NVIDIA RTX Spark marks a major shift for Windows PCs by combining a 20-core NVIDIA Grace Arm CPU, Blackwell RTX graphics, and up to 128GB of unified memory inside one superchip. Built for Windows 11 on Arm, it targets local AI, creator workflows, and high-end gaming, with support for large on-device AI models, NVIDIA’s CUDA ecosystem, DLSS, ray tracing, and Copilot+ PC features. The first RTX Spark laptops and compact desktops are expected in fall 2026, but pricing, battery life, and sustained real-world performance remain the key details to watch before buying.
The NVIDIA RTX Spark
If you've heard the name and you're not sure what it means in practice (a chip, a laptop, a piece of software, or a rival to something you already own), this is a clear, no-hype breakdown of what the RTX Spark is, what it's for, why it matters, and whether it's worth waiting for.
NVIDIA RTX Spark: key specs at a glance Type: Arm-based system-on-chip (SoC) for Windows on ArmCPU: 20-core NVIDIA Grace (Arm), co-designed with MediaTekGPU: NVIDIA Blackwell RTX, 6,144 CUDA cores, fifth-generation Tensor Cores (FP4)CPU-to-GPU link: NVIDIA NVLink-C2C interconnectMemory: Up to 128GB unified (shared by CPU and GPU)AI performance: Up to 1 petaflop (FP4)Operating system: Windows 11 (Copilot+ PC), with x86 and x86-64 apps via the Prism emulatorDevices: Slim laptops (from 14mm, around three pounds, 14 to 16 inches) and compact desktopsAvailability: Fall 2026Price: No official price yet; a Taiwan-based report estimates the first laptops above US$4,000 So, what exactly is the NVIDIA RTX Spark?
The RTX Spark is a superchip , NVIDIA's term for a single processor that fuses together components that, until now, usually lived as separate parts inside your computer.
In a typical laptop, you have a CPU from one company (Intel, AMD, or Qualcomm) and, if you're lucky, a separate NVIDIA graphics chip alongside it. The RTX Spark collapses both into one piece of silicon:
A 20-core NVIDIA Grace CPU , built on the Arm architecture (the same broad family of chips that powers smartphones and Apple's M-series Macs), with custom CPU design work done in collaboration with MediaTek. A Blackwell RTX GPU with 6,144 CUDA cores and fifth-generation Tensor Cores, NVIDIA's latest graphics and AI engine. The two halves are linked by NVLink-C2C , NVIDIA's high-speed chip-to-chip interconnect, and they share a single pool of up to 128GB of unified memory . That last point is the one to remember. In a normal PC, your system memory (RAM) and your graphics memory (VRAM) are separate, and a typical gaming laptop might have only 8 to 16GB dedicated to the GPU. The RTX Spark instead gives the CPU and GPU a shared memory pool of up to 128GB that either can address directly. As we'll see, that single design choice unlocks much of what makes this chip interesting.
NVIDIA rates the whole package at up to 1 petaflop of AI performance (using the efficient FP4 number format) while still being designed to fit in slim, battery-friendly laptops.
Crucially, the RTX Spark is built to run Windows on Arm , a version of Windows designed for Arm-based processors. NVIDIA has dabbled in Arm-powered Windows hardware before (it supplied the chip in Microsoft's old Surface RT over a decade ago), but the RTX Spark is a far more serious effort, developed hand-in-hand with Microsoft.
RTX Spark vs. DGX Spark: what's the difference?
Because NVIDIA has reused the "Spark" name, it's easy to mix up two very different products:
DGX Spark is a small, NVIDIA-built desktop appliance for AI developers and researchers. It runs Linux, is built around the GB10 Grace Blackwell Superchip, and sells for roughly US$3,500 to US$4,700 depending on configuration. It is a specialist tool for prototyping AI models.RTX Spark , the subject of this article, is a chip that goes inside consumer and prosumer Windows laptops and compact desktops made by Microsoft and a range of major PC manufacturers, Acer among them.The two are built on closely related underlying silicon, but they target completely different buyers and operating systems. If you want a polished Windows laptop, you want RTX Spark. If you want a dedicated Linux AI dev box, that's DGX Spark.
