Microsoft has introduced its most powerful and brightest laptop in its history, the Surface Laptop Ultra, equipped with the NVIDIA RTX Spark chip.
Once upon a time, Microsoft gambled on using an NVIDIA chip based on ARM architecture for its first flagship Surface tablet and ultimately had to book a $900 million loss; but today, the company is taking its chances again. The Redmond-based firm unveiled the Surface Laptop Ultra, a computer whose beating heart is NVIDIA’s new ARM-based chip.
This article was updated on September 6 with official and new information regarding the specifications of the Surface Laptop Ultra.
In a sense, the Surface Laptop Ultra also serves as Microsoft’s response to a long-standing demand. For years, a group of professional users has been asking for a powerful, cool, and bezel-free laptop; a device that, without the experimental design and special hinge of the Surface Laptop Studio, focuses primarily on raw power, thermal efficiency, and the classic laptop form factor. Now that production of the Studio model has been discontinued, the Surface Laptop Ultra fills the void left by a truly professional Surface.
Andrew Hill, head of Microsoft’s Surface division, states with confidence in response to a question about the positioning of the Surface Laptop Ultra: “This is the most powerful product we have ever made.”
Now, most of the uncertainties surrounding the Surface Laptop Ultra have been resolved; following its initial reveal, Microsoft announced details on pricing, configurations, and release timing at its October 7 (October 7) event. The price of the most powerful laptop in Surface history starts at $2,599.99 for the base model and reaches up to $5,899.99 in the highest-end configuration; pre-orders are now open, and availability begins on October 16, 2026 (October 16, 2026).
NVIDIA RTX Spark Superchip in Surface Form
The Surface Laptop Ultra is powered by NVIDIA’s new “superchip,” known as the RTX Spark; a processor that appears to be nearly identical to the module used in the DGX Spark mini-PC; but is now optimized for Windows 11.
The Surface Laptop Ultra is part of a new wave of devices based on the NVIDIA N1x platform; a platform that combines an ARM CPU, RTX graphics, and unified memory in a single package. The use of Blackwell architecture also ensures the laptop is targeted for heavy AI workloads, rendering, long compilations, and professional CUDA-based tasks; an area that has not been pursued as seriously in the Windows ARM world before.
A Petaflop of AI Processing Power in a Surface Device
The RTX Spark chip in the Surface Laptop Ultra is available in two variants. The more powerful model features a 20-core Grace CPU and Blackwell graphics with 6,144 CUDA cores, supporting up to 128 GB of integrated LPDDR5X memory. The lesser variant uses an 18-core CPU and 5,120-core graphics, supporting up to 64 GB of memory. NVIDIA has specified a power range of 45 to 80 watts for both laptop models.
In the 20-core model, the CPU is built on the Arm v9.2 instruction set architecture, utilizing 10 powerful Cortex-X925 cores and 10 power-efficient Cortex-A725 cores. The CPU and GPU are connected via the high-speed NVLink-C2C interface, granting access to a shared memory pool. The memory utilizes a 256-bit interface, achieving bandwidth of approximately 300 GB/s. NVIDIA manufactures the RTX Spark using TSMC's 3-nanometer process and reports a transistor count of around 70 billion.
The base configuration of the $2,599.99 Surface Laptop Ultra is equipped with the 18-core RTX Spark, 24 GB of memory, and a 512 GB SSD. Conversely, the $5,899.99 model utilizes the full 20-core variant, 128 GB of unified memory, and a 1 TB SSD; therefore, contrary to initial reports, the base Surface Laptop Ultra model is not available with 16 GB of RAM.
One of the most significant differences between the Surface Laptop Ultra and previous Surfaces is its approach to memory. As mentioned, the top-tier configuration provides 128GB of unified memory shared between the CPU and GPU, allowing the system to dynamically reallocate this memory across different subsystems depending on the type of workload.
When initially announced, Microsoft stated that the Surface Laptop Ultra could run 120-billion-parameter AI models with a one-million-token context locally; however, it has now gone further. According to Microsoft, the device can even handle the local execution of compressed versions of 280-billion-parameter models, which until just a year ago practically required cloud infrastructure to run.
The 128GB of memory places the Surface into the realm of workstations
The Blackwell RTX Spark graphics in the highest configuration delivers up to one petaflops of FP4 compute power for AI workloads. To put this number in perspective, NVIDIA states that RTX Spark can render 3D scenes exceeding 90GB in size using OptiX, edit 12K video with 4:2:2 sampling, create 4K AI video, and run 120-billion-parameter models locally.
Internal images of the device also reveal that Microsoft has adopted a more robust cooling architecture to manage heat. The Surface Laptop Ultra features two large fans, indicating that the device was not built merely to showcase paper specifications but is designed to deliver stable performance under sustained pressure, ranging from rendering and compilation to running AI models.
