NVIDIA’s RTX Spark platform has moved from announcement to store shelves. Pre-orders opened October 7, 2026, across six PC makers, and NVIDIA says the first machines became available on October 16. The launch marks a shift for NVIDIA: rather than selling a graphics chip to laptop makers and letting them design around it, the company built an entire Windows platform, spec sheet and all, then handed it to six OEMs at once.
The platform pairs an Arm-based Grace CPU with a Blackwell RTX GPU and unified memory, the same architectural idea behind NVIDIA’s DGX-class AI systems, now squeezed into a laptop chassis. NVIDIA says the top configuration delivers up to 1 petaflop of FP4 AI compute, roughly 1,000 TOPS, which is a figure normally reserved for data center hardware, not a 2.2-pound ultrabook. Six manufacturers, Microsoft, ASUS, HP, Dell, MSI and Lenovo, are now taking orders, with starting prices that range from $2,599.99 to $4,199. Two more, Acer and GIGABYTE, join the lineup October 23, and compact RTX Spark desktops are expected in November.
Don't miss new tech stories on Google
Add Tech Insider once in the Google app and our stories appear in your news suggestions.
What RTX Spark Actually Is
RTX Spark is NVIDIA’s branding for the Windows version of its Grace Blackwell superchip design, known internally as GB10 in desktop-class form and N1X in laptop form. Wccftech reports the full GB10 configuration uses a 20-core Arm CPU, split between 10 Cortex-X925 performance cores and 10 Cortex-A725 efficiency cores, paired with a Blackwell RTX GPU that scales up to 6,144 CUDA cores. NVIDIA has described the CPU and GPU as connected at 600 GB/s, letting both chips draw from a single pool of unified memory rather than splitting system RAM from dedicated VRAM the way conventional laptops do.
That unified memory pool is the headline spec. Laptop configurations start at 24GB and scale to 128GB, all addressable by both CPU and GPU workloads. For context on how NVIDIA has used this same unified-memory approach at larger scale, see our coverage of the NVIDIA DGX Station, which pushes the same Grace Blackwell concept to 748GB for workstation-class AI training.
Laptop-class RTX Spark chips run at an estimated 45 to 80 watts depending on configuration, while desktop versions are rated around 140 watts, according to figures compiled by Wccftech. Graphics throughput tops out near 31 FP32 TFLOPS on the highest-end configuration, a number NVIDIA pairs with gaming claims of roughly 100 frames per second at 1440p in select titles, though no independent lab benchmark of that figure has been published yet.
Six OEMs, One Pre-Order Window
What makes this week different from prior RTX Spark coverage is the breadth of the launch. NVIDIA previously teased RTX Spark as an AI PC launch and OEMs individually leaked pricing for months, including an early HP RTX Spark price leak and a Surface Laptop Ultra specification leak. This week is the first time pricing, configurations, and a shipping date have become official and simultaneous across six manufacturers rather than trickling out one leak at a time.
| OEM & Model | Starting Price | Key Configuration | Pre-Order / Ship Date |
|---|---|---|---|
| Microsoft Surface Laptop Ultra | $2,599.99 | 18-core CPU, 5,120-core GPU, 24GB RAM, 512GB storage | Pre-orders opened Oct. 7; available since Oct. 16 |
| ASUS ProArt P14 | $2,599.99 | 18-core CPU, 24GB RAM, 512GB storage, 3K OLED | Retail release Oct. 16 |
| ASUS ProArt P16 | $2,799.99 | 18-core CPU, 24GB RAM, 512GB storage, 4K OLED | Retail release Oct. 16 |
| HP OmniBook Ultra 16 | $2,999 | 32GB RAM base configuration | Available since Oct. 16 |
| Dell XPS 16 Creator Edition | $3,799.99 | 18-core CPU, 32GB RAM, 512GB storage, 3.2K OLED | Retail release Oct. 16 |
| MSI Prestige N16 Flip AI+ | $3,299 (32GB) / $4,199 (64GB) | Convertible 2-in-1 laptop | Available since Oct. 16 |
| Lenovo Yoga Pro 9n | Not yet confirmed | RTX Spark laptop, part of launch wave | Oct. 16 launch wave |
The spread between the cheapest and most expensive configuration, $2,599.99 for the Surface Laptop Ultra entry model versus $4,199 for the top-spec MSI Prestige N16 Flip AI+, comes down almost entirely to memory and display. Every machine on this list shares the same CPU core-count tier and similar GPU core counts; the jump to 64GB of unified memory on the MSI model accounts for roughly $900 of the price gap on its own, which tells you how much NVIDIA and its OEM partners are charging per gigabyte of unified RAM at this launch.
