Microsoft has unveiled a new generation of AI-focused Windows PCs powered by Nvidia’s RTX Spark chips, alongside a revamped Windows 11 designed to run artificial intelligence workloads locally and support autonomous AI agents. The headline device is the Surface Laptop Ultra, a premium machine aimed at developers, creators and AI builders who want to run demanding models directly on their computers rather than relying entirely on cloud services. The devices were unveiled at a Microsoft event in San Francisco on October 7.
The launch represents a major shift in Microsoft’s PC strategy. Instead of treating AI primarily as a cloud service, Microsoft is positioning Windows as a platform for “hybrid intelligence,” where workloads can move between the PC and the cloud depending on performance, privacy and computing requirements. The new Windows 11 capabilities include Microsoft Execution Containers for safely sandboxing AI agents, while Copilot is being upgraded to use local context and local AI models with user permission.
Microsoft Launches Surface Laptop Ultra
The Surface Laptop Ultra is Microsoft’s most powerful Surface laptop to date and is built around Nvidia’s RTX Spark platform.
The base model starts at $2,599, while higher-end configurations can reach nearly $5,900. The laptop is available for preorder and is scheduled to begin shipping on October 16, 2026.
Microsoft is targeting the machine at users who need substantial local computing power, including software developers, AI researchers, creators and professionals working with demanding workloads.
Surface Laptop Ultra at a Glance
| Feature | Details |
|---|---|
| Starting price | $2,599 |
| Processor platform | Nvidia RTX Spark |
| Display | 15-inch HDR touchscreen |
| Maximum unified memory | 128GB |
| AI performance | Up to 1 petaflop FP4 |
| Local AI models | More than 120 billion parameters |
| Availability | October 16, 2026 |
| Primary users | Developers, creators and AI builders |
The high-end version combines Nvidia’s CPU and GPU architecture with large amounts of unified memory, allowing AI models that would normally require significantly more powerful systems to run locally.
Nvidia RTX Spark Brings Data-Centre-Style AI to PCs
At the centre of Microsoft’s new hardware strategy is Nvidia’s RTX Spark platform.
The system combines a Grace CPU, Blackwell-based RTX GPU and unified memory architecture. Microsoft says the platform can deliver up to 1 petaflop of FP4 AI performance and support models exceeding 120 billion parameters on the Surface Laptop Ultra.
The significance is that users can run considerably larger AI models on a personal computer.
Traditionally, running sophisticated AI models required powerful cloud infrastructure or specialised workstations.
RTX Spark attempts to bring a portion of that capability onto a laptop or desktop.
Cloud AI
User → Internet → Data centre → AI model → Result
Local AI
User → RTX Spark PC → AI model → Result
Hybrid AI
User → Windows → Local model or cloud model → Result
Microsoft is betting that the third approach will become increasingly important.
Windows 11 Is Being Redesigned for AI Agents
The hardware is only one part of Microsoft’s announcement.
The company is also changing Windows itself to make the operating system more suitable for AI agents.
Microsoft Execution Containers, or MXC, are now generally available on Windows. The technology is designed to isolate AI agents so they cannot freely access files, applications or other system resources without appropriate permissions.
This is important because AI agents behave differently from conventional software.
A normal application generally waits for a user command.
An AI agent can potentially browse the web, write code, open applications, modify files and perform multiple steps autonomously.
That creates new security risks.
Microsoft’s Agent Security Model
Microsoft is building the system around three core concepts:
- Containment: Limits what an agent can access.
- Identity: Identifies which agent performed an action.
- Manageability: Allows organisations to monitor and govern agents.
The company says these controls are intended to make Windows a safer environment for autonomous software.
Copilot Gets Local AI Capabilities
Microsoft is also expanding Copilot through what it calls hybrid intelligence.
Instead of sending every request to a cloud-based AI model, Windows can determine whether a task should run locally or remotely.
This can provide several potential benefits.
Local AI Advantages
Privacy
Sensitive information can potentially remain on the PC instead of being transmitted to a cloud service.
Speed
Local processing can reduce dependence on internet connectivity and cloud response times.
Cost
Running some workloads locally could reduce the amount of cloud computing required.
Control
Users and organisations can have greater control over where AI processing takes place.
Microsoft says Copilot will be able to use local context from a user’s PC and take actions on the user’s behalf with permission. Hybrid-intelligence features are expected to begin rolling out over the coming months on Copilot+ PCs.
Microsoft Brings AI Coding to the PC
One of the most notable demonstrations involves coding.
Microsoft’s MAI Code 1.1 Flash model is being optimised to run locally.
The model has 137 billion total parameters and 6.8 billion active parameters. Microsoft says it has been compressed using 3-bit precision, reducing its size by nearly 80% while retaining coding performance and supporting a 256K context window locally.
