Microsoft unveils Surface Laptop Ultra with Nvidia RTX Spark for on‑device AI

Microsoft revealed the Surface Laptop Ultra at a press conference held on October 7, 2026, announcing a new premium‑laptop segment that combines the Windows 11 ecosystem with NVIDIA’s RTX Spark AI platform. The device carries a list price of US$2,599, roughly €2,300, and Microsoft has already opened pre‑orders with delivery expected on October 16, 2026. The announcement was accompanied by a NVIDIA statement confirming the integration of its Blackwell architecture into Microsoft’s hardware.

Hardware and architecture

At the heart of the Surface Laptop Ultra is the combination of an RTX Spark GPU with up to 6,144 CUDA cores and an NVIDIA Grace processor with up to 20 cores, linked by a 600 GB/s interconnect. This architecture, based on the Blackwell generation, enables the laptop to deliver up to one petaflop of FP4 performance in AI tasks, according to Microsoft, matching the compute power of the 2016 NVIDIA DGX‑1 data‑center server. Planned configurations include up to 128 GB of unified memory and an SSD that can be upgraded after purchase, although the exact storage size has not been disclosed.

The Ultra’s display is a 15‑inch mini‑LED panel with touch capability and HDR support. Microsoft claims the peak brightness exceeds that of the 16‑inch MacBook Pro equipped with the M5 chip by about 25 %, positioning the device as a visually more striking alternative for professionals working with multimedia content. Although touch‑enabled, the laptop does not accept a stylus; instead, Microsoft equips the Ultra with a large haptic trackpad for cursor control.

AI performance

The RTX Spark platform, announced by NVIDIA, brings the full CUDA software stack to Windows PCs, allowing developers to run existing CUDA workloads without code changes. Microsoft states that the Surface Laptop Ultra can achieve up to one petaflop of FP4 performance, which the company says is equivalent to the compute power of the 2016 DGX‑1 server. NVIDIA adds that the chip can run large‑scale language models such as Qwen 3.8 Flash Next with 125 million parameters locally, without needing to send data to the cloud, although independent benchmarks confirming these claims have not yet been published.

As part of its on‑device AI strategy, Microsoft introduced Microsoft Execution Containers (MXC) in Windows 11, a secure execution environment for AI agents. The company also announced that Meta will ship a version of its Muse AI agent to Windows, allowing users to interact with an advanced AI assistant directly on the laptop without relying on cloud services.

Software and compatibility

The Surface Laptop Ultra runs Windows 11 on an ARM architecture. To ensure compatibility with the vast x86 application ecosystem, Microsoft uses the Prism emulation layer, which allows programs built for Intel or AMD processors to run on ARM hardware. In addition, thanks to the full CUDA stack support provided by RTX Spark, developers already working with NVIDIA GPUs can port their AI projects to the laptop without rewriting code.

Connectivity and design

In terms of connectivity, the Ultra features a magnetic USB‑C charging port – a solution similar to Apple’s MagSafe – as well as two standard USB‑C ports, a USB‑A port, an HDMI output, an SD‑card reader, and a 3.5 mm audio jack. The announced weight is approximately 4.4 lb (2 kg), and the chassis is built around the integrated ARM processor and NVIDIA chip, although battery capacity and endurance under intensive AI workloads have not been disclosed.

To encourage migration from Mac, Microsoft is offering a credit of up to US$1,000 in exchange for eligible MacBooks, reinforcing its ambition to compete directly with Apple’s high‑performance line. The company positions the Surface Laptop Ultra as a solution aimed at content‑creation professionals, software developers, engineers, and 3D work, highlighting it as an alternative to the 16‑inch MacBook Pro.

While the announcement brings ambitious promises, Gartner analysts place Microsoft outside the top six PC manufacturers by units shipped, suggesting the Ultra is part of a broader effort to grow the company’s market share in the hardware segment. Unverified aspects include the actual AI inference latency, battery life under AI loads, and the extent to which an ARM‑based system can match the performance of software that runs exclusively on x86. These metrics can only be confirmed after the official launch, but if the claims hold up, the Surface Laptop Ultra could change how professionals use on‑device AI, reducing reliance on cloud servers.