Grass Network is a DePIN project that transforms users' unused internet bandwidth into distributed infrastructure for accessing public web data. The network is primarily focused on the artificial intelligence market: collected data can be used for model training and for providing up-to-date context during inference. Grass's technical architecture combines user nodes, routers, validators, a ZK Processor, and a Data Ledger, while proofs of data provenance are linked to the Solana blockchain. The system's economic asset is the GRASS token, which is used for staking, rewards, and governance participation.
Contents
- Grass Network: How Data DePIN Works
- Grass Nodes, Routers and Public Web Data Collection
- Sovereign Data Rollup and AI Data Provenance
- GRASS Token, Staking and Network Tokenomics
- Grass Network Development in 2026 and Data DePIN Risks
1. Grass Network: How Data DePIN Works
Grass began developing as a Wynd Labs project in 2023. Its concept is based on creating a distributed network of internet connections that can be used to retrieve information from publicly accessible websites. Instead of relying on a centralized pool of IP addresses, Grass allows users to install a software node and provide a portion of their unused bandwidth.
This approach belongs to the DePIN sector, or decentralized physical infrastructure networks. However, Grass differs from projects that distribute computing power or storage: its primary resource is internet connectivity and access to the public web through geographically distributed IP addresses. This infrastructure can be used to collect text, images, and other publicly available content required by network clients.
Demand for this type of infrastructure is partly driven by the development of AI. Training large models requires substantial datasets, while retrieval-based applications and AI agents need fresh information at the time a query is processed. Grass is gradually expanding its positioning from a provider of training data to infrastructure for live web data.
The project has raised several funding rounds. In 2024, its Series A was led by Hack VC with participation from Polychain Capital, Delphi Digital, Lattice, and Brevan Howard Digital; Wynd Labs had previously announced a $3.5 million seed round. In October 2025, Grass also raised $10 million in a bridge round involving Polychain Capital and Tribe Capital to support further infrastructure and commercial development.
2. Grass Nodes, Routers and Public Web Data Collection
At the user level, Grass begins with a node. The software runs on a device and allows the network to use a portion of available internet bandwidth to make requests to public web resources. According to the project's documentation, a node is not designed to access a user's browser history or personal files; instead, it routes traffic required to retrieve publicly available data.
Not every connected node receives the same amount of traffic. Grass considers connection availability, stability, software version, and demand for a particular geographic location. In 2026, the team reported that during Stage 2, approximately 150,000 users processed around 90% of network traffic, demonstrating significant differences in the actual utilization of connected resources.
Key participants and stages of the Grass network:
- a user runs a Grass Node and provides unused bandwidth;
- the node receives authorized network requests to public websites;
- Routers connect multiple nodes to the network's verification layer;
- routers track transferred data volumes, latency, and connection status;
- Validators receive and verify web transactions submitted by routers;
- the ZK Processor aggregates proofs of valid network sessions;
- the Data Ledger links datasets with information about their provenance;
- processing transforms unstructured web data into formats suitable for AI systems.
Users receive Grass Points for participating in the network. The current model distinguishes between Uptime Points, which are associated with a node's presence on the network, and Network Points, which are earned when the connection is actually used to process traffic. As a result, remaining continuously connected does not guarantee the same level of rewards for every participant.
The Grass model should not be confused with conventional mining. Users do not perform Proof-of-Work or train AI models on their computers. Their primary contribution is a network resource: available bandwidth and an IP connection through which the infrastructure retrieves public data.
3. Sovereign Data Rollup and AI Data Provenance
The second layer of Grass is the Sovereign Data Rollup. Its purpose is not only to collect information but also to create a verifiable chain of data provenance. This is important for AI developers because datasets need information about where content originated and what operations were performed during processing.
