What Is Nvidia RTX Spark? The Personal AI PC Platform Explained

Nvidia RTX Spark is Nvidia's platform for small, Arm-based Windows PCs — slim laptops and compact desktops built around a "Superchip" that pairs a Blackwell RTX GPU with a Grace CPU on one package, the same Grace Blackwell design Nvidia uses in its data-center and DGX Spark systems, shrunk down to run on a desk or a lap. Nvidia has confirmed the chip, its core specs, and a wave of OEM partners; what it has not confirmed, as of early November 2026, is a retail price for the flagship N1X configuration, which is why so much of the pricing conversation around RTX Spark is still analyst guesswork rather than an Nvidia price list.

What Nvidia RTX Spark Actually Is

Nvidia introduced RTX Spark as a new category of Windows PC built around the idea of a "personal AI computer" — a machine designed to run large AI models, local coding agents, and creative-AI workflows directly on the device instead of leaning entirely on the cloud. Rather than a single product, RTX Spark is a chip platform: a Superchip that Nvidia's hardware partners drop into their own laptop and desktop designs, the same way Intel or AMD chips show up in machines from a dozen different brands.

At its core is Nvidia's combination of a Blackwell RTX GPU and a Grace CPU, linked together on a single module. That Grace Blackwell pairing is the same basic architecture Nvidia uses in its DGX Spark desktop AI appliance and in parts of its data-center lineup — Nvidia's Grace CPU architecture was originally designed for servers, and RTX Spark is Nvidia's attempt to bring a version of that CPU-GPU pairing into consumer and prosumer Windows hardware. The result is a chip Nvidia pitches as capable of running AI models with tens of billions of parameters locally, something that's historically required a data-center GPU or a cloud subscription.

RTX Spark was announced in mid-2026 and formally launched alongside Microsoft in October 2026, with the first wave of laptops going on sale October 16. Pandromeda has covered that launch event and the laptop lineup as individual news stories; this piece is different — it's a step back from the news cycle to answer what RTX Spark is, what's actually confirmed about it, and who should care.

Inside the Superchip: Grace Blackwell, Shrunk for a Desk or a Lap

Nvidia's RTX Spark product page describes the chip as combining fourth-generation RT cores and fifth-generation Tensor cores with a Grace CPU connected directly to the GPU rather than through a conventional PCIe link to a separate processor. That tight CPU-GPU coupling is the point: it lets the system treat CPU and GPU memory as one shared pool — Nvidia calls it unified memory — so a large AI model doesn't have to be split awkwardly across separate, smaller pools of CPU RAM and GPU VRAM the way it would on a typical gaming laptop.

On the graphics side, Nvidia lists support for DLSS 5, its latest Reflex 2 low-latency tools, hardware AV1/HEVC/H.264 encode and decode, PCIe Gen 5, and the usual DirectX 12 Ultimate and Vulkan ray-tracing stack. None of that is unusual for a modern Nvidia GPU; what's new is packaging it alongside a many-core Arm CPU built for sustained AI workloads rather than a conventional x86 chip from Intel or AMD. That's also why RTX Spark machines run Windows on Arm rather than the traditional x86 version of Windows most PCs use.

N1 vs N1X: What's Actually Confirmed

This is where coverage of RTX Spark gets messy, and it's worth being precise about it. Nvidia's own RTX Spark product page and spec sheet currently only brand and detail one chip: RTX Spark N1X. It lists two N1X configurations — a 6,144-core Blackwell GPU paired with a 20-core Grace CPU and up to 128GB of unified LPDDR5X memory, and a smaller 5,120-core GPU paired with an 18-core CPU and up to 64GB of memory — with TDPs that Nvidia lists in a range from roughly 45W up to 140W depending on the configuration.

Nvidia's own developer documentation, which lists the device IDs hardware and driver engineers use to identify each chip, is the clearest confirmation of how those two configurations map to form factors: the RTX Spark GPU identification guide shows the 6,144-core GPU shipping in both laptop and desktop designs, while the 5,120-core GPU is listed as laptop-only. In other words, the bigger chip is the one that shows up in Nvidia's compact desktops; the smaller chip is reserved for thinner laptops.

