Dialogue with Cysic Co-founder: ZK DePIN Computing Power Network Launches Testnet, Aiming for Large-scale Adoption by 2025

Author: Deep Tide TechFlow

Having fallen into the splendid and illusory rabbit hole of Web3 for some time, the public seems to habitually overlook one point: essentially, we live in a world supported by hardware, whether it is the demand for VR devices in the metaverse, the CPU and GPU consumption behind AI, or the ZK track regarded by Vitalik Buterin as "the future of Ethereum."

Although ZK has rigorous mathematical logic and significant advantages in security, privacy, and scalability, there is currently an efficiency issue in ZK proof generation. In the absence of breakthrough progress at the algorithmic level, ZK hardware acceleration has become the most direct and effective solution to empower ZK development. Vitalik also mentioned during his attendance at the Hong Kong Web3 Carnival in April this year that ZK hardware acceleration helps elevate Ethereum's efficiency and security to a new level.

As the potential of the ZK hardware acceleration track is increasingly recognized, the ZK hardware acceleration project Cysic, which will launch its testnet in mid-July, has attracted widespread attention: this testnet launch will see Cysic conduct a series of community incentive activities in phases. At this stage, Cysic will invite users to become validators through a whitelist mechanism, participate in network testing, and receive corresponding rewards; later, Cysic will also open some prover roles to the community.

As one of the earliest leading projects in the ZK hardware acceleration track, Cysic has previously completed $6 million in seed funding and $12 million in Pre-A funding, attracting participation from well-known industry VCs including Polychain Capital, Hashkey Capital, OKX Ventures, ABCDE, and receiving support from angel investors such as DAO5 partner, Celestia's first investor George Lambeth, and former Binance Labs executive director Ken Li.

Seizing the opportunity of the testnet launch, we had an in-depth exchange with Cysic co-founder Leo Fan, discussing Cysic's insights on ZK and the hardware acceleration track, project advantages and implementation logic, and the outlook for large-scale application of ZK technology. During the conversation, Leo Fan stated: ZK hardware acceleration will greatly empower Ethereum, the crypto industry, and even the real world, and with the development of ZKVM and the upcoming shipment of Cysic ASIC hardware, we have reason to optimistically predict that ZK technology will achieve true large-scale adoption by 2025.

In this issue, let us delve into Cysic's past, present, and future, along with Cysic co-founder Leo Fan's thoughts and feelings on the ZK hardware acceleration track.

Key Points

• With the arrival of the testnet in mid-July, a series of incentive activities are also in preparation: Cysic will invite users to become validators through a whitelist mechanism, participate in testing to win rewards, and later we will also open some prover roles to the community. For B-end users, Cysic aims to provide efficient, low-cost, and instant ZK services for ZK project parties, while professional miners can also connect to the Cysic network for more efficient use of hardware devices. For C-end users, ordinary laptops can also connect to the Cysic network to become validators.

• As a leading project in the ZK hardware acceleration track, Cysic's focus can be summarized in three aspects: efficiency, achieving real-time proof generation through self-developed chips; universality, providing acceleration services for almost all different algorithm types of ZK on the market; decentralization, building a ZK DePIN Network to provide services to the industry in a decentralized manner.

• The advantages of ZK lie in: not needing to trust specific individuals, but establishing trust through mathematical logic. Additionally, compared to other scalability solutions, Ethereum's entire architecture is more friendly to ZK development, and the rigorous mathematical logic of ZK also provides higher assurance for Ethereum's security and fairness.

• In Cysic, project parties have the autonomy to choose suitable Provers based on their needs, further enhancing the degree of decentralization.

• In terms of cost-effectiveness and performance-to-power ratio, ASICs outperform significantly, but due to potential issues like tape-out errors during the design and production process, there is caution around ASIC design and manufacturing.

• Cysic adopts a dual-token economic model: the Native Token serves purposes such as gas fees and rewarding ecosystem contributors, while veToken can participate in governance decisions regarding network computing power scheduling.

• With the development of ZKVM and the shipment of Cysic ASIC hardware, ZK technology may achieve true large-scale adoption by 2025.

The fastest ZK hardware acceleration service in the industry, Cysic testnet is coming soon

Deep Tide TechFlow: First, could you please introduce yourself?

