The laptop integrates Nvidia's RTX Spark N1X chip, combining up to 20 Grace CPU cores, 6,144 Blackwell GPU cores, and 128 GB of unified memory.
Base retail models start at $2,599.99 for 24 GB of RAM, with upper configurations listed at $3,699.99, $4,699.99, and reaching up to $6,000.
The 15-inch mini-LED screen provides a resolution of 3,270 by 2,180 pixels at 120 Hz, reaching 2,000 nits of peak HDR brightness.
Satya Nadella and Jensen Huang demonstrated local execution of tools like Blender alongside new background Microsoft Execution Containers.
Microsoft and Nvidia introduced the Surface Laptop Ultra on 7 October 2026 at an event in San Francisco. The device is built on Nvidia's RTX Spark N1X platform, which pairs up to 20 Grace CPU cores with up to 6,144 Blackwell GPU cores connected at 600 GB/s. The system delivers up to one petaflop of FP4 compute capability and supports up to 128 GB of unified memory running Windows on ARM. In the United States, pre-orders opened at a starting price of $2,599.99 for an 18-core CPU model with 24 GB of RAM and 512 GB of storage. Higher configurations cost $3,699.99 for 32 GB of memory and $4,699.99 for 64 GB, with top specifications reaching up to $6,000, or 6,879 euros in European pricing.
Data from the article text
The computer features a 15-inch mini-LED PixelSense Ultra touchscreen display with a resolution of 3,270 by 2,180 pixels and a refresh rate of up to 120 Hz. Display brightness reaches up to 2,000 nits in high dynamic range across a 10% window, while standard dynamic range content tops out at 1,000 nits. That maximum HDR figure compares directly to 1,600 nits on current Apple MacBook Pro units. The 18-millimeter-thick aluminum chassis integrates rounded edges, redesigned internal cooling vents to manage thermal output, and a haptic touchpad. Power delivery includes charging through standard USB-C alongside a snap-in magnetic connect port.
Data from the article text
Microsoft and Nvidia designed the machine to execute complex artificial intelligence models on local hardware rather than relying entirely on remote cloud clusters. Nvidia chief executive Jensen Huang discussed the evolving technical landscape for software creators during the presentation in California.
“What Satya and I talked about four years ago is that there will be a whole new generation of software developers. They will be AI software developers: they will build AI applications and models or use AI to develop their software. And so the computer must be revolutionized.”
Microsoft chief executive Satya Nadella outlined the operating system strategy, describing Windows as an environment for hybrid intelligence where tasks move between local silicon and external servers.
“There will never be a time where we say: this runs locally, this runs in the cloud. We will expect this hybrid intelligence to be everywhere and pervasive. And that is what I find really exciting.”
Microsoft executive Pavan Davuluri noted that the hardware targets creators and software engineers running demanding local workloads in applications such as Blender, ComfyUI, and Photoshop.
The RTX Spark platform will not remain exclusive to Microsoft hardware, as partner manufacturers including Acer, Asus, Dell, HP, Lenovo, MSI, and Gigabyte revealed matching systems. Alongside the consumer laptops, Nvidia previewed its enterprise DGX Station for Windows based on the GB300 architecture, with compact desktop devices set to follow in November 2026. Microsoft also launched Microsoft Execution Containers into general availability, providing a system-level environment designed to execute background software agents securely. Concurrently, the company is preparing Copilot updates that allow the assistant to examine local files and operate autonomously on the device with user approval. While pre-orders began on 7 October 2026, delivery dates in the United States are scheduled for 16 October, although some reports cite an 18 October release.
Pollar is free and carries no ads in the text. Its members keep it going.
Support PollarAI & Tech · San Francisco and Redmond
Announced in San Francisco, the device runs AI models locally with up to 128 GB of unified memory, with United States deliveries scheduled to begin on 16 October.
AI text based on reports by ANSA.it, ZDNet, Il Messaggero and other publishers. Changed .