in cross-bank, cross-border remittances, other blockchains on top of the private chains are still needed. Consequently, consortium chains are being discussed again. We know that SWIFT, in collaboration with 9 of the world’s major banks, is discussing how to use blockchain technology and deposit tokenization tools to solve cross-bank, cross-border capital remittances. Although still under discussion, we have already seen consortium chain technology return, starting primarily as single-node private chains.
Breakthroughs in Privacy Computing Technologies and the Convergence of Three Technologies
If cross-bank operations are involved, a huge problem remains: how much internal bank data can you allow your partners to see? Now, new technologies are beginning to emerge: privacy computing, Zero-Knowledge Proofs (ZK), and Fully Homomorphic Encryption (FHE). After ensuring privacy protection, these technologies can still perform computations, and the computation results are exactly the same as if they were performed on plaintext. These new technologies have existed for a while. I remember when the Ethereum Devcon was held in Shanghai in 2016, many speakers mentioned privacy computing technologies like Zero-Knowledge Proofs and formal verification. But to date, we haven’t seen widespread application because the performance is insufficient. In terms of cost, time, efficiency, and commercial viability, they cannot yet support commercial applications. However, based on the information I have now, FHE chips might be launched in the second half of this year, reaching a performance of roughly 1,000 transactions per second. This can obviously satisfy a portion of commercial application scenarios because some scenarios do not require real-time computation results; waiting 10 minutes, half an hour, or even an hour might be acceptable.
Under Fully Homomorphic Encryption, the convergence of Blockchain Tokens, AI Tokens, and privacy computing technologies like ZK and FHE is what can truly drive completely disruptive innovation in the business models of the agent economy. These three technologies must be superimposed to enable this disruptive innovation. Furthermore, if we assume that the efficiency, performance, and cost of Zero-Knowledge Proofs and Fully Homomorphic Encryption algorithms eventually become sufficient to support commercial use, then blockchain technology might experience another return. At that point, private chains or consortium chains might truly be unnecessary; all data, once encrypted, could be uploaded to the public chain. Even running on a public chain, the privacy protection would be robust enough to meet the world’s highest compliance requirements today. This is exactly what will happen in the next 3 or 5 years.
Yesterday, Ethereum founder Vitalik Buterin also spoke on this stage about Ethereum’s development path over the next five years. He mentioned that Ethereum does not need to compete to be the fastest chain; it only needs to secure decentralization and safety. He doesn’t need to make any performance improvements. Because of the blockchain “impossible triangle” — decentralization, security, and scalability (performance) — Ethereum solves for decentralization and security, while delegating the performance problem to others. This means leaving it to hardware acceleration, algorithm optimization, and allowing L2 and L3 to determine their own consensus mechanisms based on different application scenarios.
Token Factory Model: Medical Data Case and Global Application
To give an example illustrating how AI Tokens, blockchain Tokens, combined with cryptographic privacy protection like Zero-Knowledge Proofs (ZK) and Fully Homomorphic Encryption (FHE), can create new business models: take a hospital, for instance. Its medical data is highly valuable but also has extremely strict privacy protection requirements. Based on current tokenomics models, we envision that all commercial institutions in the future will become “Token Factories.” If this medical data is empowered by sufficiently robust privacy computing technologies like ZK or FHE, any hospital could turn its medical data into tokens, becoming a Token Factory. Anyone could call its data to compute specific needed technical features, but at the same time, it would be impossible to access any individual’s privacy-protected data. Only with the integration of these three technologies can there be a disruption of traditional business models, representing the ultimate form of the agent economy.
If there were only AI Tokens plus privacy computing technology, this business model would, of course, still be viable. A hospital could innovate its business model without the empowerment of blockchain Tokens, but its commercial scope could not expand globally. We know that all digital technologies essentially provide global services; digital products and services have always been globalized. Without blockchain empowerment, even if a hospital’s data used homomorphic encryption, anyone wanting to use that hospital’s data would need to find the hospital offline, negotiate, sign agreements, and pay through a bank. This is not the approach of a Token Factory.
The approach of a Token Factory should be: utilizing the permissionless and trustless technical characteristics of blockchain to convert all of the hospital’s data into tokens — into AI Tokens — and open them up to global users in a permissionless and trustless manner. Anyone with a need could call the data 24/7, without signing agreements or undergoing KYC onboarding, just as we currently use the Bitcoin or Ethereum networks. You consume tokens when you call the data, thereby paying the hospital, and the hospital thus becomes a Token Factory. The combination of these three elements is the ultimate manifestation and final outcome of the agent economy.