Leo Fan:

Hello everyone, I am Leo Fan, co-founder of Cysic.

Let me first share my educational background, which mainly revolves around cryptography and computer security: I studied in the mathematics base class at Sichuan University for my undergraduate degree, and from 2010 to 2014, I continued my master's studies at the National Key Laboratory of Information Security in Beijing. After that, I entered Cornell University to pursue a PhD in cryptography and information security, graduating in 2019.

Now, let me share my work experience: during my master's studies in Beijing, I came into contact with Bitcoin and participated in early Bitcoin mining. Later, during my PhD, I had several internships at companies like IBM and Yahoo. After graduation, I first joined the National Institute of Standards and Technology (NIST) in the U.S. to work on cryptographic standardization, then joined the Layer 1 project Algorand, founded by ZK technology inventor, Turing Award winner, and MIT professor Silvio Micali. At Algorand, I was mainly responsible for designing and implementing Algorand State Proofs, aiming to bring trustless interoperability to the Algorand ecosystem.

Until 2022, I noticed that the time for ZK proof generation was very long. Initially, I wanted to optimize it through algorithms and software, but the results were not significant, so I thought of accelerating proof generation through hardware methods, which led to the birth of Cysic, aiming to achieve real-time proof generation through hardware acceleration.

Deep Tide TechFlow: The Cysic testnet will launch in mid-July. Could you please share how project parties, users, or developers can participate in the Cysic ecosystem from both B-end and C-end perspectives?

Leo Fan:

From the B-end perspective, Cysic aims to provide efficient, low-cost, and instant ZK services for ZK project parties. Currently, Cysic has launched an Early Access Program (EAP) to provide enterprise-level end-to-end hardware acceleration solutions for ZK projects with computing power needs. Cysic supports a range of multi-verification systems, including Halo2, STARK, Groth16, Plonky2, ZAMA, GKR, and RapidSnark. Our powerful network is supported by over 10,000 GPU cards, ensuring unparalleled performance and reliability. Currently, 16 projects, including Scroll, Light protocol, Risc Zero, Inference Labs, Ion Protocol, Aerius Labs, and TalusNetworkEAP, have joined the EAP program, and early members will enjoy a series of ecological benefits in the future.

Additionally, the B-end can also refer to some professional miners who have a large number of graphics cards. Besides connecting to other networks like AI DePIN, they can also connect to the Cysic network in the future. Once connected, their graphics cards will be running tasks most of the time, meaning hardware will enter a more efficient usage phase. In the later stages of the testnet, we will also open prover roles to the community, allowing users who meet certain hardware conditions to apply to become provers, contributing to the mainnet launch while earning testing rewards.

From the C-end user perspective, this may refer to relatively novice users who may not have powerful hardware. They can also connect to the Cysic network through ordinary laptops to play the role of validators, thus contributing to the ZK ecosystem while earning rewards. The Cysic testnet will launch in mid-July, and related testing incentive activities are also in preparation: at that time, we will invite users to become validators through a whitelist mechanism, participate in network testing, and receive corresponding rewards.

Deep Tide TechFlow: If you had to describe what Cysic is doing in three words, how would you describe it?

Leo Fan:

We believe that after breaking through the efficiency bottleneck, whether in the crypto industry or the Web2 real world, ZK technology will usher in large-scale adoption, and Cysic aims to provide full hardware acceleration services, including FPGA, GPU, and ASIC, to achieve real-time generation of ZK proofs. By building the ZK DePIN network Cysic Network, we connect ZK project parties, computing power providers, and community validators, ultimately providing an efficient, cost-effective, and decentralized computing + validation network for the entire industry. Overall, Cysic will play an indispensable role in accelerating the large-scale adoption of ZK.

Returning to the question, I think what Cysic is currently doing can be described with the following three words:

Efficiency: Cysic aims to achieve real-time proof generation through self-developed chips. Although we are still in the stage of providing ZK acceleration services through GPUs, we are currently the fastest in the industry, and as the project progresses and ASICs are realized, we will be even faster.

Universality: There are various algorithms within the ZK domain, and there are differences between different algorithms. Cysic aims to provide ZK acceleration services that support all algorithms for the industry, further expanding our service scope and enhancing usability.

Decentralization: Cysic is currently building the Cysic Network, which is a ZK DePIN Network. We hope to provide services for the entire ecosystem through decentralized Provers and Verifiers, with participants contributing to ZK and receiving rewards.

