NVIDIA has made graphics cards for decades, but it has never built a consumer CPU. That changed on June 1, 2026, when CEO Jensen Huang stepped onto the stage at the Taipei Music Center during Computex 2026 and officially unveiled the RTX Spark, a new system-on-chip that marks the most significant thing Nvidia has ever done for mainstream computing. This isn’t a GPU update or a data center chip you’ll never touch. It’s a processor designed to sit inside the laptop you carry every day, and on paper, it’s unlike anything the PC market has seen.
The RTX Spark is the commercial name for what had been quietly developing under the N1 and N1X codenames for the past two years. It combines a custom 20-core ARM CPU, co-designed by Nvidia and MediaTek, with a Blackwell RTX GPU that packs 6,144 CUDA cores.
That GPU core count is identical to Nvidia’s desktop RTX 5070. All of it lives on a single chip built on TSMC’s 3 nm process, carrying 70 billion transistors and supporting up to 128 GB of unified LPDDR5X memory shared between the CPU and GPU at roughly 301 GB/s bandwidth. Huang called the moment “as revolutionary as the change from phones to smartphones,” and based on the numbers, that framing is hard to argue with.
What Makes the RTX Spark a Real Power Breakthrough
The efficiency angle is where RTX Spark gets genuinely interesting. The flagship N1X configuration runs a 45W to 80W TDP. A traditional gaming laptop packing an RTX 4070 requires around 120W just for the GPU side alone. According to pre-Computex data shared by analysts, the N1X is expected to deliver equivalent gaming performance at around 65W of total system power. That’s not a minor improvement. That’s a fundamental redesign of how much energy a laptop has to burn to give you serious GPU performance.
I’ve been following the N1X story since it was first rumored for Computex 2025, watched it slip past CES 2026, and kept track of every leaked benchmark along the way. When the Geekbench OpenCL results surfaced earlier this year, they were running on early silicon at 1.05 GHz and still managing to outpace Apple’s M3 Max in integrated GPU scores. Those are prototype numbers, not final silicon. What Nvidia revealed today is the finished product.
For workloads beyond gaming, the story gets even more compelling. NVIDIA says the RTX Spark delivers one petaflop of AI performance at NVFP4 precision, alongside 31 TFLOPs of FP32 compute. It supports tasks like editing 12K video, rendering massive 3D scenes, and running advanced AI agents locally without any cloud dependency. Those are the same capabilities Nvidia’s $3,999 DGX Spark desktop supercomputer is built around, now squeezed into a laptop chip with a fraction of the power draw.
The RTX Spark family also includes a lower-power N1 tier for more mainstream laptop designs. Those configurations feature either 12 or 10 CPU cores, with 2,560 or 2,048 CUDA cores, respectively, and a TDP range of just 18W to 45W. Memory support tops out at 64GB. There’s also an 18-core N1X variant with 5,120 CUDA cores that sits between the base N1 and the top-end 20-core configuration, giving OEM partners more flexibility across price points.
Why CUDA Changes the Windows on Arm Story
Until today, Windows on Arm was primarily a Qualcomm story. Snapdragon X laptops have made real progress with battery life and efficiency, but gaming performance and GPU compute have remained weak points. The RTX Spark changes that equation completely. Qualcomm runs on its own QNN and DirectML stacks. NVIDIA brings CUDA, the software ecosystem that four million developers and over 3,000 applications already run on. That’s not a small detail.
After looking into this more closely, what becomes clear is that the biggest winners here might not be gamers at all. They might be developers and researchers who currently depend on cloud GPUs for serious AI inference. The same silicon inside the DGX Spark can already run quantized versions of DeepSeek, Meta’s LLaMA, and Google’s Gemma at the 200 billion parameter scale. An RTX Spark laptop brings a meaningful slice of that capability to a portable machine for the first time, and it does it on Windows with the full CUDA software stack intact.
The coordinated rollout was notable, too. On May 29, Nvidia, Microsoft’s Windows account, and Arm simultaneously posted “A new era of PC” alongside GPS coordinates pointing to the Taipei Music Center. That kind of synchronized message signals deep platform-level integration, not just a hardware partnership.
ARM-based PCs are projected to hit 30% global market share by 2026, and Nvidia just entered that race with arguably the strongest opening chip anyone has brought to the table. What most coverage missed is that DLSS 4.5 Ray Reconstruction also landed at this keynote. It uses a more refined transformer model for ray tracing denoising, delivering cleaner visuals at a similar performance cost to previous versions, and it’s coming to RTX Spark devices.
Who Is Building RTX Spark Laptops
Dell, Lenovo, Asus, and MSI all have RTX Spark devices in the pipeline. Lenovo has internally confirmed the Legion 7 15N1X11, a gaming laptop that requires a 245W power adapter, which signals a high-performance configuration pushing the chip hard. Lenovo’s lineup also includes the Yoga Pro 7, a Yoga 9 2-in-1, and several IdeaPad Slim 5 variants. Dell was showing an embargoed XPS model as early as May 31. Asus is building a ProArt-focused laptop targeting creative professionals. First devices are expected to ship before the 2026 holiday season, with broader availability rolling into early 2027.
Pricing is still unconfirmed officially, but analysts at Dataconomy project RTX Spark laptop pricing starting in the $1,000 to $1,500 range. Sources suggest premium configurations with more memory could push considerably higher, given LPDDR5X costs and TSMC 3nm manufacturing. Industry insiders hint that Nvidia may introduce trimmed configurations with less memory to give manufacturers a path toward more accessible price points, though no timeline for budget-tier RTX Spark products has been confirmed.
What Intel, AMD, and Qualcomm Need to Watch
NVIDIA is entering the consumer PC CPU market that ships around 150 million units per year globally. It’s doing that with a chip that already outpaces Apple’s M3 Max in early GPU benchmarks on integrated graphics, at less than half the total system power of what a traditional gaming laptop GPU requires. That’s the number most people are glossing over. Competing against Apple on creative and AI workloads has been the stated ambition for Intel and Qualcomm for years. NVIDIA just showed up to that fight carrying an RTX 5070’s worth of GPU compute in a laptop SoC.
Personally, I think RTX Spark has the potential to genuinely reshape the next two years of laptop buying. Whether it delivers depends on software support, real-world battery life testing, and how fast the Windows on Arm ecosystem matures around it. But the hardware itself is the most aggressive opening Nvidia could have made. If fall 2026 delivers what was shown in Taipei today, the RTX Spark could end up being the PC chip people look back on as the turning point for portable AI computing.
