Microsoft opened preorders Wednesday for Surface PCs built on Nvidia’s new RTX Spark chip and introduced a revamped Windows 11 meant to run AI agents directly on the machine, TechCrunch reported. The company pitched the hardware as a way to run AI models locally rather than paying for cloud compute.

What Happened

Microsoft CEO Satya Nadella and Nvidia CEO Jensen Huang presented the products at an event in San Francisco during the city’s Tech Week. The centerpiece is the Surface Laptop Ultra, which starts at $2,600, according to TechCrunch. A version with a more powerful chip starts at $3,700, and the top configuration, priced at $5,900, was already listed as out of stock. The laptops become available Oct. 16, according to Nvidia.

Microsoft also announced the Surface RTX Spark Dev Box, a desktop for developers that TechCrunch said starts at $6,000. It comes with Microsoft’s developer tools preinstalled, including VS Code, GitHub Copilot CLI, the Windows Subsystem for Linux and PowerShell 7, and is due in November, according to Fortune. Dell’s XPS 16 Creator Edition, which uses the same chip, is on preorder at Best Buy for $3,800 with delivery in late October.

RelatedOpenAI Says Its Models Broke Into Two Outside Data Sites and Left It Out of Their Answers

On the software side, Windows 11 gains Execution Containers, a feature designed to make it easier to isolate AI agents in a sandbox. Microsoft said it will be available to all Windows 11 users. Fortune reported a separate feature called Autopilot that lets users name an agent, give it a role and a goal, and set it to work on tasks such as tracking deadlines, sending reminders to colleagues and compiling project updates.

Microsoft is also offering up to $1,000 in trade-in credit for MacBook Pros.

Why It Matters

RTX Spark is Nvidia’s first processor platform for Windows PCs, which puts the company in direct competition with Apple’s Mac chips and with Intel, AMD and Qualcomm for high-end laptops. Nvidia says the chip combines a Blackwell RTX GPU with up to 6,144 cores and a Grace CPU with up to 20 cores, linked at 600 GB per second, with up to 128GB of unified memory and 1 petaflop of FP4 AI compute. That shared memory pool is what allows large language models to run on the device instead of in a data center.

Microsoft framed the strategy as “hybrid intelligence,” Fortune reported, with Windows combining cloud models and models running locally. Nadella argued that a model alone is not enough for useful agents. According to TechCrunch, he said agents need orchestration and memory kept outside the model. He also said Microsoft needed, according to Nvidia, to make the desktop the most secure place for agents to run, which is the job Execution Containers is meant to do.

For enterprise IT teams, the pitch is that some AI workloads, including coding agents, could move from metered cloud usage to hardware they already own. The prices, however, put these machines in workstation territory rather than in mainstream corporate fleets.

The Hardware Lineup

Nvidia lists Acer, ASUS, Dell, HP, Lenovo, Microsoft, MSI and Gigabyte as RTX Spark PC makers, with compact desktops due in November. Nvidia also previewed DGX Station for Windows, built on its GB300 Grace Blackwell Ultra Desktop Superchip with 748GB of coherent memory and up to 20 petaflops of FP4 compute. Nvidia said that system can run models of up to a trillion parameters locally but did not give an availability date.

Huang described an agent living on the PC as a personal assistant. Microsoft’s cloud business gives it a stake in both approaches: Fortune noted that Microsoft 365 Copilot passed 30 million paid seats in July.

What to Watch

The first laptops ship Oct. 16, followed by the Dev Box and compact RTX Spark desktops in November. Independent testing will show how well large models actually run on the new hardware. The other open question is adoption on the software side: how many Windows developers build agents around Execution Containers and Autopilot, and whether companies are willing to let those agents act on employees’ machines.