Microsoft’s Windows account put the date in plain text: “Save the date: October 7 at 10:00 AM PT. Windows. @Surface. @NVIDIARTXSpark.” The Oct. 7 event, set for San Francisco at 1 p.m. Eastern, is built around what Windows Central’s preview calls “how local AI will shape the next chapter of the PC,” with CEO Satya Nadella and Nvidia CEO Jensen Huang scheduled to appear.
Local AI here means models that run on the machine’s own chip and memory instead of a data center. Microsoft and Nvidia have already published what that looks like on paper, in announcements from May 31 and early June, so most of what the event can add is price, dates and new software. Nvidia’s pitch is that a model fitting in the chip’s 128GB of unified memory needs no network connection to run.
The speakers and the stage on Oct. 7
The Windows Central preview names the theme and the speakers. The lineup, as Windows Latest reported from Microsoft’s post, includes Pavan Davuluri, who leads Windows and devices, and Marcus Ash, a corporate vice president for design and research, alongside Nadella and Huang. The livestream is on YouTube. Coverage by Windows Central, Thurrott and BetaNews agrees on the 10 a.m. Pacific start, which makes the program a midday affair on the East Coast.
What each company will actually unveil is not confirmed. Windows Central lists the Surface Laptop Ultra, a Surface RTX Spark Dev Box and a developer-focused effort called Project Zenith as the expected subjects, and Thurrott adds that Microsoft could use the stage to announce the public rollout of Windows 11 version 26H2. These are expectations from trade outlets rather than announced agenda items.
RTX Spark: the chip meant to run models locally
The chip at the center is RTX Spark, which Nvidia unveiled on May 31, 2026. It pairs a Blackwell RTX GPU with 6,144 CUDA cores and a 20-core Grace CPU over an NVLink-C2C link, delivers up to 1 petaflop of AI compute and offers up to 128GB of unified memory. Nvidia says it can run 120-billion-parameter language models locally with a 1 million token context, and its laptops of about 14mm thickness and three pounds are due this fall from ASUS, Dell, HP, Lenovo, Microsoft Surface and MSI, with Acer and Gigabyte following.
Huang’s own summary in that announcement reads: “Local agents. Frontier models. Creative workflows. RTX games. All on a laptop.” Nvidia and Microsoft also describe a joint security layer, built from new Windows primitives and an Nvidia runtime called OpenShell, that lets people set permissions for on-device agents and route requests by privacy policy.
Microsoft’s side of the partnership, described in its Computex post, frames RTX Spark as an Arm-based chip co-developed with Nvidia and MediaTek and ties it to Windows reaching more than a billion daily users.
Surface Laptop Ultra and the developer box
Microsoft’s May 31 introduction of the Surface Laptop Ultra calls it the most powerful Surface Laptop ever built and says it is engineered with Nvidia and optimized for RTX Spark. The listed hardware includes a 15-inch mini-LED touchscreen reaching 2,000 nits, up to 128GB of unified memory and the ability to run models of up to 120 billion parameters locally. It is described as available later this year, with no price given.
The Surface RTX Spark Dev Box targets developers instead: a compact aluminum chassis that doubles as the heatsink, Windows 11 Pro with WSL 2 and GPU passthrough, CUDA support, and secured-core protections. Microsoft says it will launch later in 2026 exclusively through Microsoft.com in the United States, again without a price. Neither Microsoft page mentions Project Zenith.
DGX Station and the hybrid stack
Local does not mean cloud-free. Nvidia’s Build-week post describes a single stack that spans Windows devices, Azure and local deployments, with CUDA, RTX and TensorRT running across all three. On the local end it lists RTX Spark laptops and desktops for fall 2026, plus DGX Station for Windows, a deskside machine with a GB300 Grace Blackwell Ultra chip, up to 748GB of coherent memory and 20 petaflops of FP4 performance, expected in the fourth quarter of 2026.
On the cloud end the same post places Nvidia models on Microsoft Foundry, Anthropic’s Claude running on GB300 Blackwell hardware on Azure, and the Vera Rubin platform in full production on Azure, while Foundry Local on Azure Local covers on-premises servers. OpenShell, the sandboxed agent runtime that also ships with RTX Spark, is being built into GitHub Copilot.
That leaves the open items for Oct. 7 concrete: the prices and ship dates for the Surface Laptop Ultra and the Dev Box, and whether Microsoft attaches Windows 11 changes to the same stage as the hardware.
This article was produced with the assistance of AI and reviewed by Morning Overview editors prior to publication.
More from Morning Overview
- Invisible volcanic gas at a California ski resort has killed three ski patrollers
- Verizon just moved to cut landline phone service in 9 states within weeks
- Tropical Storm Rachel is dumping up to 12 inches on four Mexican states on its way to major hurricane strength
- The NSA is again telling phone owners to switch off one location setting
