Microsoft is taking the Surface brand into decidedly more powerful territory. The company has opened preorders for two new Surface devices built around NVIDIA’s RTX Spark platform: the Surface Laptop Ultra and Surface RTX Spark Dev Box.

The new machines represent Microsoft’s push to make Windows a more serious platform for developers, creators and AI builders who want to run demanding workloads locally instead of sending everything to the cloud. The Surface Laptop Ultra starts at $2,599.99, while the desktop-style Surface RTX Spark Dev Box starts at $5,999.99. The Laptop Ultra will begin shipping October 16, while the Dev Box is scheduled to ship in November.

This is a very different kind of Surface launch from the company’s usual thin-and-light laptops and 2-in-1s. Microsoft isn’t pitching these machines primarily as everyday productivity PCs. Instead, it is positioning them as “Builder PCs” for people working with local AI models, professional creative applications, development tools and demanding graphics workloads.

Surface Laptop Ultra is the star of the lineup

The Surface Laptop Ultra is Microsoft’s most powerful Surface Laptop yet, and its defining feature is NVIDIA’s RTX Spark N1X superchip.

The platform combines an NVIDIA Grace CPU with up to 20 cores and a Blackwell-based RTX GPU with up to 6,144 cores. The system can also be configured with as much as 128GB of unified memory shared between the CPU and GPU. Microsoft says the machine can run AI models exceeding 120 billion parameters locally and deliver up to 1 petaflop of AI performance under the company’s stated FP4 measurement.

That unified-memory architecture is particularly important for local AI. Instead of treating system memory and graphics memory as separate pools, RTX Spark allows the processor and GPU to share the same large memory pool. That gives demanding AI workloads substantially more room to operate without requiring the traditional discrete-GPU setup found in many high-end workstations.

Microsoft is also leaning heavily into NVIDIA’s CUDA ecosystem, which should make the Surface Laptop Ultra considerably more interesting to developers already building AI applications around NVIDIA’s software stack.

The company says users can prototype agents, review code, test larger models and keep their development environments running alongside local AI workloads. When the workload gets too large for the machine, Microsoft’s broader hybrid-intelligence approach allows cloud computing to take over rather than forcing everything to remain local.

The hardware is still unmistakably Surface

Despite the workstation-class ambitions, Microsoft hasn’t turned the Laptop Ultra into a giant gaming laptop.

The machine is less than 18mm thick and weighs under 4.5 pounds, with Microsoft’s specifications listing a 4.41-pound weight. It uses a new thermal architecture that Microsoft says provides up to 2.5 times the thermal capacity of its current Surface Laptop generation. That should matter when the RTX Spark hardware is being pushed continuously rather than briefly during a benchmark.

The 15-inch PixelSense Ultra touchscreen is another major upgrade. It reaches up to 2,000 nits of peak HDR brightness, with Microsoft calling it the brightest peak HDR display it has put in a Surface Laptop. The screen supports 100% DCI-P3 color coverage, Dolby Vision IQ and a 120Hz refresh rate, while the display is designed for both professional creative work and gaming.

Microsoft has also enlarged the precision haptic touchpad by 30% compared with the Surface Laptop 8th Edition. That may sound like a small change next to a 128GB memory configuration, but it makes sense for a machine aimed at video editors, designers and developers who will spend hours navigating complex applications.

Finally, a Surface with ports for grown-up workloads

Microsoft has also gone unusually generous with connectivity.

The Surface Laptop Ultra has two additional USB-C/USB4 ports alongside a third USB-C/USB4 port with the company’s new Magnetic Connect charging system. It also includes HDMI 2.1b, USB-A, a full-size SD card reader and a 3.5mm headphone jack. The laptop can drive up to three 4K external displays.

The magnetic USB-C connection is an interesting twist on the traditional Surface Connect approach. Microsoft describes it as the world’s first built-in magnetic USB-C charging port on a laptop. The magnetic cable attaches to the right-side USB-C port, while the port remains capable of handling ordinary USB-C data and video connections.

That combination of USB-C, HDMI, USB-A and SD makes the Laptop Ultra much more practical for professional workflows than a machine that requires a collection of dongles just to connect a camera, monitor and storage drive.

The SSD is also removable, with configurations ranging from a 512GB Gen 4 SSD to 1TB and 2TB Gen 5 options. Microsoft says replacement components and service options are available, although availability can vary by market and over time.

There are a lot of configurations — and they get expensive fast

The $2,599.99 starting price gets you an RTX Spark N1X configuration with an 18-core CPU, 5,120-core GPU, 24GB of unified memory and a 512GB SSD running Windows 11 Home.

From there, Microsoft offers considerably more powerful configurations. The 20-core CPU and 6,144-core GPU version starts at $3,699.99, while memory options extend through 32GB, 48GB, 64GB and 128GB. Storage can reach 2TB.

At the extreme end, a Nightfall configuration with 128GB of unified memory and a 1TB SSD costs $5,899.99. A 64GB/2TB configuration comes in at $4,699.99, while the 64GB/1TB model costs $4,299.99.

