Every AI model needs training data. Sourcing that data at scale — crawling the web, scraping APIs, aggregating public datasets — requires enormous bandwidth. Grass routes this data collection through its network of 2.4 million connected devices, paying users in GRASS tokens for their unused upload capacity. AI companies pay Grass for the data; Grass distributes that revenue to node operators.

Most dePIN requires hardware purchases. Grass requires only a PC and an internet connection. The browser extension or desktop app runs in the background and earns passively. Hardware cost: zero. This dramatically lowers the barrier to participation — explaining the 2.4M device count, which dwarfs most other dePIN networks by an order of magnitude.

The key question for Grass is revenue sustainability. AI companies are the customers. As the AI training data market matures, prices may compress. Compare Grass to Ocean Protocol, which takes a more institutional approach to data tokenisation with smart contract-based marketplaces. Both are betting that data will be a major on-chain asset class — the mechanisms differ significantly. See Bittensor for the third major AI data strategy.