What is the RTX Spark built for? NVIDIA is positioning the RTX Spark as a genuine all-rounder aimed at three audiences at once: AI users, creators, and gamers.
AI and personal agents This is the headline use case, and it's where the chip's unique design shows up most clearly. Thanks to its large unified memory, NVIDIA says an RTX Spark machine can run a 120-billion-parameter large language model with up to a one-million-token context window locally , meaning a serious AI model running on your own laptop rather than in a data center.
The bigger idea NVIDIA and Microsoft are selling is personal AI agents : software that doesn't just answer questions but actually does work for you across your applications, such as drafting, editing, generating images and video, writing code, and searching your files. The full NVIDIA CUDA software stack (the foundation most of the world's AI is built on) runs natively on the chip, and popular local-AI tools like llama.cpp and ComfyUI are already on board.
Content creation For creative professionals, the chip brings the full NVIDIA creator toolkit to a portable machine. NVIDIA cites capabilities like rendering 90GB+ 3D scenes , editing 12K video with professional 4:2:2 color, and generating 4K AI video. Adobe is going a step further and rearchitecting Photoshop and Premiere from the ground up for RTX Spark, promising up to 2x faster AI and editing performance, and apps like DaVinci Resolve, Blender, and OTOY's Octane are part of the ecosystem.
Gaming The RTX Spark is also a capable gaming chip, supporting the technologies PC gamers care about: ray tracing, the full DLSS suite (including the new DLSS 4.5), NVIDIA Reflex, and G-SYNC. NVIDIA says it can drive AAA games at 1440p and over 100 frames per second .
One detail matters more than it might seem. Because Windows on Arm runs many older games through an emulation layer, anti-cheat software has historically been a sticking point for online multiplayer. NVIDIA and Microsoft have specifically lined up native support from Easy Anti-Cheat and BattlEye , plus Xbox PC app support. That signals they're taking game compatibility seriously rather than treating it as an afterthought.
Why the RTX Spark is a big deal
A few things make this more than just another laptop chip:
NVIDIA is entering the Windows CPU market. For decades, the brains of a Windows PC came from Intel or AMD, and more recently Qualcomm. With the RTX Spark, NVIDIA (the company best known for graphics) is now supplying the central processor too. This is a meaningful shake-up of a market that has been remarkably stable, and it gives Microsoft a powerful new partner for its Windows-on-Arm ambitions.
Unified memory changes what a laptop can do. Putting up to 128GB of shared, high-bandwidth memory behind both the CPU and GPU is something normally seen in workstations and servers, not thin-and-light laptops. It is the key enabler for running large AI models on-device, a capability that simply hasn't been practical on mainstream portable hardware before.
The "personal AI" pitch is genuinely new. NVIDIA and Microsoft aren't just selling speed; they're selling a different relationship with your computer, one where AI agents work alongside you. To make that safe, they've built a security layer: new Windows "security primitives" for identity and containment, plus NVIDIA's OpenShell runtime, which lets you control what agents are allowed to do, keeps sensitive work on local models based on your privacy rules, and can mask personal information in any queries that do get sent to the cloud. Whether it lives up to the promise is something only real-world use will tell, but the ambition is real, and it's the actual reason this product exists.
When will the NVIDIA RTX Spark be released?
The NVIDIA RTX Spark is expected to launch in fall 2026 . Microsoft has confirmed its own Surface device in the first wave, alongside several other major PC manufacturers, with additional models (including Acer's) confirmed to follow. NVIDIA and its partners have not yet announced pricing.
Can you buy the RTX Spark chip by itself? No. This is an important point to clear up: the RTX Spark isn't a graphics card or an upgrade part you can drop into an existing PC. It's a system-on-chip , meaning the CPU, GPU, and memory are fused together and soldered into the machine at the factory. To get an RTX Spark, you buy a complete laptop or compact desktop built around it. There is no standalone version to install yourself.
What you can choose is the form factor and configuration. NVIDIA describes laptops as slim as 14mm and as light as three pounds , in sizes from 14 to 16 inches, with premium machined-aluminum bodies and color-accurate tandem OLED displays. There will also be small, ultra-efficient desktop versions built to run AI agents around the clock. Remember the "up to" in the specs: the 128GB memory and 1-petaflop figures represent the top of the range, so different models will offer different amounts of memory and performance at different prices.