Other models are also being prepared for local execution, including an upcoming Nvidia Nemotron model and DeepSeek V4 Flash.
This could make powerful AI coding assistance available even when a user does not want to send code and project information to a cloud service.
Microsoft Is Building a Wider RTX Spark Ecosystem
The Surface Laptop Ultra is not the only device coming to market.
Microsoft says RTX Spark-powered Windows PCs are also being introduced by:
- ASUS
- Dell
- HP
- Lenovo
- MSI
- Microsoft Surface
The ecosystem includes laptops, desktops and developer-focused systems.
Microsoft’s official RTX Spark lineup includes six Windows laptops designed around the platform, including the Surface Laptop Ultra, ASUS ProArt P14 and P16, Dell XPS 16, HP OmniBook Ultra 16 and Lenovo Yoga 9i 2-in-1.
The broader hardware strategy is important because Microsoft cannot transform Windows into an AI-first platform using Surface devices alone.
It needs manufacturers across the Windows ecosystem to adopt the architecture.
Surface RTX Spark Dev Box Targets Developers
Microsoft has also introduced the Surface RTX Spark Dev Box, a workstation designed specifically for developers and AI engineers.
It starts at approximately $6,000 and includes tools such as Visual Studio Code, GitHub Copilot CLI, Windows Subsystem for Linux and PowerShell 7.
The system is designed for users who need more sustained AI computing power than a conventional laptop can provide.
Microsoft says the Dev Box will begin shipping in the United States in November.
Microsoft Is Also Targeting Apple
The new hardware puts Microsoft and Nvidia directly against Apple’s high-end Mac lineup.
The Surface Laptop Ultra starts at a higher price than Apple’s $1,999 base MacBook Pro, although the specifications and target workloads are different.
Microsoft is positioning its new machines around local AI performance rather than simply traditional laptop specifications.
The company claims RTX Spark systems can deliver significant gains in local AI workloads compared with a 16-inch MacBook Pro with Apple’s M5 Pro.
For example, Microsoft cites up to 2.1 times faster time to first token for the RTX Spark Windows ecosystem.
The competitive battle is therefore increasingly moving from CPU performance and battery life toward AI inference capabilities.
Pricing Could Become a Major Challenge
The biggest obstacle to Microsoft’s local-AI vision could be cost.
AI hardware requires large amounts of fast memory, powerful processors and advanced cooling systems.
The memory shortage has already increased costs for high-end AI machines.
Reuters reported that Nvidia recently increased the price of its DGX Spark AI desktop to around $6,950, up approximately 75%, partly because of rising memory costs.
That raises a broader question about who can afford powerful local AI.
Microsoft originally promoted AI PCs partly as a way to reduce reliance on expensive cloud computing. But if high-performance local AI machines become extremely expensive, the technology could initially remain concentrated among developers, professionals and businesses with large hardware budgets.
Microsoft Also Introduces Windows AI Supercomputers
At the top end of the new ecosystem, Microsoft is preparing Windows systems powered by Nvidia DGX Station technology.
These machines use Nvidia’s GB300 Grace Blackwell Ultra Desktop Superchip and are designed to bring data-centre-class AI capabilities to local systems.
Microsoft says the systems will eventually be able to run models with more than one trillion parameters locally.
That represents a dramatic expansion of the definition of a personal computer.
The Windows ecosystem could increasingly range from lightweight AI laptops to powerful local AI workstations and desktop supercomputers.
The Bigger Picture
Microsoft’s new Nvidia-powered PCs and revamped Windows 11 represent a shift from the traditional cloud-first AI model toward a combination of local and cloud computing. The company’s “hybrid intelligence” strategy is designed to let Windows determine where AI workloads should run while giving users and organisations more control over data, permissions and autonomous agents.
The strategy also gives Microsoft a way to make Windows more relevant in the emerging agentic-AI era. If AI agents become capable of performing meaningful computer tasks independently, the operating system will need to control their access to files, applications and data. Microsoft is attempting to build those controls directly into Windows while Nvidia supplies the computing power needed to run increasingly sophisticated models locally.
Looking Ahead
The first major test will be whether developers and professional users are willing to pay premium prices for local AI performance. The Surface Laptop Ultra begins shipping October 16, while other RTX Spark systems from Microsoft’s hardware partners are also scheduled to arrive from mid-October.
Over the longer term, Microsoft’s success will depend on whether hybrid intelligence becomes a mainstream Windows feature rather than a premium capability for AI specialists. If local models become more powerful and hardware prices fall, PCs could increasingly perform tasks that today require cloud AI services. That could reshape both the Windows software ecosystem and the economics of AI computing.
Get the day’s top stories in your inbox
One concise email. No spam, unsubscribe anytime.