Our earlier look at the Dell XPS 16 RTX Spark Creator Edition flagged the $3,799.99 price point weeks before this wider pre-order wave confirmed it, and the number has held steady into launch.
Laptop Spec Tiers, Side by Side
NVIDIA and its OEM partners are shipping two distinct RTX Spark tiers rather than one fixed chip. The lower laptop tier uses an 18-core processor and a 5,120-core Blackwell GPU, while the higher tier scales to 20 cores and 6,144 GPU cores, matching the desktop-class GB10 spec. Both tiers share the same unified-memory architecture, just at different ceilings.
| Spec | Laptop Tier (N1X, base) | Full Tier (GB10, desktop-class) |
|---|---|---|
| CPU cores | 18-core Arm | 20-core Arm (10 Cortex-X925 + 10 Cortex-A725) |
| GPU cores | 5,120 Blackwell CUDA cores | Up to 6,144 Blackwell CUDA cores |
| Unified memory | 24GB and up | Up to 128GB LPDDR5x |
| AI performance | Lower-tier FP4 throughput, not independently benchmarked | Up to 1 petaflop FP4 (~1,000 TOPS) |
| Graphics throughput | Not separately disclosed | Up to 31 FP32 TFLOPS |
| Power envelope | Approx. 45-80W | Approx. 140W (desktop) |
| Display outputs | Varies by OEM | Up to 4 displays (3x DisplayPort, 1x HDMI 2.1a) |
This two-tier structure explains why the Surface Laptop Ultra and the entry ASUS ProArt P14 land at the same $2,599.99 price despite coming from different manufacturers: both ship the baseline 18-core, 5,120-core, 24GB configuration. Buyers who want the full 128GB unified-memory ceiling have to step up to the higher-priced models or wait for the desktop variants NVIDIA has scheduled for November.
Desktops Arrive in November, Acer and Gigabyte Systems Planned
The October 16 launch covers laptops only. NVIDIA says compact RTX Spark desktops, the form factor closer to the original DGX Spark workstation concept, are planned for November 2026, after the laptops. That timing gives NVIDIA and its partners a second launch moment heading into the holiday shopping window, and it gives buyers who want the full 128GB desktop configuration without laptop thermal constraints a reason to wait.
NVIDIA also named Acer and Gigabyte systems as planned additions to the lineup, though launch coverage identified only the six original OEMs as having laptops on pre-order, and confirmed pre-order status for the two newcomers is unavailable. Pricing and model names for either company’s RTX Spark hardware have not been confirmed, but their inclusion signals NVIDIA wants broader retail shelf coverage fast rather than letting the category stay a six-brand club. For a sense of how previous NVIDIA AI PC desktop concepts have scaled memory even further, see our coverage of the Microsoft Surface RTX Spark Dev Box, a developer-focused desktop that debuted with 128GB RAM ahead of this consumer wave.
Why NVIDIA Is Selling a Platform, Not Just a Chip
The RTX Spark strategy breaks from how NVIDIA has historically sold laptop silicon. In past generations, NVIDIA shipped a GPU die to Dell, HP, ASUS, and others, and each OEM built a distinct machine around it with its own CPU pairing, memory configuration, and thermal design. RTX Spark instead standardizes the CPU, GPU, and unified-memory architecture at the platform level, leaving OEMs to differentiate on chassis, display, and price tier rather than core silicon. That is a notable shift toward the kind of reference-platform control Apple exercises over its own silicon, and it is also a hedge against the generic commodity-laptop pricing pressure that has squeezed margins across the Windows PC market for years.
It also reflects how aggressively NVIDIA is pushing local AI compute as a laptop selling point rather than a cloud subscription. Our earlier report on RTX Spark builder PCs covered NVIDIA and Microsoft framing the 1-petaflop figure as a developer and creator tool, and the enterprise side of that pitch, including the security tradeoffs of running agentic AI workloads locally rather than in a managed cloud environment, was the focus of our piece on RTX Spark agentic PCs and enterprise security gaps.
