NVIDIA used this year's COMPUTEX expo to announce two new AI-focused products: the RTX Spark PC lineup and the OpenShell platform. Together, they're designed to run secure, high-performance AI agents natively on Windows machines, cutting the need for constant cloud connections.
RTX Spark: a new class of AI hardware
The RTX Spark PCs are built specifically for local AI workloads. The machines are meant to handle tasks that usually require cloud servers, from running large language models to processing real-time data from sensors and cameras. By keeping everything on the device, the system reduces latency and keeps sensitive data from leaving the user's control.
NVIDIA described the new PCs as a way to bring data-center-level AI performance to a desktop form factor. The company did not release specific hardware specifications during the keynote, but positioned the line as a direct response to enterprise demand for secure, on-premise AI.
OpenShell: the software that powers AI agents
The other major piece of the announcement was OpenShell, a platform that allows developers to build and deploy AI agents on Windows without relying on remote inference. OpenShell integrates with the RTX Spark hardware to provide a full stack for local AI, from model training to execution.
NVIDIA said OpenShell supports multiple AI frameworks and includes built-in security measures to keep the agent operations isolated from the rest of the system. The platform is intended for industries like healthcare, finance, and manufacturing where data privacy regulations often forbid sending information to external clouds.
Why the shift to local AI matters
For years, most AI workloads have run on cloud infrastructure. But concerns over data sovereignty, network reliability, and cost have pushed some companies to look for local alternatives. NVIDIA's announcement at COMPUTEX suggests the company sees a growing market for PCs that can handle these jobs without a round trip to a data center.
The combination of RTX Spark hardware and OpenShell software could allow organizations to deploy AI agents that work offline or with intermittent connectivity. That's especially relevant for edge computing scenarios, such as factory floors, retail stores, and remote research sites.
NVIDIA did not announce pricing, availability, or a list of partners for either product. The company said more details will come later this year. Developers and enterprise customers at COMPUTEX are likely to push for concrete timelines in the coming days.
