You are looking at the first mainstream Windows laptop that feels less like a “PC with a GPU” and more like a portable AI workstation with a keyboard attached. The Microsoft Surface Laptop Ultra, built around NVIDIA’s new RTX Spark platform, is Microsoft’s attempt to redefine what a creator or developer notebook should be in the age of local large language models, personal agents, and 12K timelines that no longer wait politely for progress bars.

At first glance, Surface Laptop Ultra looks like Microsoft’s familiar premium clamshell – clean lines, minimalist branding, and that slightly understated confidence the Surface lineup has leaned into for years. It comes in two finishes, Platinum and Nightfall, wrapped around a 15-inch chassis that somehow still manages to stay precise and light despite what is hiding inside. Open the lid, though, and it becomes pretty clear this machine was not designed for email, spreadsheets, and the occasional Netflix binge. It was built for people who treat a laptop like a portable production studio or a personal data center – developers, 3D artists, AI researchers, video editors, and anyone who thinks “let me just run that model locally” before they even consider the cloud.

At the heart of this machine is NVIDIA’s RTX Spark superchip, a fusion of NVIDIA’s new Blackwell RTX GPU and an ultra-efficient Grace CPU into a single SoC, manufactured on a 3-nanometer process and tied together with NVLink. Think of it as NVIDIA taking what it learned with its DGX Spark developer desktop – a box originally meant to sit next to data scientists and ML engineers – and shrinking that philosophy into something thin and light enough to go into a premium laptop. In practical terms, RTX Spark in this form factor gives you up to 6,144 CUDA cores, 20 CPU cores, and up to 128GB of unified memory on one chip, with around 1 petaflop of AI compute in FP4 – enough on-device horsepower to run 120-billion-parameter models with million-token contexts at interactive speeds, not “go-make-coffee” speeds.

Unified memory is the quiet revolution here. Traditional Windows laptops split RAM between CPU and discrete GPU, then make you pay the bandwidth and latency tax every time you shuttle tensors back and forth. RTX Spark’s unified pool means the Surface Laptop Ultra can dynamically allocate that memory to whatever needs it most – the GPU chewing through a diffusion model, the CPU orchestrating an agentic workflow, or both simultaneously. If you have ever watched your VRAM max out minutes into a training run or a render job while system RAM sat half empty, this design directly targets that pain. You are effectively getting workstation-style memory behavior in a machine that still looks at home in a lecture hall or a coffee shop.

Microsoft’s own positioning for Surface Laptop Ultra leans hard into this idea of “world makers” – people building systems, infrastructure, and tools that other people’s work will run on. It is more than a marketing line. The machine is tuned around workflows where “background” tasks are anything but: long compile cycles, large 3D scenes, multi-model AI pipelines, and agentic workloads that might run for hours at a time. This is also why Microsoft is pairing the Laptop Ultra with the Surface RTX Spark Dev Box on the desktop side, a mini-tower shaped like the top of an Xbox Series X that uses the same RTX Spark silicon with a higher thermal envelope to handle sustained AI jobs, training, and fine-tuning. Between the two, you are looking at an ecosystem where “local-first AI” is not a slogan – it is a design constraint.

That constraint is most obvious when you look at what Microsoft and NVIDIA are promising in terms of AI performance. With roughly 1 petaflop of AI compute and 128GB of unified memory available in its top configuration, Surface Laptop Ultra is specced to run models in the 120-billion-parameter range locally, with support for long context windows that make personal agents substantially more capable because they can keep more of your history and data in working memory. NVIDIA is explicitly framing RTX Spark as the hardware foundation for “personal AI computers,” where the PC is no longer just a tool you operate, but a teammate running agents that draft, analyze, design, and monitor in the background. For Windows, which has been racing to make Copilot+ and on-device AI feel like a native part of the OS rather than a bolt-on chatbot, this is the silicon that makes those ambitions plausible at laptop scale.

Of course, raw compute is only part of the story. If you are dropping what will likely be a high-three-figure or low-four-figure price on a creator-class machine, you care about the basics: screen, keyboard, trackpad, ports, thermals, and battery. On the display front, Surface Laptop Ultra swings for the fences. The 15-inch PixelSense Ultra panel uses mini-LED backlighting, hits up to 2,000 nits of peak HDR brightness, and delivers around 262 pixels per inch, making it the brightest display Microsoft has shipped on any Surface so far. For creators and color-critical work, the company is talking up “high-precision color accuracy,” which is table stakes in this category but still welcome, especially when paired with HDR workflows and high-res timelines.

The input story continues that trend. The haptic touchpad is the largest ever on a Surface, putting it more in line with the spacious glass pads on modern MacBooks and some of the better Windows ultrabooks. Haptic feedback gives Microsoft more control over how clicks feel and sound, and in theory, should also help with durability because there is no physical diving board mechanism to wear out or attract dust over time. The keyboard, while less talked about in the official launch material, is likely to stick with the familiar Surface Laptop formula – reasonably deep travel, satisfying but not loud, tuned for daily typing rather than gaming. It is the sort of layout you can comfortably draft a long-form article on without feeling like you are hammering against a glass slab.

One area where Surface Laptop Ultra noticeably diverges from the dongle-life trend is ports. Microsoft did the extremely un-2026 thing of simply putting “the ports you actually use” directly on the device: HDMI, USB-C, USB-A, an SD card slot, and a headphone jack. For photographers and video shooters, the SD card alone is a huge quality-of-life upgrade compared to laptops that make you fish for a card reader in your bag. For developers and IT folks, having USB-A and HDMI means legacy peripherals and projectors simply work without a dock or adapter circus. It sounds basic, but in a world where many premium notebooks have gone USB-C-only, this feels intentionally creator-centric rather than design-driven minimalism.