RTX Spark vs. the Rest of the AI PC Market
RTX Spark’s closest conceptual rival is Apple’s unified-memory Mac hardware, which has spent the past several years building a reputation among developers for running large local AI models thanks to shared CPU-GPU memory pools that avoid the VRAM ceilings of conventional discrete GPUs. NVIDIA’s pitch is that RTX Spark brings that same unified-memory advantage to Windows, while adding CUDA, TensorRT, and the broader NVIDIA software stack that most AI development tooling already targets. No independently verified head-to-head benchmark between RTX Spark and any Mac configuration has been published, so performance claims in either direction should be read as platform specifications rather than lab results.
Against conventional Windows AI laptops, the comparison is more clear-cut. Most “AI PC” laptops sold in 2026 rely on an integrated NPU rated in the tens of TOPS, paired with 16GB to 64GB of conventional system RAM that is not shared with the GPU. RTX Spark’s up-to-128GB unified pool and roughly 1,000-TOPS FP4 ceiling put it in a different compute class entirely, but that capability comes at a starting price of $2,599.99, well above the $1,000 to $1,800 range where most NPU-based thin-and-light laptops sell. NVIDIA is not pricing RTX Spark to compete with mainstream Copilot+ PCs; it is pricing it against developer workstations and creator laptops.
That positioning also puts RTX Spark in indirect competition with NVIDIA’s own higher-end hardware. The DGX Station’s 748GB memory pool and 20-petaflop ceiling dwarfs anything in the laptop lineup, but it also costs dramatically more and targets a different buyer, enterprise AI teams rather than individual developers and creators. RTX Spark sits deliberately between consumer laptops and DGX-class workstations, a gap NVIDIA has not had a dedicated product for until now.
Historical Context: From DGX Spark to a Six-OEM Retail Launch
RTX Spark did not appear out of nowhere. NVIDIA spent much of 2026 previewing pieces of this architecture under different names, including early DGX Spark branding and the N1X codename that still shows up in technical reporting. What changed between those earlier previews and this week is distribution: a single concept that was once a developer-kit curiosity is now a six-brand consumer retail SKU with firm pricing and a ship date. That is a faster jump from announcement to shelf than NVIDIA’s own DGX hardware made, and it mirrors the speed at which Apple Silicon moved from a single MacBook Air launch to covering an entire Mac lineup within about two years, except NVIDIA is attempting it across six external OEM partners simultaneously rather than inside its own hardware division.
Software support has been tracking the hardware closely. Utilities like CPU-Z and HWMonitor added RTX Spark recognition ahead of this launch, a sign that the enthusiast tooling ecosystem expected this pre-order wave and wanted day-one compatibility ready. Our earlier coverage of CPU-Z and HWMonitor adding RTX Spark support noted that this kind of early tooling support typically precedes a firm retail date by only a few weeks, which lines up with what shipped this week.
Market Impact: What a Six-OEM Launch Signals
A simultaneous six-OEM launch is a different market signal than a single-vendor exclusive. It tells component suppliers, app developers, and competing chipmakers that NVIDIA expects RTX Spark volume broad enough to justify parallel supply chains across Microsoft, ASUS, HP, Dell, MSI, and Lenovo at once, rather than testing demand with one partner first. For Intel and AMD, whose x86 platforms still dominate the laptop CPU market, RTX Spark’s Arm-based Grace CPU is a second front, after server and data center Arm competition, that has already drawn stock-price attention. For software vendors building local AI tools, 128GB of unified memory addressable by both CPU and GPU workloads removes a ceiling that has forced many local-model projects to either quantize aggressively or use the cloud.
The pricing structure also tells a story about where NVIDIA sees margin. Every OEM on the launch list is pricing above $2,500, with most clustering in the $2,600 to $3,800 range and only the highest-memory configuration crossing $4,000. That is a narrower price band than NVIDIA’s standard RTX laptop GPU lineup, which spans from sub-$1,000 budget gaming laptops to $4,000-plus workstation-class machines. By keeping RTX Spark’s entire price range premium, NVIDIA is signaling this platform is not meant to cannibalize its existing GeForce RTX laptop business; it is a new, adjacent category aimed at buyers who would otherwise consider a Mac Studio, a cloud AI subscription, or a dedicated workstation.