Imagine if all of an individual’s private data — years of past physical examination and medical data — after being encrypted, could use a permissionless and trustless business model to issue requests to global insurance companies online and on-chain: “My encrypted data is here; you have your actuarial models. You can come to me, use your actuarial models under homomorphic encryption to compute my data, and then provide me with a personalized insurance plan that suits me best, with optimal cost and effectiveness.” Therefore, perhaps under the integration of these three technologies, the method of financial services in the agent economy will be completely transformed. There will no longer be insurance brokers or intermediaries, and you will not belong to any single financial institution; you are not their exclusive client, but you are a client of all financial institutions. Because your data is there, you can seek the “optimal solution” for yourself across the entire network in a trustless and permissionless manner.
The Essence of AI Tokens and the Payment System of the Agent Economy
This is a conclusion I’ve drawn recently from exploring the development of privacy computing, Zero-Knowledge Proofs, and Fully Homomorphic Encryption technologies, combined with the trustless and permissionless commercial characteristics of blockchain technology, and the AI Agent economy brought about by AI Tokens. I want to particularly emphasize that there is currently a misunderstanding that refers to AI Tokens as the monetary unit of the AI intelligent economy. This is not the case; AI Tokens are the means of production of the agent economy, not its monetary unit. From electricity, to chip computing power, to large models, to algorithms, to applications — this is the five-layer structure of Token economics proposed by Jensen Huang of Nvidia. This five-layer structure actually describes the production process of agent outcomes; it depicts how the products or services of agents are generated. However, to acquire, purchase, or enjoy a product or service, you need to pay with money, and this money can only be digital currency. Because the currency an AI Agent can use is not the currency convenient for human use; it must be programmable money, divisible money, and money capable of real-time settlement.
Because when an AI Agent calls an API provided by someone else — for example, when an AI Agent calls a hospital’s data Token — it cannot use the traditional bank account system to say, “I’ll pay you first, and once the money arrives tomorrow, we’ll see what service you provide.” It must confirm the arrival of funds in real-time, and perhaps each call only costs a few cents, a dime, or a dollar. Whether it’s a dollar, a dime, or a cent, the cost of current banking payment systems cannot support an AI Agent making such a micro-payment to call an API. Therefore, digital currency is the “blood” within it; it is the “blood” of the agent economy.
Of course, the agent economy will also form another type of digital asset. In the agent economy, it is not just money that needs to be transformed and tokenized into digital currencies and digital assets. Whether it’s a central bank’s CBDC, a bank’s deposit tokenization, or stablecoins issued by commercial institutions, they are actually all doing the same thing: tokenizing money or funds. The reason money or funds must be tokenized is that if they are not, they cannot be used by machines, because the currency machines use must be readable and understandable by machines. For an AI Agent to use this currency, it must be programmable. The current US Dollar, Hong Kong Dollar, and RMB are not programmable; they must be tokenized before they can become programmable, recognized by machines, and used by machines.
Conclusion: New Business Models, New Assets, and the Future Financial System
To summarize, the AI agent economy is trustless commerce. Being trustless and permissionless will significantly reduce business costs. As we know, to build trust in commerce, the entire society operates a massive system — a system so massive that even prisons are prepared to ensure commercial credit. Once you breach trust and violate criminal law, you might go to jail. Accountants, lawyers, courts, police departments — the entire society runs a massive system to ensure that people do not breach trust in commerce, or that the cost makes breaching trust prohibitive. But if commerce is trustless and permissionless, you no longer need everyone to share the operating costs of such a massive system.
On the foundation of such a new business model, new asset classes will be produced. In this new commercial system of AI agents — which is trustless, permissionless, and vastly cheaper — the new asset class is called “native digital assets.” Thus, Crypto has native digital assets, like Bitcoin and Ethereum; it also has twin digital assets (digital twins), such as various tokenized financial assets, which already exist in the real world but simply have a digital twin on the chain. AI will have these too: AI native digital assets and AI twin digital assets. AI twin digital assets can be summarized by Tokenization, or asset tokenization; AI native digital assets are bound to emerge in the coming years. AI native digital assets will build completely new business models and will simultaneously require a completely new financial service system, because it will be a financial service designed for machines to use, not for humans. The financial service system that has operated very well up to now was designed for humans. In the future, a new financial service system, and even a capital market system, specifically established for AI, for machines, and for AI native digital assets, will definitely emerge.
That concludes my sharing for today. Thank you, everyone!
Follow us
Twitter: https://twitter.com/WuBlockchain
Telegram: https://t.me/wublockchainenglish