Now Microsoft also provides a specific figure for the new cooling system; the Surface Laptop Ultra’s thermal system can deliver up to 2.5 times greater heat rejection capacity compared to the seventh-generation Surface Laptop 15. The smaller motherboard design of the RTX Spark has also allowed Microsoft to allocate more space to fans and the battery.
RTX Spark Chip vs. M5 Pro
At the full announcement event for the Surface Laptop Ultra, Microsoft turned to one of its most important rivals, the MacBook Pro M5 Pro, to demonstrate the RTX Spark’s power. According to a slide displayed at the event, which showed NVIDIA’s preliminary test results, Windows PCs equipped with RTX Spark can reduce the Time to First Token for AI processing by up to 2.1 times, generate images up to 4.3 times faster, and produce videos up to 6.2 times faster than the M5 Pro.
NVIDIA claims the RTX Spark’s advantage is not limited to AI model processing. The chip giant speaks of running AAA games at 1440p resolution with over 100 frames per second, leveraging Ray Tracing, DLSS, and Reflex. Adobe is also redesigning parts of Photoshop and Premiere to better utilize the new architecture. According to NVIDIA’s claim, optimized versions of Adobe applications can perform up to two times faster in a set of graphics and AI processing tasks; however, this figure is not a direct comparison with the M5 Pro.
Brilliant Display and Unprecedented Trackpad
Along with the powerful RTX Spark chip, which promises “all-day battery life,” graphics performance on par with the RTX 5070 laptop processor, and one petaflop of AI processing power, the Surface Laptop Ultra also features a 15-inch Mini-LED touchscreen with a 262 pixel-per-inch density.
Microsoft has also made the Surface trackpad and display “Ultra”
Microsoft claims that the panel used in its new laptop, with a maximum brightness of 2,000 nits in HDR mode, is “the brightest display we have ever used in our products”; the 15-inch Mini-LED PixelSense Ultra display has a 3:2 aspect ratio, and Microsoft states that it was designed for users who make color- and exposure-critical decisions during their workflow.
Microsoft says that in addition to the brightest display in its laptop history, the Surface Laptop Ultra also features the largest haptic trackpad in the history of Surface products. We now have a more precise figure: the new trackpad is more than 30 percent larger in area than the one used in the seventh-generation Surface Laptop, providing haptic feedback for precise pointer control in professional software.
Microsoft does not yet provide a specific battery life figure for the Laptop Ultra and simply uses the phrase “all-day battery life”; however, a more important claim is that the laptop maintains its performance level even when unplugged from power. Such a feature is particularly important for a system with a powerful GPU, as many professional Windows laptops impose significantly stricter power limits when disconnected from power.
Design, ports, and ambitious claims
Microsoft states that the Surface Laptop Ultra will be available in two colors: Dark Gray (Nightfall) and Silver (Platinum), weighing less than 2 kilograms. The device's thickness has also been revealed; despite its powerful hardware and dual-fan cooling system, the laptop's aluminum body is less than 18 millimeters thick.
The laptop's port selection includes three USB-C ports, one USB-A, HDMI output, a full-size SD card reader, and a 3.5mm headphone jack. Consequently, the initial guess regarding the presence of three USB-C ports has proven correct; Microsoft has ensured the port array is comprehensive enough that professional users will not need dongles for many daily applications.
One of the most interesting details revealed in the full unveiling is a new magnetic charging system that Microsoft has integrated around one of the USB-C ports. Unlike the older Surface Connect, the connector itself remains a standard USB-C; however, the charging cable attaches to the port's perimeter using a magnet. This allows the same port to remain usable for data and video transfer.
The device's charger offers 140 watts of power, and the cable can even be reversed to plug into the adapter, allowing the other end to be used as a standard USB-C cable for charging other devices.
Repairability is also a more prominent feature of the Laptop Ultra design. Internal components are designed for easier disassembly rather than extensive use of adhesives, and Redmond’s team has included signs and QR codes inside the device to guide repair technicians. A repair guide and spare parts will also be provided, and the laptop’s SSD will be replaceable by users or qualified repair professionals.
Microsoft has finally heard content creators’ voices.
The Surface Laptop Ultra is not the only device arriving this autumn with NVIDIA’s new chips; however, Microsoft plays a deep role in the success of other laptops and mini PCs equipped with RTX Spark as well.
Microsoft and NVIDIA have announced that they have collaborated for years to fully prepare Windows for ARM-based devices and specifically for the RTX Spark chip. A significant part of this collaboration relates to the new generation of AI agents; assistants that do more than just respond, as they can perform tasks on behalf of users, such as fixing code bugs, managing workflows, or analyzing large datasets.
Microsoft is speaking about new security and restriction mechanisms in Windows for the future of agent-based artificial intelligence; mechanisms intended to keep agents in a controlled environment, make their data access levels transparent, and prevent them from directly interfering with critical parts of the operating system.
