# Cysic Network Whitepaper ### TL;DR ComputeFi is Cysic's vision for a decentralized compute economy, transforming GPUs, ASIC miners, and servers into a programmable, verifiable, and liquid resource. Today, compute remains siloed and centralized, driving up costs for AI, zero-knowledge proofs, and mining. ComputeFi solves this by matching providers and requesters through a decentralized marketplace, where tasks are executed and verified via cryptography, redundancy, or consensus. With Cysic’s vertically integrated hardware stack, from custom ASICs to GPU clusters and portable miners, ComputeFi creates scalable, real-world compute liquidity, positioning compute as the missing pillar of Web3 infrastructure alongside DeFi, storage, and bandwidth. ## Introduction In the digital economy, computation has become the most valuable raw material. Artificial intelligence models grow larger each year, consuming billions of GPU hours. Zero-knowledge proofs now underpin the scalability of blockchains, requiring millions of parallel proof generations to sustain Ethereum and its rollups. Cryptocurrency mining, once dominated by hobbyists, has evolved into an industrial-scale activity that secures hundreds of billions of dollars in value. Yet the supply of compute remains siloed, expensive, and centralized. Access to GPUs is largely monopolized by cloud providers. Proof generation is fragmented into closed prover services. Mining power is concentrated in the hands of a few industrial pools. For developers, researchers, and ordinary users, compute remains both scarce and inaccessible. At Cysic, we believe compute should be treated not as a privilege but as a publicly accessible resource — liquid, programmable, and verifiable. Just as decentralized finance (DeFi) unlocked capital from banks, ComputeFi unlocks compute from closed platforms. ComputeFi is the foundation of this vision. It is a decentralized compute infrastructure that financializes GPUs, ASICs, and other accelerators, transforming them into a globally accessible marketplace. Within ComputeFi, workloads of many types, from ZK proofs to AI inference to mining, can be requested, executed, and verified transparently. ### Background and Motivation The last decade has seen the progressive decentralization of key digital resources, such as finance, storage and bandwidth. But one resource remains stubbornly centralized: compute. Cloud providers like AWS, Google Cloud, and Azure dominate the AI landscape. Mining has consolidated into industrial-scale facilities. Prover networks remain experimental, and AI compute markets are fragmented into niche platforms. Compute is not only the most valuable digital commodity. It is also the most complex. Unlike storage or bandwidth, computation varies in type, complexity, and verification method. Training an LLM, generating a zkSNARK proof, and mining a Bitcoin block all require radically different hardware, software, and verification logic. This complexity explains why no project has yet achieved a general-purpose compute market. Existing attempts remain siloed: - GPU rental platforms like Render and IO.net address AI inference but not proofs or mining. - Prover-as-a-service companies like Succinct or RiscZero address ZK workloads but not AI or HPC. - Mining pools tokenize hashpower but cannot flexibly reallocate it to new domains. The result is fragmentation: islands of compute liquidity that cannot be bridged into a unified economy. To unlock the next wave of innovation, compute must become: - Programmable: Just as capital in DeFi became composable via smart contracts, compute must be accessible through APIs and programmable workflows. - Verifiable: Users should not need to trust providers blindly. Correctness must be guaranteed by cryptographic proofs, redundancy, or consensus mechanisms. - Liquid: Compute should be treated as an asset class. A GPU hour, an ASIC hash, or a zkSNARK proof should be interchangeable and tradable across domains. - Accessible: Anyone, from an AI lab to a student with a gaming PC, should be able to request or provide compute. This is the motivation behind ComputeFi: to make compute a liquid and trustless resource layer for the digital economy. ### What is ComputeFi ComputeFi is the financialization of compute resources. It turns raw computation into a programmable, tradable, and verifiable primitive, just as DeFi transformed idle capital into liquidity pools. In ComputeFi, compute cycles and hashrates are contributed by providers, requested by users, and verified by the network. The result is a unified marketplace where ZK proofs, AI inference, and mining workloads coexist. ComputeFi is built on four foundational principles: - Hardware-Agnostic: The network is open to GPUs, ASIC miners, CPUs, and specialized accelerators. No single hardware vendor dominates the protocol. - Workload-Agnostic: ComputeFi supports multiple domains, from blockchain proving to AI to HPC. Modules define how each workload type is executed and verified. - Verifiable: Correctness is enforced at the protocol level. Different workloads employ different mechanisms — zkSNARK validity for proofs, redundancy for AI inference, and hash verification for mining. - Composable: Developers can integrate ComputeFi into dApps, protocols, or AI services via standardized APIs, treating compute as a first-class resource. Cysic is uniquely positioned to realize this vision because it integrates silicon design, infrastructure, and blockchain coordination. Unlike software-only protocols, Cysic can onboard real hardware, from ASICs to portable miners, ensuring that ComputeFi is not just theoretical, but practical and scalable. ## Cysic Network To understand how Cysic Network operates, we first define the actors, their interactions, and the threats the system must defend against. ### Roles There are several different actors in Cysic Network: - Compute Providers: These are individuals or organizations who contribute raw computational power to the network. Providers may range from a single GPU in a desktop computer to industrial-scale clusters of ASIC miners. Providers register their hardware, receive