Deep Tide TechFlow: Currently, ZK hardware acceleration solutions mainly include CPU, GPU, FPGA, and ASIC. What are the differences between these methods? Which method does Cysic prefer for ZK hardware acceleration?

Leo Fan:

Cysic has a significant voice on this issue, as there is a competition called ZPrize for ZK hardware acceleration. Cysic missed the first ZPrize, but we are one of the designers and judges for one of the tracks in this year's competition.

In the entire ZK or hardware field, there are three important metrics:

First is Performance Per Dollar: how much it costs to acquire the corresponding hardware and how much computing power you get for that cost. Here, since CPUs do not perform as well as the other options, our discussion of hardware acceleration mainly focuses on GPUs, FPGAs, and ASICs. From a cost-effectiveness perspective: ASIC > GPU > FPGA, provided that ASICs are produced in sufficient quantities, which the current ZK can meet.

The second metric is Performance Per Watt: the energy consumption required to run these hardware solutions. In this regard, ASICs still lead significantly, with GPUs slightly outperforming FPGAs.

The last metric is Go to Market Time: we know that ASICs need to have chips designed, and tape-out is an important step in the ASIC chip design process, where the generated masks are sent to the wafer fab for chip manufacturing. If there are mistakes during the tape-out process, it can lead to high costs, so there is caution around ASIC design and manufacturing.

In the short term, Cysic is also following this path. From the end of last year until early 2025, we will provide acceleration services through GPUs, but our ASICs will likely meet everyone in the second or third quarter of next year, at which point we will shift to primarily providing ASIC services. This is an inevitable process, as Cysic is committed to achieving real-time ZK proof generation. For the currently mainstream ZK Circuits, we hope to achieve proof times of 1 to 5 seconds, and to reach this goal, only ASICs can accomplish it.

Deep Tide TechFlow: How does Cysic achieve real-time proof generation? What changes or innovations will this bring to ZK, Ethereum, and the entire crypto industry?

Leo Fan:

I believe this will definitely bring significant innovation. Let's take the combination of ZK and Machine Learning as an example. Currently, if you want to run a Machine Learning Circuit with ZK, it will be very time-consuming: running a GPT2 without ZK takes about one second, but if you want to prove a GPT2 Circuit with ZK, even the fastest solution takes several minutes, and using a CPU could take several hours. Such long times are completely unacceptable for user experience.

Cysic's ASIC chips will enable real-time ZK proof generation, which not only promotes the integration of ZK with machine learning technology but also provides a great user experience. Of course, the combination of Crypto and AI is also very close now, so this example is a very Crypto Native one.

Additionally, I think ZK acceleration will greatly empower Ethereum and the entire crypto industry. The biggest change is that tasks that previously took several minutes can now be completed in seconds, allowing for faster confirmations of more transactions and cross-chain transactions on L2, and this confirmation is based on mathematical logic rather than the "innocent until proven guilty" assumption like OP, which is a significant advancement for the crypto industry.

Achieving real-time ZK proof generation in a decentralized manner, dual-token model incentivizing the ZK DePIN network

Deep Tide TechFlow: In past solutions, it has always been difficult to balance efficiency and decentralization, while Cysic has achieved decentralization of validators/provers while realizing real-time proof generation. Could you briefly introduce how Cysic achieves this? How does the PoC consensus adopted by Cysic further enhance decentralization?

Leo Fan:

As mentioned earlier, there are two important roles in the ZK DePIN network Cysic Network built by Cysic: provers and validators.

Validators are responsible for verifying the proofs generated by provers, and the decentralization of validators is relatively easy, as even mobile phones, laptops, and tablets can participate in verifying a ZK proof, which has very low hardware requirements, and the verification time is usually less than one second.

The decentralization of provers is somewhat challenging in the early stages (i.e., the GPU phase), because to provide a good user experience, meaning relatively short proof generation times, requires provers to meet higher hardware requirements, which can be a barrier for many. However, when Cysic's hardware begins to ship, this situation will be greatly alleviated, as compared to the current mainstream GPU acceleration, Cysic's hardware can achieve up to ten times the efficiency in ZK acceleration, and is relatively inexpensive, allowing everyone to purchase Cysic's hardware at a lower cost to reach the level of professional miners.

