Table of Contents
Chutes wants token ownership to represent a durable claim on the network's compute.
On August 24, Chutes outlined the "conviction lock," a proposed mechanism for distributing GPU capacity according to time-weighted locked stake. The announcement gives investors, operators, and AI builders an early look at how Chutes could grow from a company-led service into infrastructure shared by independent providers across jurisdictions.
How Conviction Lock Would Work
Under the proposal, participants would lock stake to establish a proportional right to the compute available across the Chutes network. A provider that needs more capacity would lock more stake, while a holder that does not want to operate a service directly could delegate its access to another provider.
Chutes described conviction as a time-weighted, on-chain score that builds gradually and takes time to unwind. That structure is intended to favor providers with a durable commitment to the network instead of allowing short-term stake movements to create unstable claims on capacity.
The available compute would still fluctuate as miners add or remove hardware, much like Bitcoin's hash rate changes over time. Chutes said actual usage would also influence allocation, allowing unused capacity to absorb burst demand elsewhere rather than leaving GPUs idle.
Once allocated, providers could use the compute to run inference, fine-tune models, or resell capacity to customers. The proposal therefore gives locked tokens a direct utility within the network, although Chutes has not yet published the complete implementation details or launch schedule.
Chutes.ai Would Become One Provider Among Many
The larger objective is to separate the network from the company that currently operates its most visible service. Chutes said Chutes.ai should be understood as one provider using stake to access GPUs supplied by the network, not as the network itself.
In the model described by the team, independent providers could operate in the United States, the United Kingdom, Singapore, and other jurisdictions. Each provider would hold or control its own locked share of network access, comply with local law, and serve customers through the same underlying compute infrastructure and end-to-end encryption.
That federation is designed to reduce the risk that one company, founder, or regulatory regime becomes a single point of failure. If a jurisdiction restricts a particular model, the local provider could stop serving it and delegate the affected portion of its access to another provider that can offer the model lawfully. The restricted provider remains compliant, while the broader network can continue operating through a different jurisdiction.
What Conviction Lock Could Mean for Chutes
Export controls, model licenses, and regional restrictions are making AI distribution more fragmented. Chutes is betting that a federated provider model can handle that complexity without making one operator responsible for every jurisdiction. Its borderless, permissionless "Linux of AI" framing depends on preserving common infrastructure while allowing each provider to control what it can lawfully serve.
The announcement remains a design proposal, and the next updates will need to explain how capacity will be measured, how allocation will respond to demand, what safeguards will govern delegation, and when the mechanism will reach users. Those details will determine whether conviction lock can support decentralized service delivery without sacrificing compute utilization or regulatory compliance.