workloads, execute them, and submit results back to the network. - Task Requesters: These are users, dApps, or enterprises that require computation. A requester may be a rollup operator needing zkSNARK proofs, a zkVM prover network requiring proving cycles, or a miner outsourcing hashpower to a decentralized pool. Requesters specify workload requirements (e.g., GPU memory, runtime constraints, deadline) when submitting tasks. - Verifiers: Independent participants who check the correctness of results. For some workloads, verification is trivial (e.g., hash difficulty checks, verifying ZK proofs). For others, it requires cryptographic proofs or redundancy (multiple providers executing the same AI inference). Verifiers anchor trust in the network. - Marketplace: The coordination layer of ComputeFi. The marketplace matches tasks with providers, balancing performance, fairness, and reliability. The marketplace presents a user-facing interface where requesters can submit tasks and monitor execution. For instance, the lifecycle of a ZK proof task is: the task requestor first specifies the essentials of a task, such as the software version, deadline, and reward. Multiple compute providers then proceed bidding for this task, with the rank of weighted sum of reserve tokens and bid. After the proof task is carried out by the winning provider, the result is sent out to multiple randomly selected verifiers to verify the result. After all the verification is done, the reward will be distributed to participated providers and verifiers and whole process is recorded on Cysic Network blockchain. In ComputeFi, compute is represented as workload-specific units: - GPU cycles, expressed in FLOPs (floating-point operations) - ASIC cycles, expressed in hashes per second. - Proofs, expressed in cycles in zkVM. The protocol normalizes heterogeneous resources into a comparable model so that different workloads can be fairly priced, allocated, and traded. ### Threat Model To make sure every task runs smoothly, we need to defend against some attacks. We list some canonical attacks below: - Malicious Results: Providers may attempt to return incorrect or fabricated results. - Dropouts: Providers may fail to deliver tasks due to hardware failure or intentional disruption. - Collusion: Groups of providers or verifiers may attempt to manipulate outcomes. - Sybil Attacks: An adversary may create many fake identities to increase influence in scheduling or verification. - Censorship: Schedulers or dominant providers may attempt to block certain tasks. ### Architecture Cysic Network is designed as a modular stack, ensuring flexibility while preserving coherence across domains. It is built using Cosmos CDK as a layer-1 blockchain. The layered architecture from bottom to top can be described as: - Hardware Layer: This is the physical layer of the Cysic Network, where CPU/GPU/FPGA servers, mining rigs and portable computing devices, including cellphones and miners, constitutes the foundation of the network. - Consensus Layer: As Cysic Network is built upon Cosmos CDK, which uses CometBFT algorithm. CometBFT follows the Byzantine Fault Tolerance (BFT) consensus model, meaning it is designed to handle situations where some nodes in the network behave maliciously or fail to follow the protocol correctly. It tolerates up to one-third of the nodes failing or behaving maliciously without compromising the system’s integrity. The consensus in Cysic Network, Proof-of-Compute, is developped based on this consensus mechanism. In this updated consensus mechanism, not only the staked tokens, but the amount of computation pledged in the system is taken into the consensus reaching process. - Execution Layer: This layer is responsible for job scheduling, workload routing, bridging, voting and some other basic functionalities of the network. The functionalities are achieved by various smart contracts deployed on the EVM-compatible blockchain. - Product Layer: The product layer is the interaction portal of multiple products in Cysic Network, which currently includes a ZK proof market, an AI inference framework, a crypto mining framework and some other products. These are domain-specific modules (ZK proving, AI inference, mining, HPC workloads). Each service defines how tasks are executed and verified. The layered architecture provides several advantages. By separating concerns, it ensures that improvements or changes in one layer do not disrupt others. For example, new proof systems or AI models can be added at the product layer without modifying the underlying consensus or hardware logic. This modularity accelerates innovation, improves scalability, and makes the system resilient to evolving workloads. It also creates a clear interface for developers and hardware providers, enabling rapid onboarding of new compute resources while preserving protocol stability. ### Dual Token Model The goal of the Cysic Network is to establish a decentralized and reliable proving and verification service that fosters community growth and self-sustainability. The token will be used to incentivize provers, verifiers, and validators within the protocol, establishing an effective governance and reward distribution mechanism. Cysic Network uses a dual-token model, consisting of the network token and governance token. Each token plays a specific role in the network, working together to build the Cysic Network ecosystem: - **$CYS**: The $CYS token is the native token of the Cysic Network and is used to pay transaction fees, block rewards, and other network-related activities. $CYS ensures the liveness and vitality of the network through its transaction fee mechanism and serves as one of the incentives for users to participate in network activities. - **$CGT**: $CGT is the governance token and is non-transferable. It can be obtained by staking $CYS in a 1:1 conversion ratio. The un-staking process takes longer than the staking process, as implemented in the Cosmos SDK. Compute providers contribute their computing power to the pool, which in turn provides services to ZK, AI and crypto mining projects. In addition to receiving block rewards from