Remember when the hottest laptop spec was "it has a DVD burner"? Now Microsoft and NVIDIA are selling laptops built for AI agents that keep working in the background while you're off doing something else. Clippy could never.

RTX Spark Leaves the Keynote

At Microsoft's Windows AI and Surface event in San Francisco on October 7, NVIDIA opened preorders for RTX Spark laptops, and Microsoft announced general availability of Microsoft Execution Containers (MXC). MXC is OS-level plumbing that lets AI agents run safely and persistently in the background, with Windows controlling and watching what they do. The laptops ship October 16, compact RTX Spark desktops follow in November, and a Windows version of NVIDIA's DGX Station is expected in Q4.

The specs are a lot: up to 1 petaflop of FP4 AI performance, up to 128GB of unified memory, and support for models up to 120 billion parameters running locally. Systems are coming from Acer, ASUS, Dell, HP, Lenovo, Microsoft (including a Surface Laptop Ultra), MSI, and Gigabyte. Pricing wasn't disclosed, which is usually the hardware world's way of saying "sit down first."

The Real Headline Is the Sandbox

The chip gets the spotlight, but MXC is the bigger story. Satya Nadella put it bluntly: Microsoft "needed to make the desktop the most secure place for agents to execute." That's an admission that letting autonomous software click around your files with your permissions was never a great idea, and that the operating system itself has to enforce what agents can and can't do.

For businesses, this changes the math. Until now, "AI agent" mostly meant "send your data to someone's cloud and hope." Local hardware that runs serious models, plus OS-level containment and governance, makes on-device agents a realistic option for workloads where privacy, latency, or compliance used to rule them out. The catch is that capable hardware doesn't fix sloppy agent design. An agent with broad permissions is still a liability, no matter how fast it runs.

The PC isn't dead. It just got a coworker, and that coworker's access had better be locked down.

If you're wondering which of your business workflows could actually hand off to an agent, and what guardrails it would need, let's talk through it. We'd much rather design those permissions with you up front than help clean up afterward.

Source: Unite.AI