AnswersPublished22 sources
How Nvidia and Microsoft Aim to Run AI Agents Locally on Windows
Nvidia and Microsoft’s October 7, 2026 initiative pairs RTX Spark PCs with Windows agent controls to run more AI work locally. RTX Spark offers up to 6,144 GPU cores, 20 CPU cores, 1 petaflop of FP4 AI performance and 128GB of unified memory; those are maximum specifications, not the configuration of every model.[36...
Published byEdited with GPT-6 LunaImages generated with GPT Image 2
Research answer
Key takeaways
- Nvidia and Microsoft’s October 7, 2026 initiative pairs RTX Spark PCs with Windows agent controls to run more AI work locally.
- RTX Spark offers up to 6,144 GPU cores, 20 CPU cores, 1 petaflop of FP4 AI performance and 128GB of unified memory; those are maximum specifications, not the configuration of every model.[36]
- Microsoft Execution Containers became generally available for Windows on October 7, with policies for controlling which files and networks agents can access.[2]
AI Prompt
openai.comCreate a landscape editorial hero image for this Studio Global article: How did Nvidia and Microsoft’s Windows-based local AI agent initiative, announced by Jensen Huang and Satya Nadella in San Francisco on Octo. Article summary: The initiative aimed to make Windows PCs a place where AI agents could do useful work locally, rather than sending every task to a cloud service: NVIDIA supplied the computing hardware, while Microsoft supplied the execu. Topic tags: general, general web, documentation. Style: premium digital editorial illustration, source-backed research mood, clean composition, high detail, modern web publication hero. Use reference image context only for broad subject, composition, and topical grounding; do not copy the exact image. Avoid: logos, brand marks, copyrighted characters, real person likenesses, fake screenshots, UI text, readable text, watermarks, charts with fak
Nvidia and Microsoft’s plan is to make Windows PCs capable of running more AI agents and models on the device, rather than relying on a cloud service for every task. Nvidia provides RTX Spark hardware; Microsoft provides Windows infrastructure for containing and controlling agents. The companies describe this as hybrid AI: local execution where it fits, with cloud services still available when needed.2
8
34
How the local-agent setup works
Microsoft Execution Containers (MXC) became generally available on Windows on October 7, 2026. Organizations can use them to set policies for which files and networks an agent may access, with those policies enforced at runtime.2 MXC is intended to contain agents rather than give them unrestricted access to the PC. NVIDIA’s OpenShell runtime integrates with MXC for policy-managed execution and personally identifiable information obfuscation.
32
The aim is to let agents run on a Windows PC while keeping their access subject to rules. That does not, by itself, establish that an agent is secure in every configuration: controls depend on the policies and software in use.
Local execution also does not mean every operation must stay on the PC. NVIDIA describes tools that can route work between local and cloud AI, and names GitHub Copilot and OpenClaw among the agent applications being adapted for RTX Spark.30
34
RTX Spark brings local AI compute to Windows PCs
RTX Spark is the hardware side of the initiative. NVIDIA lists configurations with up to a 6,144-core Blackwell RTX GPU, a 20-core Grace CPU, 1 petaflop of FP4 AI performance and 128GB of unified memory.36 These are maximum specifications; other configurations have lower specifications.
35
NVIDIA says developers can use RTX Spark to run coding agents locally, then test and deploy software on the same machine.30 Microsoft opened preorders for the Surface Laptop Ultra and announced partner devices from ASUS, Dell, HP, Lenovo and MSI. NVIDIA said RTX Spark laptops were available on October 16.
2
42
DGX Station targets larger workloads
For heavier enterprise work, NVIDIA also presented DGX Station for Windows, a deskside system based on the GB300 Grace Blackwell Ultra Desktop Superchip. NVIDIA lists up to 748GB of coherent memory and 20 petaflops of FP4 compute, and says it supports models with up to 1 trillion parameters.45 Those figures and capability claims come from NVIDIA’s product information; they are not a guarantee that every model or workload will run at a particular speed.
At GTC Berlin, NVIDIA positioned RTX Spark for personal agents on Windows, DGX Spark for developers using Linux, and DGX Station for enterprise workloads. The company said the Windows version of DGX Station would arrive in December.1
15
What GTC Berlin added
NVIDIA announced that its Underdog Desktop Agent was available for RTX Spark, describing the app as fully local.1 That is NVIDIA’s description of the product, not independent evidence that every feature or related service always runs on-device. NVIDIA also said many RTX Spark laptops have unlocked bootloaders to support developer customization.
58
Taken together, the announcements outline a shift in where some AI work can happen: personal agents and development tasks on PCs, with larger systems for enterprise workloads and cloud computing remaining part of the mix. Local compute and access controls may offer more control over where tasks run and what agents can access, but they do not make security automatic.2
15
34
Studio Global AI
Continue your research
This page includes a source-backed answer you can continue inside Studio Global.
People also ask
What is the short answer to "How Nvidia and Microsoft Aim to Run AI Agents Locally on Windows"?
Nvidia and Microsoft’s October 7, 2026 initiative pairs RTX Spark PCs with Windows agent controls to run more AI work locally.
What are the key points to validate first?
Nvidia and Microsoft’s October 7, 2026 initiative pairs RTX Spark PCs with Windows agent controls to run more AI work locally. RTX Spark offers up to 6,144 GPU cores, 20 CPU cores, 1 petaflop of FP4 AI performance and 128GB of unified memory; those are maximum specifications, not the configuration of every model.[36]
What should I do next in practice?
Microsoft Execution Containers became generally available for Windows on October 7, with policies for controlling which files and networks agents can access.[2]