Nvidia is also moving in the same direction, bringing NVIDIA OpenShell to Windows; a capability built on these new security foundations that allows developers and content creators to run AI agents with large contextual data on their devices while maintaining privacy and local execution.
Microsoft has referenced in its blog post the structural changes to Windows to fully leverage the power of RTX Spark, and has also mentioned multiple developers who have been convinced to support Windows on Arm architecture.
Part of the changes has occurred at the Windows Scheduler level. The Redmond team has optimized Workload Profile Scheduling (WPS) for the heterogeneous RTX Spark architecture so that Windows can distribute processing more intelligently among the 20 CPU cores. Support for the Microsoft Power and Thermal Framework (MPTF) is also available for simultaneously managing the chip's power consumption, performance, and temperature during long workloads on RTX Spark.
AI agents without security restraints are more dangerous than useful, and Microsoft knows this.
Windows has also been optimized for the unified memory architecture of RTX Spark; Microsoft has increased the limit on system memory that the GPU can access and improved the management of memory page sizes in shared regions. The result should enable loading larger AI models and heavier rendering projects without the need for continuous data copying between CPU and GPU memory. Windows ML also provides the ability to natively use TensorRT for AI processing.
The Windows on Arm software ecosystem is now much broader than in previous generations. Applications such as Blender, DaVinci Resolve, Cinema 4D, Redshift, Topaz Photo, CapCut, Cubase, Bitwig Studio, and Affinity run natively on Arm, and major Adobe applications, including Photoshop and Premiere, also have native versions. MATLAB currently offers official support for Windows on Arm via Prism.
Nvidia's entry into the Windows on Arm world is important not just for Microsoft but changes the entire equation for Windows on Arm. Qualcomm has been the key player in this space, with Snapdragon X chips holding the largest share of Windows on Arm devices; however, Nvidia's presence, particularly in workstation laptops and hybrid-gaming devices, can further incentivize developers to optimize their apps and games for Arm—an outcome that ultimately benefits Qualcomm as well.
On the other hand, Intel and AMD now face fresh pressure. If part of the Windows laptop market’s future hinges on the combination of Arm, powerful graphics, and local AI processing, the traditional dominance of the x86 architecture in personal computers will be challenged like never before.
Microsoft is also working to address one of Windows on ARM's longstanding weaknesses: game execution. Riot Games plans to natively release games such as Valorant and League of Legends for this platform, and KRAFTON is adding PUBG: Battlegrounds to its list of supported titles. Support for native anti-cheat solutions like Easy Anti-Cheat and BattlEye is also crucial, as the lack of this infrastructure previously kept many online games away from Windows on ARM.
The Surface Laptop Ultra is not a gaming device; however, it intends to take gaming seriously.
Microsoft has also optimized the Prism emulation layer for the RTX Spark architecture to ensure that x86 games and applications without native ARM versions run more efficiently. Despite its Blackwell-based graphics, the Surface Laptop Ultra will not be officially classified as a gaming laptop; however, Microsoft aims to demonstrate that Windows on ARM no longer necessarily implies significant compromises in running games and professional software.
The Surface Laptop Ultra is available in a total of eight configurations. The version equipped with the RTX Spark N1X, featuring an 18-core CPU and a 5120-core CUDA GPU, starts at $2,599.99 for the model with 24GB of memory and a 512GB SSD. The same version with 24GB of memory and a 1TB SSD costs $2,999.99, while the model with 32GB of memory and a 1TB SSD is priced at $3,299.99.
The more complete RTX Spark N1X with a 20-core CPU and a 6144-core CUDA GPU starts at $3,699.99 for the 32 GB memory and 1 TB SSD model. The 48 GB and 64 GB models with a 1 TB SSD are priced at $3,999.99 and $4,299.99 respectively, while the 64 GB version with a 2 TB SSD costs $4,699.99. At the top of the family sits the 128 GB model with a 1 TB SSD, priced at $5,899.99.
The Surface Laptop Ultra's price tag clearly indicates that Microsoft does not intend to position its new product merely as a faster version of the regular Surface Laptop. The primary target is the MacBook Pro domain and portable workstations; a market where GPU power, available graphics processing memory, CUDA support, execution of large AI models, and stable performance under sustained loads become more critical.
On paper, the combination of 128 GB unified memory, Blackwell graphics, the CUDA ecosystem, and a chassis thinner than 18 mm appears attractive; even Nvidia's initial figures against the M5 Pro are impressive; however, the real test begins on October 16; that is when it will become clear to what extent Microsoft and Nvidia have been able to deliver on their difficult promise: the power of a workstation in the form of an ARM laptop that should not force the user to compromise when unplugging the power cord or running legacy Windows applications.
Comments