the Cysic Network, users can also buy $CYS and stake them to gain voting power to govern the computing pool. The distribution of computing power can be dynamically adjusted based on several key factors, with the external token rewards from different projects being one of the main factors. The Cysic Network requires computing providers to reserve a certain amount of $CGT initially to defend against malicious behavior. All eligible providers can connect to the Cysic Network by staking $CYS as collateral to maintain the reliability and sustainability of the network service. ## Hardware-Software Co-Design, Application and Case Studies Cysic’s competitive advantage lies in its **full-stack integration of hardware and protocol design**. While most networks focus solely on software coordination, Cysic builds the silicon and infrastructure that power the protocol itself. This vertical integration ensures efficiency, scalability, and resilience against hardware supply shocks. - **ZK ASICs:** Cysic designs specialized ASICs tailored for zero-knowledge proofs, achieving orders-of-magnitude improvements in throughput and energy efficiency compared to commodity GPUs. These chips are purpose-built to sustain real-time Ethereum block proving and other cryptographic workloads. - **GPU Clusters:** General-purpose GPUs remain essential for AI inference and training. Cysic operates optimized clusters with custom CUDA kernels for verifiable workloads, enabling both high-throughput and trust-minimized execution. - **Portable Miners:** Consumer-friendly Dogecoin and Bitcoin miners, branded under Cysic, are designed for low energy cost and easy onboarding. These devices integrate directly into the network, bringing retail users into the ComputeFi economy. - **Heterogeneous Onboarding:** Beyond Cysic’s own devices, the protocol supports commodity GPUs and enterprise HPC clusters. This inclusivity ensures global participation and prevents centralization around a single hardware class. By aligning hardware and protocol development, ComputeFi ensures that supply is not just theoretical but **grounded in real, deployable infrastructure**. ### Applications The versatility of ComputeFi allows it to support multiple industries and use cases simultaneously. Key applications include: - **AI Marketplaces** Developers can deploy machine learning models and serve inference queries on-demand. Requesters pay per inference, while providers execute tasks and verifiers ensure correctness. This enables verifiable LLM outputs, pay-as-you-go inference services, and decentralized alternatives to centralized AI APIs. - **Blockchain Infrastructure** Rollups and privacy-focused L1s require vast zkSNARK proving capacity. Instead of relying on closed prover-as-a-service models, they can outsource workloads to ComputeFi. The network delivers cryptographically verified results while distributing revenue back to providers. - **Mining Integration** Traditional hashpower is integrated as a form of compute liquidity. Bitcoin and Dogecoin miners participate in ComputeFi while continuing to secure their native networks. In parallel, mining rigs can be flexibly reallocated to secondary workloads, such as ZK or AI tasks, when profitable. - **Scientific HPC** Researchers in fields like genomics, climate modeling, and 3D rendering often face prohibitive cloud costs. ComputeFi provides access to affordable and verifiable compute capacity without relying on centralized providers, opening new opportunities for open science. Through these applications, ComputeFi positions itself as the **general-purpose compute backbone** of the decentralized web. ### Case Studies #### Ethereum Block Proving Ethereum rollups produce millions of transactions daily that require zkSNARK proofs for finality. Traditional provers face bottlenecks in throughput and cost. By distributing proving workloads across a global pool of providers, ComputeFi enables real-time Ethereum block proving. Benchmarks demonstrate reduced latency, lower cost per proof, and higher resilience compared to centralized prover services. #### Verifiable AI Inference Consider a use case where a decentralized marketplace offers GPT-based inference. Requesters submit natural language queries; providers run them on GPUs; verifiers cross-check outputs with redundancy or probabilistic methods. The result is an AI API where users pay only for correct results, without trusting a black-box provider. This opens the door to verifiable AI assistants, content moderation systems, and autonomous agents. #### Dogecoin Mining Integration Cysic’s portable Dogecoin miners exemplify how consumer hardware can plug directly into ComputeFi. Retail users deploy low-energy miners at home, contributing hashpower to secure Dogecoin while also earning participation rights in the compute marketplace. Over time, this expands decentralization of mining and builds grassroots adoption of ComputeFi. ## Conclusion Compute has become the most essential resource of the digital era, yet it remains the least accessible. Centralization by cloud providers and industrial miners has created scarcity, inflated costs, and limited innovation. To unlock the next wave of growth in AI, blockchain, and scientific research, compute must be made programmable, verifiable, and liquid. **ComputeFi is Cysic’s answer.** By transforming GPUs, ASIC miners, and accelerators into a decentralized marketplace, ComputeFi unifies previously siloed domains, ZK proving, AI inference, mining, and HPC, into a single global economy. Its layered architecture ensures modularity and scalability, while Cysic’s vertical integration of silicon, infrastructure, and protocol delivers unmatched real-world efficiency. Through ComputeFi, compute is no longer just a service — it becomes an **asset class**, accessible to anyone, anywhere. Just as DeFi unlocked capital and Filecoin decentralized storage, ComputeFi establishes the missing pillar of Web3 infrastructure: a **decentralized compute economy**. **Cysic’s mission is clear:** to build the foundation of this economy and empower the next generation of applications across AI, cryptography, and beyond.