Should you buy the NVIDIA RTX Spark? The honest answer right now is that it's promising, but it's worth waiting for the full picture. Here's how to think about it.
The RTX Spark looks like a strong fit if you are:
Someone who wants to run AI models and agents privately on your own machine, without sending everything to the cloud. A creator who works with video, 3D, or AI-generated content and wants serious power in a portable, long-battery-life form. A buyer who values a thin, premium laptop and is comfortable being an early adopter of a new platform. It may be worth holding off if you:
Rely on a wide range of older or niche Windows software, since some apps still run through emulation on Windows on Arm. NVIDIA and Microsoft have done substantial work here (native versions of major creative apps, expanded emulation, and proper anti-cheat support), but the long tail of programs and games is best confirmed by independent reviews once devices ship. Need maximum sustained performance for hours-long gaming or AI workloads. The chip is designed for efficiency in slim chassis (Microsoft's developer version, for reference, runs in a 100-watt thermal envelope), and how well it holds up under sustained heavy load is something to watch in real-world testing. The two biggest unknowns as of now are price and real-world battery life and sustained performance . NVIDIA and its partners haven't announced official pricing, though a Taiwan-based report estimates the first RTX Spark laptops will likely cost more than US$4,000, and premium components like tandem OLED suggest these won't be budget machines. Battery life and sustained performance, meanwhile, won't be clear until reviewers get their hands on shipping hardware.
Our recommendation: if the personal-AI and creator capabilities excite you, keep an eye on the fall 2026 launch, watch for independent reviews, and compare specific configurations once pricing is public.
Stay in the loop Acer's own RTX Spark systems are on the way. We're not ready to share the details just yet, but when we are, you'll want to be among the first to know. Keep an eye out for our RTX Spark announcement, and sign up for the Acer newsletter to be the first to hear when it lands
Frequently asked questions Is the NVIDIA RTX Spark an Arm chip?
Yes. The RTX Spark uses a 20-core NVIDIA Grace CPU based on the Arm architecture, and it runs Windows on Arm. Older x86 and x86-64 apps run through Microsoft's Prism emulation layer.
Can I buy the RTX Spark chip separately and install it myself?
No. The RTX Spark is a system-on-chip with the CPU, GPU, and memory fused together and soldered in at the factory. You buy a complete laptop or compact desktop, not a standalone part.
When does the NVIDIA RTX Spark come out?
RTX Spark laptops and compact desktops are expected in fall 2026, starting with Microsoft Surface and other major manufacturers, with more models (including Acer's) to follow.
How much does the RTX Spark cost?
NVIDIA and its partners have not announced official pricing. A Taiwan-based report estimates that the first RTX Spark laptops will likely cost more than US$4,000 , in line with NVIDIA positioning them for the high-end market. For reference, the DGX Spark, which uses the same GB10 chip, sells for roughly US$3,500 to US$4,700.
What is the difference between RTX Spark and DGX Spark?
DGX Spark is a Linux-based desktop appliance for AI developers, built on the GB10 Grace Blackwell Superchip and priced around $4,000. RTX Spark is a chip inside consumer and prosumer Windows laptops and desktops. They share related silicon but target different users.
Is the RTX Spark good for gaming?
Yes. It supports ray tracing, the full DLSS suite (including DLSS 4.5), NVIDIA Reflex, and G-SYNC, and NVIDIA says it can run AAA games at 1440p and over 100 frames per second. Native anti-cheat support from Easy Anti-Cheat and BattlEye helps with online multiplayer.
Can the RTX Spark run AI models locally?
Yes. With up to 128GB of unified memory and up to 1 petaflop of FP4 AI performance, NVIDIA says it can run a 120-billion-parameter large language model with up to a one-million-token context window directly on the device.
Specifications and capabilities described here are based on NVIDIA's and Microsoft's announcements as of mid-2026 and may change before the products ship. Figures stated as "up to" refer to maximum available configurations.
Recommended Products