Validators receive web transactions from routers, verify them, and organize them into batches. The ZK Processor generates validity proofs for network session data and submits these proofs to L1. This means the entire collected content does not need to be stored directly on Solana: the blockchain instead serves as a settlement and verification layer for proofs.
| Component | Primary Function | Resource Processed | Role in Data DePIN |
|---|---|---|---|
| Grass Node | Web traffic transmission | Bandwidth | Public data retrieval |
| Router | Connects nodes and validators | Network requests | Routing and accounting |
| Validator | Verification and batching | Web transactions | Network session validation |
| ZK Processor | Proof generation | Session data | On-chain verification |
| Data Ledger | Links data and proofs | Datasets and metadata | Data provenance verification |
An additional stage involves transforming the original content. Edge Embedding Models and data preparation processes are designed to clean, normalize, and structure collected information so that it can be used for AI training, fine-tuning, vector databases, and other applications. Grass therefore aims to function not simply as a distributed proxy network but as part of the AI data production pipeline.
At the same time, the degree of decentralization should be evaluated separately for different layers of the system. Grass's technical documentation originally described the validator model as centralized around a single validator, with a subsequent transition toward a validator committee. Therefore, having a distributed network of user nodes does not by itself mean that the entire data verification and management system is fully decentralized.
4. GRASS Token, Staking and Network Tokenomics
The ecosystem's native GRASS token was issued on Solana with a fixed total supply of 1 billion tokens. According to the official tokenomics, 30% of the supply is allocated to the community, 22.8% to Foundation & Ecosystem Growth, 25.2% to early investors, and 22% to contributors.
Within the community allocation, 170 million GRASS are reserved for future incentives, 30 million for Router Rewards, and 100 million for Airdrop One. The first airdrop began on October 28, 2024 and represented 10% of the total token supply. The project stated that the distribution targeted more than 2 million early participants, while independent post-launch data recorded approximately 1.5 million addresses receiving tokens during the first few days.
GRASS is used in the staking system through routers. Holders can delegate tokens to an available Router on Solana and receive a share of rewards associated with its operation. Returns are not fixed and depend on program rules, infrastructure activity, and the amount of available rewards.
The token is also connected to governance. GRASS holders are intended to participate in discussions about network parameters, incentives, and future development. However, the token's economic utility should be distinguished from its market value: the GRASS price depends on liquidity, demand, token unlock schedules, and broader cryptocurrency market conditions rather than directly on the amount of data processed by a specific node.
In 2026, the user reward model changed. For participation in Stage 2 between October 14, 2024 and June 8, 2026, Grass decided to distribute the corresponding rewards in USDC instead of issuing additional GRASS. The team separately stated that this program would not increase the token's circulating supply. This demonstrates that bandwidth rewards and GRASS tokenomics can operate as related but distinct mechanisms.
5. Grass Network Development in 2026 and Data DePIN Risks
In 2026, Grass is expanding its model from collecting training datasets toward infrastructure that gives AI systems access to current internet data. One direction is Live Context Retrieval, which allows AI models to obtain fresh information during inference instead of relying exclusively on their training data. The project is also adjusting its plans according to actual demand. Grasshopper was designed as an always-connected hardware node, but in July 2026 the team decided not to proceed with its large-scale deployment and redirected resources toward other priorities.
For users, the main risks are associated with demand for bandwidth and the reward system. A node may receive limited traffic because of its geographic location, connection quality, or the large number of available participants. Grass Points should also not be considered a fixed monetary return, as the format and conditions of rewards are determined separately.
The quality of AI data depends not only on proof of provenance but also on the accuracy and relevance of the original content. A ZK proof can verify certain properties of the data retrieval process, but it cannot guarantee that the information itself is correct. Website policies, data regulation, and the further decentralization of the infrastructure also remain important factors.
Overall, Grass Network combines distributed internet bandwidth, public data collection, and blockchain-based mechanisms for verifying data provenance. The prospects of this Data DePIN model will depend on actual demand for training datasets and live data for AI applications, while GRASS adds an economic layer through staking and governance.
