Nvidia ARM Laptop Chip N1X Confirmed for Computex: CUDA and RTX 5070 GPU Onboard

On Friday morning, Nvidia, Microsoft, and Arm each posted the identical message to X — "A new era of PC" — followed by the GPS coordinates of the Taipei Music Center, where Nvidia CEO Jensen Huang is scheduled to deliver his GTC Taipei keynote on June 1. One day later, VideoCardz reported that Dell has an embargoed XPS laptop launch powered by the N1X set for May 31. Computex 2026 opens June 2. The choreography leaves no room for ambiguity: after three years of leaks, delays, and near-misses, Nvidia is about to announce its first laptop CPU — a chip that would put RTX 5070-class graphics and the full CUDA software stack inside a Windows ARM machine for the first time.

For anyone in the market for a laptop in the next 12 months, or for the AI and machine learning researchers who currently depend on cloud GPUs for serious inference work, that announcement is worth watching.

"A New Era of PC": How Nvidia Signaled N1X at Computex

The coordinated Friday tease was unmistakably deliberate. All three posts — from Nvidia, Microsoft's Windows account, and Arm's handle — appeared simultaneously at 9:00 a.m. Pacific time, each carrying identical language and the same geographic coordinates pointing to the Taipei Music Center. Microsoft's Pavan Davuluri, who leads Windows and Devices, separately clarified that the tease has nothing to do with a new version of Windows — narrowing the possible subjects squarely to hardware.

Asus joined the same day, posting a winking emoji alongside a teaser image aimed at its ProArt laptop lineup. Dell has an embargoed XPS launch with the N1X set for May 31, with Jensen Huang's keynote to follow at 11:00 p.m. ET on Sunday night. Lenovo has confirmed multiple N1X models internally, including the Legion 7 and several Yoga and IdeaPad Slim designs. MSI is also preparing N1X-based systems, giving the platform at least four confirmed or strongly indicated launch OEMs before a single official word from Nvidia.

What N1X Is and How It Works

The N1X is Nvidia's first laptop system-on-chip designed specifically for Windows ARM PCs. Per Nvidia's own Hot Chips 2025 presentation of the GB10 Superchip — the desktop variant of the same silicon — the chip pairs two distinct chiplets in a 2.5D package built on TSMC's 3nm process: a CPU die designed by MediaTek and a GPU die built on Nvidia's Blackwell architecture, connected at 300 GB/s bidirectional bandwidth via Nvidia's NVLink C2C interconnect.

On the CPU side, 20 Arm v9.2 cores split into two clusters of 10 performance and 10 efficiency cores share 32 MB of L3 cache. The GPU carries 6,144 CUDA cores across 48 Streaming Multiprocessors — the same shader count as the desktop GeForce RTX 5070 — alongside fifth-generation Tensor Cores with NVFP4 precision and dedicated ray tracing cores. Peak AI throughput reaches 1,000 TOPS at NVFP4 precision. FP32 floating-point performance stands at 31 TFLOPs. Memory runs as a unified LPDDR5X-9400 fabric on a 256-bit bus delivering roughly 301 GB/s of bandwidth shared between CPU and GPU.

The desktop DGX Spark ships with 128 GB of that unified memory at $3,999 and can run AI models up to 200 billion parameters locally. Laptop versions of the N1X are expected to arrive at smaller memory configurations; analysts at Dataconomy project N1X laptop pricing in the $1,000–$1,500 range, though no Nvidia-confirmed pricing has been disclosed.

Why CUDA Changes What Windows ARM Laptops Can Do

CUDA is the key differentiator that neither Qualcomm's Snapdragon X series nor Apple Silicon can offer on Windows. Qualcomm's platform runs on its proprietary QNN and DirectML stacks; Apple Silicon does not run Windows. The N1X changes the equation by bringing Nvidia's entire developer ecosystem — TensorRT, PyTorch's CUDA backend, llama.cpp for CUDA, TensorRT-LLM, and every other CUDA-native AI workflow — to a portable Windows machine.

For machine learning researchers and AI developers, the practical implication is the ability to prototype, fine-tune, and run inference on large models locally without a cloud subscription or a dedicated workstation. The same silicon in the DGX Spark desktop already supports quantized versions of DeepSeek, Meta Llama, and Google Gemma variants at the 200-billion-parameter scale. A laptop-class version of that silicon would make such workloads genuinely portable for the first time.

How Does Nvidia N1X Compare to Qualcomm Snapdragon and Apple Silicon?

Qualcomm's Snapdragon X platform holds the current Windows ARM franchise, winning design slots at Lenovo, Dell, HP, Samsung, and Microsoft Surface. Its advantages are a mature Windows software partnership and a broad OEM portfolio across price points. Its persistent weakness is GPU performance — a gap the N1X is explicitly designed to exploit.

