What the Surface Laptop Ultra Is—and Why 1 Petaflop Matters
The Surface Laptop Ultra is a slim 15-inch Windows on Arm notebook built around NVIDIA’s RTX Spark chip that delivers 1-petaflop AI workstation performance in a portable chassis, enabling large on-device models, creator workflows, and gaming without depending on the cloud. That 1-petaflop figure refers to the laptop’s ability to perform one quadrillion floating-point operations per second for AI-style workloads, putting it into traditional workstation territory despite its sub-18 mm profile. In practice, this means the machine can run complex generative models, real-time video effects, and multi-layer 3D scenes locally instead of round-tripping heavy data to remote servers. For AI developers and creative professionals, petaflop computing portable in a standard laptop form factor changes how you size projects: models that once required a desktop tower or dedicated server can now be trained, fine-tuned, and deployed from a carry-on-friendly AI workstation laptop.
Inside the RTX Spark Chip: AI Workstation Power in Laptop Form
At the core of the Surface Laptop Ultra specs is NVIDIA’s RTX Spark, an Arm-based superchip that merges CPU and GPU into a single AI accelerator. RTX Spark combines a 20-core Grace CPU with 6,144 Blackwell RTX cores, connected through NVLink so both processors share up to 128GB of unified memory instead of splitting RAM and VRAM. This design removes many of the memory bottlenecks that slow large models and complex renders on traditional laptops. According to PCMag, the Surface Laptop Ultra “offers the template for a new consumer category that can drive a full petaflop of AI-ready compute power.” For professionals, that means faster inference, smoother multi-app pipelines, and less need to downscale assets or prune models just to fit into limited GPU memory. Copilot+ features are backed by an integrated NPU, while the full CUDA core array is available for AI frameworks and GPU-accelerated apps.
120-Billion-Parameter Local Models and an AI-First Software Stack
The standout RTX Spark chip performance metric is its support for running AI models with up to 120 billion parameters locally, with no cloud tether. Unified memory lets these models sit in a single address space, so workloads like code assistants, diffusion image generators, and agentic AI pipelines can run concurrently without crashing into memory limits. Microsoft is aligning the software stack to match. Windows on Arm gains kernel-level execution for agentic AI and smarter memory management tuned for the 128GB unified pool, dynamically assigning resources to GPU and AI-heavy tasks. The Prism emulation layer now taps RTX Spark’s GPU, giving x86-bound tools a performance lift instead of relying only on CPU emulation. Adobe has reworked Premiere Pro and Photoshop for native RTX Spark support, and apps like Blender, DaVinci Resolve, Cinema 4D, CapCut, Topaz Photo and Affinity by Canva already run natively on this Windows on Arm AI PC.
Slim Chassis, Mini-LED Display and Gaming on Windows on Arm
Despite its AI workstation laptop performance, the Surface Laptop Ultra keeps a slim profile: under 18 mm thick, under 2 kg in weight, and less than 4.5 pounds according to PCMag, with an all-metal chassis in Platinum or Nightfall finishes. Microsoft claims all-day battery life, though heavy AI and gaming workloads will shorten that. The 15-inch mini-LED PixelSense Ultra display reaches up to 2,000 nits peak HDR brightness at 262 pixels per inch, making HDR grading and outdoor use more practical. A 30%-larger haptic trackpad, HDMI, USB-C, USB-A, full-size SD card reader, headphone jack, and a user-replaceable SSD round out creator-friendly hardware. On the gaming side, Riot Games’ League of Legends and Valorant plus titles like PUBG and Alan Wake 2 have confirmed support, using a mix of native Arm builds, GPU-accelerated Prism emulation, and new native anti-cheat on Arm to make competitive play viable on this Windows on Arm AI PC.
Computex Launch and the Shift to AI-First Portable Computing
By unveiling the Surface Laptop Ultra at Computex, Microsoft is signaling that AI-first portable computing is no longer an experiment but its main direction for premium PCs. Earlier Surface models used Qualcomm Snapdragon X Arm processors; this generation pivots to RTX Spark as the flagship for combining Copilot+ features with petaflop-grade local compute. Microsoft is also reframing who this machine is for. Brett Ostrum described the target users as those “building the systems, the breakthroughs and the infrastructure the world runs on and gets changed by,” highlighting developers, researchers, and high-end creators rather than casual users. Internally, leaders are equally blunt: Surface Product Leader Andrew Hill calls it “the most powerful thing we’ve ever made.” Shipping later in 2026 with pricing still unannounced, the Surface Laptop Ultra sets a new bar for what a thin-and-light AI workstation laptop can be in everyday professional workflows.