From a software perspective, Cysic is currently developing a ZK DePIN network, and we have designed a PoC (Proof of Compute) mechanism to better fit the ZK DePIN scenario. PoC is actually more like a combination of Proof of Stake and Proof of Work. For each epoch, which is about 100 blocks, a batch of nodes is selected based on the computing power scale of the provers, and then based on the staking weight of this batch of nodes, it is decided who will propose the blocks in this epoch. This not only ensures that the network remains stable but also further enhances decentralization.

Deep Tide TechFlow: We noticed that Cysic has adopted a dual-token model combining utility tokens and governance tokens. Could you introduce the advantages of this dual-token design and how Cysic achieves effective governance and incentives for the ecosystem through it?

Leo Fan:

In the dual-token model, one is the Native Token, and the other is the veToken.

The Native Token is mainly used for gas fees, rewarding ecosystem contributors, etc., but if participants want to participate in governance and decide on network computing power allocation, they need to stake Cysic's Token to obtain corresponding voting rights.

We know that there are many ZK projects, and many of them use different algorithms, so how to schedule computing power in the Cysic network becomes crucial. If the computing power is well scheduled, it will bring high returns to the Cysic network, thereby better empowering network participants. Having voting rights is equivalent to having the right to participate in computing power scheduling. For example, if a certain project performs very well recently, should we provide more computing power? Through this mechanism, it not only helps capture the value of the entire ZK ecosystem but also allows Cysic's Token to circulate better within the entire ZK ecosystem.

Deep Tide TechFlow: The hardware requirements for ZK hardware acceleration services easily evoke thoughts of the DePIN track. What extensions does Cysic's ecosystem construction have in DePIN?

Leo Fan:

We hope to see more hardware join the Cysic network. It is worth noting that the hardware in the Cysic network is not just idling; it is genuinely running services and creating value through proof generation and verification. This is our biggest difference compared to many AI and DePIN projects. Currently, the Cysic network is in the internal testing phase, and hundreds of servers are continuously running tasks.

Additionally, the Cysic network also touches on traditional users. Due to ZK's advantages in privacy computing, we will also provide better services for traditional users through the DePIN network.

Deep Tide TechFlow: Whether at last year's EDCON Black Mountain Conference or this year's Hong Kong Web3 Carnival, Ethereum co-founder Vitalik Buterin has clearly expressed the view that "ZK is the future of Ethereum." Why does Vitalik hold ZK in such high regard? What do you think are the advantages of ZK technology compared to other scalability solutions and its empowerment for Ethereum?

Leo Fan:

I think the reasons can be divided into two aspects:

First, ZK does not require trust in specific individuals. Unlike OP's "innocent until proven guilty" trust mechanism, ZK's trust is built on rigorous mathematical logic, and this technology has been developing for three to four decades, with cryptographic systems based on mathematical assumptions becoming increasingly complete. Therefore, compared to other solutions, placing trust in mathematics is safer.

Second, compared to other scalability solutions, Ethereum's entire architecture is more friendly to ZK development. After generating a proof using ZK, it can be quickly verified in Ethereum, and the generated ZK proof can represent a block on L2, which may correspond to 1,000 transactions. In other words, verifying a ZK proof on Ethereum is equivalent to verifying 1,000 transactions, making ZK a superior scalability solution without significantly altering Ethereum's underlying structure.

Finally, regarding ZK technology's empowerment for Ethereum: I believe the rapid development of ZK has significantly improved Ethereum's settlement time through L2; additionally, through ZK Bridge, Ethereum can enhance efficiency while avoiding many security vulnerabilities. Furthermore, ZK also greatly empowers how Ethereum generates blocks. ZK has a Verifiable Delay Function (VDF) module that uses rigorous mathematical functions to generate random numbers, producing corresponding ZK proofs to verify that the Delay Function has been properly calculated, which provides good assurance for Ethereum's security and fairness.

Deep Tide TechFlow: In the face of the current efficiency issues with ZK, Vitalik has also focused on three solution directions: 1. Parallelization and aggregation trees; 2. Using SNARK algorithms and hashes to improve efficiency; 3. Using ASIC for ZK hardware acceleration. As a project focused on ZK hardware acceleration, what do you think are the advantages of ZK hardware compared to the other two solution directions?