All of this power has to go somewhere thermally, and this is where the collaboration between Microsoft, NVIDIA, and the underlying Windows on Arm platform gets interesting. Traditional gaming laptops rely on brute-force cooling: big heatpipes, double-wide fans, and chassis designs that are thick, heavy, and loud. RTX Spark aims for something different. NVIDIA is targeting performance roughly comparable to an RTX 5070 Laptop GPU for graphics and gaming, but in a much more power-efficient envelope, thanks to the Arm-based design and unified memory. Microsoft stresses that Surface Laptop Ultra “does all of it quietly,” promising a combination of ultra-efficient CPU architecture and tuned thermals that lets creators push hard without the fans turning their workspace into a hair dryer. Engadget’s early look at the Dev Box notes a 100-watt thermal envelope on the desktop unit, while RTX Spark laptops sit in the 45- to 80-watt range – aggressive numbers for the class of workloads these machines are targeting.

Battery life is another pillar, especially for a device that is supposed to be both a portable AI workstation and a daily driver. Microsoft is claiming “all-day battery life” based on internal testing of pre-release units, with the usual caveats about usage and settings. That phrase is easy to dismiss as marketing, but here it is backed by an SoC that was designed from the ground up for efficiency and by NVIDIA’s public messaging around RTX Spark being suitable for slim laptops with all-day endurance, not just plugged-in monsters. If real-world reviews land anywhere near those claims under mixed workloads – say, a few hours of coding, some AI runs, meetings, and video playback – Surface Laptop Ultra could be one of the first AI-heavy laptops that does not force you to decide between “plugged-in performance” and “actually mobile.”

Zoom out, and Surface Laptop Ultra is part of a broader strategic pivot for both Microsoft and NVIDIA. For Microsoft, this is a showcase device for a new chapter of Windows PCs that are explicitly “accelerated by NVIDIA RTX Spark,” aimed at developers, creators, and businesses who want local AI capabilities without running everything in Azure all the time. For NVIDIA, it is a way to extend its data center dominance into client devices, turning Windows laptops and desktops into scaled-down AI nodes with familiar CUDA, TensorRT, and RTX tooling, backed by unified memory and serious FP4 throughput. If you are an AI developer, the pitch is straightforward: build and test agents, models, and pipelines on your desk or your lap, then burst to the cloud only when you actually need the extra scale.

The Dev Box completes that story. It uses the same class of RTX Spark superchip as the Laptop Ultra but in a compact aluminum chassis that acts as a heatsink, with a higher thermal envelope that is meant to sustain long-running training jobs, agent pipelines, and fine-tuning tasks. Microsoft is pre-configuring it with Windows 11 Pro tuned for developers – dark theme, cleaned-up taskbar, Do Not Disturb, Developer Mode, and tooling like Visual Studio Code and GitHub Copilot ready to go. In other words, the Laptop Ultra is the machine you carry into meetings, onto planes, and into coffee shops, while the Dev Box is the box that sits on or under your desk grinding away at the heavy jobs between commits.

For the broader PC industry, RTX Spark – and by extension laptops like Surface Laptop Ultra – is being treated as something of an earthquake moment. Early analysis compares its graphics performance to an RTX 5070 Laptop GPU while highlighting its AI performance as the real headline, effectively dragging server-class AI strength down into a client device and pairing it with a neural processing unit to handle Microsoft’s Copilot+ features. OEMs like HP are already lining up their own RTX Spark machines, positioning them as creator- and AI-first notebooks rather than just another spec bump. It is not hard to imagine that, a year from now, “does it support RTX Spark or an equivalent AI platform?” becomes as important a buying question for high-end laptops as “does it have a decent GPU?” was in the last decade.

Of course, none of this comes cheap. If NVIDIA’s DGX Spark pricing is any indication – originally around the mid-$4,000 range for a 128GB, multi-terabyte configuration before component shortages nudged it up – fully loaded RTX Spark systems tend to land in the “serious investment” bracket. Analysts expect top-tier RTX Spark laptops with the full 128GB unified memory to end up in the $3,000 to $4,000 range once OEMs start announcing actual SKUs. Microsoft has not detailed pricing for Surface Laptop Ultra yet, but given its positioning and the hardware inside, it is fair to think of it less as a consumer notebook and more as a mobile workstation for people and teams who can actually leverage that 1 petaflop-class compute.

So who is this machine really for? If your day consists of browser tabs, Office docs, some light photo editing, and the occasional indie game, Surface Laptop Ultra is overkill in the nicest possible way. You will be paying for an AI and graphics engine that spends most of its time idling. But if you are the kind of person who writes and debugs agents, runs 70B and 120B-parameter models locally, edits high-res footage, or treats your laptop as both a dev environment and a render node, this is the first Surface that feels like it was built specifically with your pain points in mind. You get workstation-style unified memory, serious GPU grunt, and a display and port setup that actually matches creator reality, all in a chassis that does not scream “gaming rig” the moment you walk into a client meeting.

The more interesting question is what this says about the future of Windows laptops in general. With RTX Spark, Microsoft and NVIDIA are effectively stating that the next era of PCs will be judged less on single-threaded CPU benchmarks and more on how much local AI work they can handle – how many agents can you run, how large a model can you fit, how long can you sustain that workload away from a wall socket. Surface Laptop Ultra is the first high-profile answer to that question. It looks like a familiar premium laptop, but under the hood, it is something more radical: a personal AI workstation that happens to fold shut and slip into a backpack.