Pre-release Geekbench prototype results show the N1X scoring approximately 3,096 in single-core and 18,837 in multi-core tests. The Qualcomm Snapdragon X Elite scores approximately 2,693 single-core — roughly 15% behind. The N1X trails AMD's Ryzen AI MAX+ 395 (21,035 multi-core) and Intel's Core Ultra 9 285HX (22,104 multi-core) by roughly 10–15% in multi-threaded CPU work, though those chips live in dedicated x86 laptop form factors rather than integrated SoC designs. These are prototype scores from a June 2025 pre-production sample; results with finalized Windows drivers are likely to differ.

Tom's Hardware, which reviewed the GB10 desktop, notes that because the CPU and GPU share the same LPDDR5X memory pool, the GPU operates at roughly 273 GB/s — meaningfully less than discrete-GPU laptops running GDDR memory. Gaming on the GB10 was described as possible but "not the platform's strongest suit." Software compatibility is also a live variable: earlier this year, sources close to Nvidia's OEM partners described Windows bug-fixing for the N1X as "a nightmare," though the imminent Computex reveal signals that work has advanced significantly. Not all legacy x86 apps run without friction under Windows ARM emulation, and game compatibility and driver support remain works in progress.

Apple Silicon — the M4 and M4 Max — remains the benchmark for ARM laptop efficiency and unified-memory AI performance, with a far more mature software ecosystem than Windows ARM. The N1X's primary advantages over Apple are platform openness, Windows and x86 app compatibility through emulation, CUDA portability, and the Nvidia GPU software stack. The two platforms are not in direct competition: the N1X challenges Qualcomm on Windows, not Apple on macOS.

Lenovo, Dell, Asus, MSI: Which N1X Laptops Are Coming

Based on confirmed OEM teases, supply chain reports, and Computex media materials, the initial N1X device lineup shapes up as follows. Dell has an embargoed XPS laptop with N1X set to be shown on May 31. Lenovo has internally confirmed the Legion 7 15N1X11 gaming laptop — which will require a 245W power adapter, signaling a high-performance configuration — alongside IdeaPad Slim 5 variants, a Yoga Pro 7, and a Yoga 9 2-in-1. Asus has teased a ProArt laptop targeting creative professionals. MSI is preparing at least one N1X system. Earlier supply chain reports also cited a Microsoft Surface variant.

First devices are expected to reach market before the 2026 holiday season, with broader availability into early 2027. Premium laptop pricing is expected given LPDDR5X memory costs and TSMC 3nm manufacturing; the absence of the DGX Spark's specialized ConnectX-7 networking card in laptop designs will reduce costs from the desktop's $3,999 baseline.

What Happens at Computex: Key Dates for Laptop Buyers

Jensen Huang is scheduled to deliver his GTC Taipei keynote at the Taipei Music Center on June 1 at 11:00 a.m. local Taipei time, which is 11:00 p.m. ET on Sunday, May 31. The keynote will be livestreamed globally via Nvidia's YouTube channel. Computex 2026 proper runs June 2–5, with 1,500 exhibitors expected; OEM partners have invited press to separate laptop unveilings in the hours following Huang's address. Any hardware announced on June 1 is expected to be available for hands-on demos at OEM booths starting June 2.

If the announcement proceeds as expected, anyone weighing a laptop purchase in the second half of 2026 now has a new variable to factor in — one that arrives with desktop-class GPU credentials and a software ecosystem no Windows ARM chip has previously offered.

Frequently Asked Questions

What is Nvidia's N1X chip?

The Nvidia N1X is an ARM-based laptop system-on-chip developed with MediaTek, built on TSMC's 3nm process. It combines a 20-core ARM CPU with a Blackwell-architecture GPU carrying 6,144 CUDA cores — the same shader count as the desktop RTX 5070 — in a single package with a unified LPDDR5X memory pool. It is Nvidia's first chip designed specifically for Windows ARM laptops.

When will Nvidia N1X laptops be available to buy?

Nvidia has not officially confirmed ship dates. Based on OEM leaks and supply chain reports, first devices from Dell, Lenovo, Asus, and MSI are expected to reach market before the 2026 holiday season, with wider availability into early 2027. Dell is set to reveal an embargoed XPS model on May 31, one day before Computex officially opens.

What does CUDA support on a laptop mean for developers?

CUDA on a laptop means AI and machine learning workloads written for Nvidia's GPU software stack — PyTorch CUDA backend, TensorRT, TensorRT-LLM, llama.cpp CUDA builds — can run natively on an N1X device without code changes or framework substitution. That is not currently possible on Windows ARM laptops using Qualcomm's Snapdragon platform, which requires Qualcomm's QNN or DirectML as alternatives.

How does Nvidia N1X compare to Qualcomm Snapdragon X Elite?

Pre-release Geekbench prototype scores show the N1X scoring approximately 15% higher than the Snapdragon X Elite in single-core CPU performance. The N1X's primary advantage is its Blackwell-architecture GPU and the full CUDA software ecosystem, which Qualcomm's platform cannot match. Both run Windows on ARM; software compatibility for legacy x86 apps depends on Windows emulation, which has improved substantially but is not universal for all applications.