What the Surface Laptop Ultra Is and Who It’s For
The Surface Laptop Ultra is a slim AI-focused Windows on Arm notebook that combines a 20-core Arm CPU, an NVIDIA Blackwell GPU, up to 128GB of unified memory, and a 15-inch mini-LED display to deliver desktop-class local AI computing in a portable chassis. Microsoft positions it as more than a thin-and-light email machine: it is a 1-petaflop AI workstation aimed at developers, creators, and technologists who want to run large models locally instead of depending on cloud GPUs. According to Microsoft’s Brett Ostrum, it targets “those building the systems, the breakthroughs and the infrastructure the world runs on and gets changed by.” It is also the first Surface laptop powered by NVIDIA silicon since the early Surface RT era, marking a notable architectural shift toward Windows on Arm and unified memory for high-performance workloads.
RTX Spark Specs: 20 Arm Cores, Blackwell GPU and Unified Memory
At the heart of the Surface Laptop Ultra is NVIDIA’s RTX Spark chip, a package that combines a 20-core Arm CPU with a Blackwell GPU tied together through NVLink C2C. The Blackwell GPU delivers 6,144 CUDA cores, giving this 128GB RAM laptop serious parallel compute for AI inference, 3D work, and GPU-heavy content creation. Unified LPDDR5X memory reaches up to 128GB with as much as 300 GB/s of bandwidth, shared dynamically between CPU and GPU so RTX Spark specs do not rely on separate VRAM pools. This design removes many of the typical bottlenecks when moving tensors or textures between processor and graphics memory. On the software side, Windows on Arm and Microsoft’s Prism emulator now support RTX Spark’s GPU, so even x86 apps can tap into the Blackwell GPU CUDA cores for acceleration where supported.
1-Petaflop Local AI Computing and 120B-Parameter Models
Surface Laptop Ultra performance is defined by its AI ceiling as much as its raw CPU and GPU numbers. RTX Spark can hit up to 1 petaflop of AI compute in this sub-18 mm chassis, turning the machine into a compact AI workstation. Microsoft and NVIDIA say this is enough to run AI models with up to 120 billion parameters entirely on-device, which shifts workloads that once demanded cloud clusters onto a single notebook. For practitioners, that means fine-tuning large language models, running diffusion image generators, or testing multimodal pipelines without sending data off the machine. Local AI computing cuts latency and gives more predictable performance when networks are unreliable or constrained. It also has privacy advantages for sensitive datasets, since prompts, outputs, and intermediate embeddings never leave the laptop unless you choose to sync them elsewhere.
Display, Design and Thermals: More Than an AI Dev Box
While the compute story stands out, Microsoft also tuned the hardware around it. The 15-inch mini-LED PixelSense Ultra panel runs at 2880 x 1920 with a density of 262 ppi and peaks at 2,000 nits in HDR, making it the brightest Surface Laptop display to date and well suited to HDR grading or outdoor use. The chassis stays under 4.5 pounds and under 18 mm thick, yet still fits HDMI, USB-C, USB-A, a full-size SD card reader, and a headphone jack, plus a user-replaceable SSD. A larger haptic trackpad improves everyday ergonomics. To keep this 1-petaflop AI setup under control, Microsoft worked with NVIDIA on the Microsoft Power and Thermal Framework and uses a dual-fan cooling system so sustained GPU and CPU loads can run longer without throttling during AI training runs or long 3D renders.
Windows on Arm, App Ecosystem and RTX Spark Use Cases
The Surface Laptop Ultra runs Windows on Arm, but its app library is growing around RTX Spark. Adobe Photoshop and Premiere Pro have been rearchitected as native Arm builds and tuned specifically for RTX Spark, with new GPU-driven pipelines for live filters, compositing, and real-time video editing. Other creative tools such as Blender, DaVinci Resolve, Cinema4D, Redshift, Topaz Photo, CapCut, Cubase, and Affinity by Canva also run natively, so many AI-augmented workflows already benefit from the Blackwell GPU CUDA cores. For software that still depends on x86, Microsoft’s Prism emulator can now tap RTX Spark’s GPU, smoothing the transition. On the entertainment side, titles like League of Legends, Valorant, PUBG, Alan Wake 2, Naraka: Bladepoint, and War Thunder are confirmed, though gaming support continues to expand as more developers ship Arm-compatible builds or enhanced emulation paths.