Decentralized Compute
Computing power, such as GPU processing for AI workloads, sourced from a distributed network of independent providers and coordinated through a blockchain-based marketplace.
Decentralized compute refers to computing resources — commonly GPU processing power for AI training and inference workloads, though also general-purpose computation — sourced from a distributed network of independent hardware providers, rather than from a single centralized cloud computing company. A blockchain-based marketplace typically coordinates matching compute buyers with available providers, handling payment, and in some designs, verifying that the computation was actually performed correctly.
Decentralized compute is one of the most prominent categories within DePIN (see Kaavon's "DePIN" glossary definition), the broader trend of using token incentives to bootstrap physical infrastructure networks. Proponents argue it can tap into underutilized hardware capacity more efficiently and cheaply than centralized cloud providers, while critics note that verifying decentralized computation was actually performed correctly and matching quality-of-service expectations can be more difficult than with an accountable centralized provider operating under a traditional service contract.
Examples
- A decentralized compute network might let someone with an idle GPU rent it out for AI model training, earning token rewards coordinated through a blockchain-based marketplace.
- Verifying that decentralized compute work was genuinely performed correctly, rather than faked or under-delivered, is one of the harder open problems facing these networks compared to a centralized cloud provider.