Perplexity AI and Nvidia have partnered to launch a "Portable Computer" that runs AI agents locally on desktop GPUs instead of cloud providers.
This shift represents a move away from the centralized cloud model that currently dominates the industry. By processing data on-device, the partnership aims to eliminate recurring costs for users and address growing concerns regarding data privacy.
The system is designed to run on several hardware platforms, including Nvidia DGX Spark and RTX-powered Linux machines [6]. A version for Windows PCs launched July 28, 2026 [1], targeting a market of approximately 1.4 billion Windows devices worldwide [3]. While some reports indicate an initial launch as early as February 2026 [2], the Windows rollout occurred later this summer.
Unlike cloud-based AI, which charges per request or subscription, this local implementation offers zero token costs [4]. The desktop agent is capable of routing tasks across 20 different models [5], allowing the software to select the most efficient tool for a specific query without sending that data to a remote server.
By utilizing the user's own hardware, Perplexity AI aims to provide an always-on computing experience. This approach ensures that sensitive user data remains on the local machine rather than being stored or processed by major cloud providers [2, 6].
“The partnership moves AI processing from the cloud to user GPUs to eliminate token costs.”
The transition toward local AI execution reduces the dependency on massive data centers and the high energy costs associated with cloud computing. By leveraging the existing power of consumer GPUs, Perplexity and Nvidia are challenging the subscription-based token economy and positioning hardware ownership as a primary gateway to AI privacy.