Microsoft is offering the Laptop Ultra in Platinum and Nightfall, although colors and configurations vary depending on the model.

That pricing makes one thing clear: this isn’t Microsoft’s answer to the MacBook Air. The Surface Laptop Ultra is aimed at a much narrower audience that actually needs enormous amounts of unified memory and serious local compute.

Microsoft also has a $6,000 desktop

For people who don’t need the Laptop Ultra’s portability, Microsoft is introducing the Surface RTX Spark Dev Box.

It’s essentially the desktop counterpart to the Laptop Ultra, built around the same RTX Spark platform but designed to sit on a developer’s desk. The configuration Microsoft is selling for $5,999.99 includes the 6,144-core GPU, 20-core CPU, 128GB of unified memory and a 2TB SSD.

Microsoft is targeting the Dev Box directly at AI engineers, developers and technical teams. It comes prepared with tools including Visual Studio Code, Git, GitHub CLI, GitHub Copilot, Windows Subsystem for Linux, Python and Node, so the idea is to get developers working rather than spending hours configuring an AI development environment.

The Dev Box also makes Microsoft’s broader argument about local AI more obvious. If developers can perform inference, model evaluation and experimentation on their own hardware, they don’t necessarily need to send every development task to a cloud GPU.

That’s not about eliminating cloud computing. Microsoft is instead pitching a hybrid model in which local hardware handles the workloads it can handle efficiently, while cloud infrastructure remains available for jobs that need more compute.

Microsoft wants Windows to become the home for local AI

The Surface Laptop Ultra is only one piece of a much larger RTX Spark push.

Microsoft says RTX Spark PCs are being introduced across an ecosystem of manufacturers including ASUS, Dell, HP, Lenovo and MSI. The company claims these systems can deliver up to 2.1 times faster time to first token, 4.3 times faster AI image generation and 6.2 times faster AI video generation than a 16-inch MacBook Pro with Apple’s M5 Pro, although those figures come from Microsoft’s own testing and should be treated accordingly.

That comparison tells you exactly who Microsoft has in its sights.

For years, the MacBook Pro has been an attractive option for developers and creative professionals who want workstation-class performance in a relatively portable machine. Microsoft now wants Windows to have an answer that combines that portability with NVIDIA’s massive AI software ecosystem.

The Surface Laptop Ultra is consequently less about replacing an ordinary Surface Laptop and more about creating a new tier above it.

Local AI is the real reason this machine exists

The biggest shift here isn’t the screen, the ports or even the 128GB memory option. It’s Microsoft’s attempt to make powerful local AI a normal part of a professional Windows workflow.

The company says the Laptop Ultra can run models exceeding 120 billion parameters locally. That doesn’t mean every 120-billion-parameter model will magically run at useful speeds, but it demonstrates how far beyond conventional Copilot+ PCs this hardware is intended to go.

Microsoft is also building security controls around local AI agents. Microsoft Execution Containers are designed to contain agents and restrict what they can access, while future Windows capabilities are expected to use Microsoft Entra to distinguish agent activity from user activity and extend Microsoft’s Agent 365 controls to local agents.

That’s an important piece of the puzzle. Giving an AI agent access to powerful local hardware is one thing. Giving it access to a user’s files, development environment and business data is another.

Microsoft clearly sees the PC becoming an active participant in AI workflows rather than simply being the screen through which users access cloud-based assistants.

A $2,599 Surface is a very different Surface

The Surface Laptop Ultra is an unusual product for Microsoft because its biggest selling point isn’t convenience. It’s excess.

You can buy much cheaper Windows laptops for everyday work. You can buy a MacBook for less. You can buy a conventional gaming laptop with a powerful GPU. And you can certainly build a desktop workstation for $5,899.

But none of those comparisons really capture what Microsoft is attempting here.

The Surface Laptop Ultra combines a large unified-memory pool, NVIDIA’s Blackwell graphics architecture, a high-end ARM CPU, CUDA support, a 15-inch 2,000-nit display, a relatively thin chassis and a surprisingly complete set of ports. It is effectively Microsoft’s attempt to squeeze a local AI workstation into something that still behaves like a premium laptop.

Whether that combination is worth the price will depend heavily on the workloads. For someone writing documents, browsing the web and occasionally asking Copilot a question, absolutely not. For a developer running large models locally, a creative professional working with demanding applications, or an AI engineer who wants serious compute without constantly renting cloud GPUs, the equation looks very different.

Microsoft is now taking preorders for the Surface Laptop Ultra starting at $2,599.99, with availability beginning October 16. The Surface RTX Spark Dev Box is also available for preorder in the U.S. for $5,999.99, with shipments expected to begin in November.

The bigger story, though, is what these machines represent. Microsoft isn’t simply making the fastest Surface yet. With NVIDIA’s RTX Spark, it’s trying to turn Windows into a serious local AI platform for people who need their PCs to do considerably more than run Office and a web browser.