Microsoft and Nvidia have unveiled a new high-powered Surface laptop designed to run demanding artificial intelligence tasks directly on Windows PCs, reducing reliance on cloud computing.
The companies introduced the Surface Laptop Ultra at a Microsoft event in San Francisco on Oct 7, with Nvidia CEO Jensen Huang joining Microsoft CEO Satya Nadella on stage. The device is powered by Nvidia’s new RTX Spark chips and is aimed at developers, creators and other users who need substantial computing power for AI workloads.
Microsoft is positioning the launch as part of a broader effort to make Windows a platform where AI agents can perform tasks such as writing code and handling complex projects locally rather than sending every workload to cloud data centers. The strategy could allow some computing currently performed through Microsoft’s Azure infrastructure to shift to powerful PCs in homes and businesses. Apple is pursuing a similar direction with its latest Mac computers, while Nvidia sees Windows PCs as an opportunity to expand beyond a market largely controlled by Intel and Advanced Micro Devices.
AI moves from cloud to PC
The Surface Laptop Ultra uses Nvidia’s RTX Spark architecture, which combines a Blackwell RTX GPU, a Grace CPU and unified memory. Microsoft said the system can support AI models exceeding 120 billion parameters locally and deliver up to 1 petaflop of AI performance.
Microsoft also demonstrated AI models and applications that can operate directly on Windows devices. Its Copilot system can divide workloads between local hardware and the cloud, allowing more sensitive or cost-sensitive tasks to be handled on the PC.
READ: Microsoft drafts AI code to keep models under human control (September 14, 2026)
Microsoft said Nvidia’s Nemotron models can run locally, while Meta’s Muse personal assistant is also coming to Windows as a native application. The shift could improve privacy and reduce the amount of information that needs to be sent to cloud servers, while potentially lowering cloud computing costs for users and businesses. Microsoft is also introducing Microsoft Execution Containers, or MXC, to restrict what AI agents can access when operating on a computer. The technology aims to prevent autonomous AI systems from accessing sensitive information or taking unauthorized actions. Microsoft said OpenAI, Nvidia and Anthropic will use the technology.
“We needed to make the desktop the most secure place for agents to execute,” Nadella said, according to Reuters. Huang described the security technology as an important development for deploying AI agents on personal computers.
Price could limit adoption
The biggest challenge for the new AI PCs could be their cost.
Microsoft is offering the Surface Laptop Ultra starting at $2,599, with the highest-end configuration reaching $5,899. The top model includes a 20-core processor, 128GB of memory and 1TB of storage. Microsoft has opened preorders, with availability beginning Oct. 16.
Microsoft’s separate Surface RTX Spark Dev Box, aimed at developers and AI engineers, starts at $5,999 and is scheduled to begin shipping in November.The high prices come as a shortage and rising cost of memory chips have pushed up prices for AI-capable computers. Nvidia recently raised the price of its DGX Spark AI desktop by about 75% to $6,950, according to Reuters.
Anshel Sag, an analyst at Moor Insights & Strategy, said the industry has reached a point where the software is ready for local AI but the hardware needed to run it remains expensive. “Now the software is ready and the hardware is too expensive to actually run it locally,” Sag told Reuters, adding that local AI could increasingly become something affordable mainly to users with larger budgets.
Microsoft takes aim at Apple
The launch puts Microsoft and Nvidia more directly into competition with Apple’s high-end Mac computers.
Microsoft’s entry-level Surface Laptop Ultra costs more than Apple’s $1,999 base MacBook Pro, although the machines have different specifications. The companies are competing for developers and professional users who increasingly want to run AI models and applications directly on their computers.
READ: Nvidia unveils AI agent safety platform to prevent models from breaking containment (September 28, 2026)
For Nvidia, the move also represents an expansion into a traditional PC market dominated by Intel and AMD. Microsoft, meanwhile, is betting that powerful local hardware can complement rather than replace cloud AI. The company’s Copilot strategy will continue using cloud infrastructure for the most demanding workloads while shifting suitable tasks to local models.
The approach reflects a broader change in AI computing, as companies seek to balance the capabilities of large cloud-based models with the privacy, speed and potentially lower operating costs of running AI directly on personal devices.