A cheaper "N1" tier — without the X — shows up constantly in press coverage and analyst notes as the chip expected to bring RTX Spark pricing down toward mainstream laptop territory. But as of this writing, Nvidia has not published official N1 specifications, a device ID, or a product page for it. Secondary reports disagree on basic details, with some describing a 12-core Arm CPU and a separate GeForce-branded GPU, and others describing an 18-core/5,120-core configuration similar to the smaller N1X tier. Until Nvidia or an OEM ships an actual N1 product, treat every N1 spec you read — including in this article — as unconfirmed.

Desktop vs Laptop: Same Silicon, Different Jobs

Nvidia frames the two form factors around different use cases rather than different performance tiers. The compact desktop, according to Nvidia's marketing copy, is pitched as a machine that can "run agents around the clock" — something you leave powered on to handle background AI tasks, similar in spirit to Nvidia's own DGX Spark appliance, but built around Windows rather than Nvidia's Linux-based DGX OS. The laptop configurations are pitched on battery life and portability, with Nvidia calling RTX Spark laptop silicon some of the most power-efficient RTX hardware it has made, aimed at creators and developers who want serious local AI and GPU power without being tethered to a wall outlet.

For a closer look at which specific laptops are actually shipping with RTX Spark inside, Pandromeda has tracked every confirmed RTX Spark laptop model and release date separately, including entries from Asus, Dell, HP, Lenovo, MSI, and Microsoft's own Surface Laptop Ultra.

What the Specs Actually Mean

Nvidia's headline performance claim for RTX Spark is up to 1 petaflop of FP4 AI compute and the ability to run AI models with up to 120 billion parameters and a 1-million-token context window locally, without a cloud connection. Nvidia also notes that with up to 128GB of unified memory, roughly 90GB can be addressed by the GPU for large 3D scenes or models — a nod to creative professionals doing 3D rendering or video work, not just AI developers.

It's worth being clear about what these figures are and aren't. They are Nvidia's own published specifications and performance framing for the chip, not independent, third-party benchmark results. Nvidia has not published side-by-side benchmark numbers against competing Arm AI PCs (such as Qualcomm's Snapdragon X2 Elite machines or AMD's Ryzen AI Max lineup) on its RTX Spark product page, and Pandromeda has not independently benchmarked RTX Spark hardware. Treat the petaflop and parameter-count figures as Nvidia's stated specs, useful for understanding what the platform is built to do, rather than as guaranteed real-world performance in every app.

Pricing: What's Official and What's Still Guesswork

This is the single most confused part of the RTX Spark conversation, so it's worth separating cleanly.

What's actually confirmed, on an official Nvidia or OEM page: Nvidia itself has not posted a general retail price for RTX Spark chips or a baseline system price on its product page — shopping links route to Nvidia's marketplace rather than a fixed price list. Two specific Microsoft products that use RTX Spark do have confirmed, official starting prices: the Surface Laptop Ultra, which Microsoft lists starting at $2,599.99 with its N1X configuration, and the Surface RTX Spark Dev Box, a mini-PC Microsoft prices starting at $5,999.99. Pandromeda has a dedicated breakdown of that Dev Box's specs, price, and release date if you want the full configuration options.

What's analyst estimate, not Nvidia or OEM pricing: Morgan Stanley analysts, in a note cited widely by tech press, estimated that systems built on the larger N1X variant are unlikely to be priced below roughly $2,899, while systems built on a cheaper N1 variant would likely start around $1,799 or higher. Other outlets have floated slightly different floors — some putting N1X nearer $2,500 and N1 nearer $2,000. None of these figures come from Nvidia, Microsoft, Asus, Dell, HP, Lenovo, or MSI directly; they're analyst modeling based on component costs and early channel checks, and they disagree with each other on specifics, which is itself a sign of how little has actually been confirmed. Nvidia has not disputed or confirmed any of these numbers publicly.