Leo Fan:

In fact, the first solution of parallelization and aggregation trees and the third solution of ZK hardware acceleration have a cooperative relationship. The parallelization and aggregation tree solution is essentially about creating a Parallel EVM or aggregating proofs and batch verification, and hardware acceleration can also speed up this process.

The second solution involves using SNARK algorithms and hashes to improve efficiency. SNARK (Succinct Non-Interactive Arguments of Knowledge) is essentially ZK, and what we accelerate is SNARK. Many projects have numerous parameters in their SNARKs, and if the parameters are set too aggressively, it can impact security. The advantage of ZK hardware acceleration is that it can achieve acceleration without compromising the security foundation.

ZKVM Support and Cysic's Self-Developed Hardware Shipment, ZK May Achieve Large-Scale Adoption by 2025

Deep Tide TechFlow: Recently, the ZK trend has reached Solana, where the privacy protocol Light Protocol proposed a new primitive aimed at Solana: ZK Compression, which aims to enable native ZK computation on Solana, significantly reducing the costs of token and account management on Solana. What are your thoughts on ZK Compression? What opportunities and challenges do you think this will bring to the ZK hardware acceleration track that Cysic is in?

Leo Fan:

ZK Compression is a recently emerging ZK technology on Solana. It mainly uses the succinctness of ZK to reduce state, thereby achieving cost reduction. The same technology is also used in ZK-L2 and ZK bridges. For Cysic, the core of this technology is no different from the technology we are currently developing; it is just a ZK application in a new scenario (Solana), and Cysic's existing technology can provide good support.

Deep Tide TechFlow: How long do you think it will take for ZK technology to achieve large-scale adoption in specific scenarios? What are the important opportunities to drive this vision?

Leo Fan:

I believe that the arrival of large-scale adoption of ZK may happen as soon as next year, and it will not be limited to the crypto industry but will truly realize applications in many landing scenarios. This mainly comes from two aspects of opportunity:

First, on the software side, Cysic is very focused on the development of ZKVM, which aims to help developers so they no longer have to write many ZK Circuits but can directly write corresponding RUST or Go projects and then transform them into ZK Circuits through ZKVM. This greatly improves the development speed of ZK and lowers the software development threshold for ZK. Cysic will continue to explore and launch better solutions on the software side.

On the other hand, Cysic's self-developed hardware will ship next year, which will significantly enhance the efficiency of ZK proof generation, thus becoming a key force driving the large-scale adoption of ZK technology.

Deep Tide TechFlow: Could you share with us some significant recent progress made by Cysic, and what will be the focus of Cysic's work in the second half of 2024?

Leo Fan:

Over the past year, Cysic has made many breakthrough advancements: by the end of 2023, the team has roughly determined the development direction for ZK chip design. Previously, our work may have only focused on accelerating certain heavily weighted modules, but by the end of last year, we confirmed the ZK chip design direction based on ZKVM, which has allowed our hardware products to take on very flexible forms. We will have two products:

ZK Air will be a product aimed at ordinary users. This is a lightweight, portable ZK DePIN device, about the size of a MacBook charger, which can connect to phones or computers via Type-C to achieve efficient ZK proof generation. ZK Air provides computing power greater than that of top consumer-grade graphics cards, comparable to 10 RTX 4090s.

ZK Pro can maximize the acceleration of ZK proof production, allowing users to compute ZK proofs in real-time, designed similarly to traditional mining machines. ZK Pro is more suitable for large ZK projects, such as zkRollup, zkML, etc. In terms of performance, ZK Pro has seen a significant improvement compared to Air, with its computing power comparable to 50 NVIDIA RTX 4090 graphics cards, according to Cysic's official disclosure.

Additionally, in terms of GPUs, we have developed corresponding general-purpose code ourselves, which has greatly improved compared to CPUs and is nearly twice as fast as existing open-source solutions on the market. Before the ASICs meet everyone, we aim to provide better ZK hardware acceleration services for the industry through GPUs and other solutions.

In the upcoming 2024, the Cysic testnet will launch in mid-July, and the mainnet may go live in the third quarter. Most of the development work has been completed, and we are continuously conducting stress tests and other tasks.

Moreover, another major focus of our work is the hardware design based on ZKVM, aiming to provide the community with very specific performance.