Perplexity and Nvidia debut on-device AI agent that eliminates cloud token fees
Perplexity has introduced Portable Computer, an AI agent platform that runs entirely on users' existing hardware, starting with Nvidia's DGX Spark and Linux systems with RTX GPUs. Developed with Nvidia, the system keeps models, files, and workloads on the local machine, avoiding any cloud-based billing. The company states that tasks completed locally incur no token costs, marking a significant push toward decentralized AI processing.
The new platform, called Portable Computer, is designed to run AI tasks directly on a user's own hardware rather than relying on remote servers. It launches first on Nvidia's DGX Spark systems and Linux machines equipped with RTX graphics cards, with development support from Nvidia.
By keeping models, files, and workloads local, the system eliminates the metered billing that typically accompanies cloud-based AI services. Perplexity emphasizes that any task completed on the device carries no token fees, positioning the release as a step toward more decentralized AI computing.
This shift could reduce ongoing costs for heavy AI users and address privacy concerns, since sensitive data would remain on local devices. However, it may also widen the gap between those who can afford high-end hardware and those who rely on cheaper cloud services. Businesses and developers could gain more control over their AI workflows, while casual users might see limited benefit until the technology reaches more accessible devices.