The practical takeaway: if you see a specific dollar figure attached to "RTX Spark price," check whether it's attached to an actual OEM SKU (like Microsoft's two confirmed products above) or to the word "estimate," "analyst," or "Morgan Stanley." Those are very different kinds of claims.

The Hardware: Who's Actually Building RTX Spark Machines

Nvidia's RTX Spark product page lists launch partners across both laptops and desktops, including Asus's ProArt P16, Dell's XPS 16, HP's OmniBook Ultra 16, Lenovo's Pro 9 lineup, MSI's Prestige N16 Flip AI+, and Microsoft's Surface Laptop Ultra. Microsoft is also the first partner to confirm a standalone RTX Spark desktop product for general sale, the Surface RTX Spark Dev Box, which Microsoft pitches primarily at developers building and testing local AI agents rather than at mainstream buyers.

That spread of partners mirrors how Nvidia has typically operated with GPU silicon: it builds the chip and reference designs, and lets PC makers decide on chassis, pricing, and target audience. Expect meaningful differences in battery life, port selection, and build quality between an Asus creator laptop and a Microsoft productivity-focused Surface, even when both are running the same N1X chip underneath.

RTX Spark vs DGX Spark: Don't Confuse the Two

Nvidia also sells a separate product called DGX Spark, and the naming similarity causes a lot of confusion. The short version: DGX Spark is Nvidia's own Linux-based "AI supercomputer on your desk," sold directly by Nvidia and aimed at AI researchers and engineers who want a dedicated local development box running Nvidia's DGX software stack. RTX Spark is the Windows-based consumer and creator platform covered in this article, sold through PC makers rather than Nvidia itself, and built to run ordinary Windows apps alongside AI workloads rather than a specialized AI research OS. They share a very similar underlying Grace Blackwell chip design, which is exactly why the names collide so often in headlines. Pandromeda has broken down the full differences between RTX Spark and DGX Spark if you're trying to decide which one, if either, fits what you're trying to do.

Who RTX Spark Is Actually For

Strip away the marketing, and RTX Spark is aimed at a fairly specific slice of buyers rather than the mainstream laptop shopper. The clearest fits are developers who want to run and iterate on local AI models or agents without renting cloud GPU time, creative professionals doing 3D, video, or AI-assisted content work who need more unified memory than a typical gaming laptop offers, and IT or research teams who want always-on, local AI appliances that don't involve sending data to a third-party cloud.

For most everyday buyers who just want a fast Windows laptop for browsing, office work, or light gaming, RTX Spark is almost certainly overkill, especially while confirmed and estimated pricing both sit well above $2,500. A conventional Windows laptop with a modern x86 or Arm chip, or Nvidia's regular GeForce RTX laptop GPUs, will do everyday tasks and most gaming for considerably less money, and won't carry any of the app-compatibility quirks that still occasionally show up on Windows-on-Arm hardware. RTX Spark's real pitch is to people whose work specifically benefits from large, local, unified-memory AI compute — not to anyone shopping for "the newest Nvidia laptop chip" in the abstract.

What's Next for RTX Spark

Three things are worth watching. First, whether Nvidia ever formally announces the cheaper N1 tier that analysts keep pricing in advance of any official spec sheet — until that happens, every N1 number in circulation, including the ones in this article, is secondhand. Second, whether more OEMs beyond Microsoft publish firm retail prices for their N1X laptops and desktops in the weeks after the October 16 launch, which would let buyers compare real sticker prices instead of analyst floors. Third, how RTX Spark performs once independent reviewers get retail units in hand, since Nvidia's own petaflop and parameter-count figures are the only performance numbers publicly available right now. Pandromeda will keep tracking confirmed pricing and new models as OEMs formally announce them.

Written byMara Lindqvist

Mara Lindqvist edits the hardware desk. She covers graphics cards, processors, memory and storage, the foundries and chip designers behind them, and what the numbers on a spec sheet mean for people choosing a PC. Specifications in her stories come from manufacturer spec pages and datasheets.

More from Mara LindqvistAll Hardware coverage