What to Expect Next
- Expect compact RTX Spark desktops to undercut the laptop entry price once they arrive in November, since desktop cooling and chassis costs are typically lower than laptop equivalents at the same silicon tier.
- NVIDIA has said Acer and GIGABYTE systems are also planned, but their preorder status and pricing are not confirmed yet, so it is too early to say how they will compare with the Surface Laptop Ultra and ASUS ProArt P14.
- Expect the 128GB unified-memory configuration to become the benchmark figure every local-AI tooling vendor references when describing minimum requirements for running larger open-weight models on a laptop.
- Expect continued pressure on Intel and AMD’s laptop CPU roadmaps as NVIDIA’s Arm-based Grace design proves out in a six-OEM retail launch rather than a single reference design.
- Expect scrutiny of NVIDIA’s 1-petaflop and 100-fps gaming claims to intensify once independent reviewers get RTX Spark laptops in hand after the October 16 ship date, since no outlet has published an independently verified benchmark yet.
Buying Advice for This Launch Wave
For buyers deciding which configuration to buy now that RTX Spark laptops are on sale, the memory tier matters more than the brand. NVIDIA’s platform starting price is $2,599, and entry configurations with 24GB to 32GB of unified memory are enough for lighter local-AI workloads and everyday creator tasks, but they will not comfortably run the largest open-weight models without aggressive quantization. The jump to 64GB, reported at $4,199, is where RTX Spark’s unified-memory advantage becomes most apparent for serious local AI development. Buyers who specifically want the full 128GB ceiling will pay considerably more, and a 128GB listing was reported out of stock, so they may prefer to wait for more availability or for the compact desktops NVIDIA has planned for November.
Display and chassis preference should decide between otherwise similar configurations. The ASUS ProArt P14 and P16 differentiate mainly on screen size and resolution (3K versus 4K OLED), the Dell XPS 16 Creator Edition pairs its $3,799.99 price with a 3.2K OLED panel, and the MSI Prestige N16 Flip AI+ is the only confirmed convertible 2-in-1 in the lineup, a form factor that may matter more to creators than to developers running local inference workloads.
Frequently Asked Questions
When did NVIDIA RTX Spark pre-orders open?
Pre-orders opened October 7, 2026, across six OEMs: Microsoft, ASUS, HP, Dell, MSI, and Lenovo, according to reporting from Wccftech and Notebookcheck.
When do RTX Spark laptops ship?
The confirmed shipping and retail availability date for the initial six-OEM laptop wave is October 16, 2026.
How much do RTX Spark laptops cost?
NVIDIA lists RTX Spark systems from US$2,599, with regional and configuration differences. Starting prices on the launch models range from $2,599.99 for the Microsoft Surface Laptop Ultra and ASUS ProArt P14 up to $4,199 for the 64GB MSI Prestige N16 Flip AI+ configuration, while 128GB configurations were reported at US$6,799 and above.
What chip powers RTX Spark?
RTX Spark laptops use NVIDIA’s Grace Blackwell design, an Arm-based CPU (18 or 20 cores depending on tier) paired with a Blackwell RTX GPU (5,120 or 6,144 CUDA cores) and unified memory shared between both.
How much memory do RTX Spark laptops have?
Configurations range from 24GB up to 128GB of unified LPDDR5x memory, addressable by both the CPU and GPU.
When are RTX Spark desktops coming?
Compact RTX Spark desktop systems are expected in November 2026, roughly a month after the initial laptop wave ships.
Which other OEMs are joining RTX Spark?
NVIDIA said Acer and Gigabyte systems were also planned, though their pre-order status and specific pricing have not been confirmed.
How does RTX Spark compare to a Mac for local AI work?
Both platforms use unified memory shared between CPU and GPU, which helps run larger AI models locally than conventional discrete-GPU laptops. NVIDIA’s advantage is CUDA and TensorRT software compatibility with the broader AI development ecosystem; no independently verified benchmark comparing RTX Spark directly to any Mac configuration has been published yet.
Sources: Wccftech, NVIDIA, Microsoft Surface, ASUS ProArt, and the NVIDIA blog.
