Microsoft and NVIDIA are collaborating to integrate AI agents directly into Windows PCs, allowing these agents to run locally rather than relying solely on cloud infrastructure. This initiative aims to transform the personal computer into a more powerful tool for AI, with a focus on security and performance for a new generation of AI-driven applications. NVIDIA founder and CEO Jensen Huang highlighted the historical connection between NVIDIA and Windows, stating, “If not for Windows there would be no GeForce.”
The Shift to Local AI and Agentic Computing
The core idea behind this collaboration is to enable AI agents to execute directly on a user's PC, turning the device into a personal assistant. Microsoft's EVP of Windows and Devices, Pavan Davuluri, emphasized the creation of OS-level infrastructure, Microsoft Execution Containers (MXC), designed to allow agents to run safely and persistently in the background. MXC ensures that agents are secured, observed, and governed by the operating system, a critical aspect for enterprise adoption and general user trust. According to Microsoft CEO Satya Nadella, making the desktop the most secure place for agents to execute is paramount.
This local execution model offers several advantages, including enhanced data privacy, reduced latency, and the ability to run AI tasks without an internet connection or constant cloud interaction. Huang believes that MXC will revolutionize how agents are built and deployed, similar to how Windows and DirectX transformed application development.
RTX Spark: Powering AI on Laptops and Desktops
A key component of this new AI PC era is NVIDIA RTX Spark, which integrates NVIDIA's full AI stack into Windows laptops and compact desktops. Preorders for RTX Spark laptops are now available, with general availability starting October 16, while compact desktops will follow in November. The Surface Laptop Ultra, for instance, is built around NVIDIA RTX Spark, offering up to 128GB of unified memory and up to a petaflop of AI compute. This hardware configuration allows it to run models that traditionally would not fit on a standard machine.
RTX Spark combines an NVIDIA Blackwell RTX GPU with up to 6,144 cores and an up to 20-core NVIDIA Grace CPU, connected at 600 GB/s. This setup delivers one petaflop of FP4 AI performance and up to 128GB of unified memory, making it effective for local AI processing. It can run large models like Qwen 3.8 Flash Next, a 125B model with 51B n-gram, locally without requiring cloud resources or sending data to external servers. This capability is significant for developers, creators, and even gamers. Developers can leverage the full NVIDIA CUDA platform, moving models and workflows without rewriting code, while creators benefit from 5th-generation Tensor Cores and hardware-accelerated video capabilities.
NVIDIA DGX Station for Windows: Enterprise AI on the Desktop
For enterprise developers and researchers, NVIDIA also previewed the NVIDIA DGX Station for Windows. This deskside AI supercomputer brings GB300 Grace Blackwell-class AI infrastructure directly into the Windows ecosystem, a significant development as previous DGX Stations ran on Linux. This means that capabilities that once required renting cloud clusters or maintaining separate Linux environments can now be accessed within Windows.
The DGX Station for Windows runs on the GB300 Grace Blackwell Ultra Desktop Superchip, offering 748GB of coherent memory and up to 20 petaFLOPS of FP4 AI compute. This level of power is sufficient to run models up to a trillion parameters locally. This eliminates the need for Fortune 500 companies, which are largely standardized on Windows, to maintain two separate environments for heavy AI workloads and productivity tools. Developers can now fine-tune and inference large models and build always-on AI agents that connect directly with their existing Windows applications and infrastructure without leaving their primary machine. Linux AI toolchains remain accessible through WSL when needed.
Practical Implications for Developers and Users
The introduction of local AI agents on Windows PCs through RTX Spark and DGX Station has several practical implications. For developers, it means the ability to create more sophisticated AI applications that run directly on user devices, offering improved responsiveness and data security. The unified NVIDIA AI stack across various hardware, from RTX Spark laptops to DGX Stations, simplifies the development process by allowing consistent workflows. For end-users, this signifies a new era of personal computing where AI assistants can operate more efficiently and privately, integrated directly into their everyday tools and operating system.
This strategic partnership between Microsoft and NVIDIA aims to make AI more accessible and powerful for a broad range of users and applications, from everyday tasks on laptops to advanced model development in enterprise settings. The focus on secure, local execution of AI agents marks a significant step toward a more integrated and private AI experience on personal computers.