What AI PCs Are and How AMD Got a Head Start
AI PCs are laptops and desktops built to run complex models such as large language models, AI agents, and creative tools with local AI processing instead of relying on remote cloud servers, so they deliver faster responses, better privacy, and more consistent performance even without a network connection. In that race, AMD AI processors arrived early. The company’s Ryzen AI Max+ 395 “Strix Halo” SoC has been shipping in laptops and mini PCs since early 2025, pairing 16 Zen 5 CPU cores with a 40‑CU RDNA 3.5 iGPU and up to 128GB of unified memory. At Computex, AMD executives pointed to a portfolio of 35 Strix Halo products already in market and welcomed NVIDIA’s entrance. Their confident message was clear: AMD built this class of machine first, and now it wants to defend that lead as the ecosystem grows.
Inside NVIDIA RTX Spark Laptops: A New SoC for Local AI
RTX Spark laptops are NVIDIA’s answer to this new category, built around its first true client SoC for Windows PCs. The chip combines a 20‑core Arm CPU based on the Grace architecture with a Blackwell GPU, 1 PFLOPs of AI performance, and up to 128GB of unified LPDDR5X memory, all tuned for local AI processing as well as gaming and professional workloads. According to Wccftech’s Computex hands‑on, these RTX Spark laptops and mini PCs are designed as premium systems that can run demanding AI workloads directly on-device. They also inherit technology from NVIDIA’s existing DGX Spark desktop platform, making the mobile GB10‑derived silicon part of a broader client strategy. While the silicon will not ship until later this year, NVIDIA is positioning RTX Spark laptops as the reference design for its vision of AI‑first Windows PCs.
NVIDIA vs AMD AI: Hardware Similar Goals, Different Paths
On paper, AMD AI processors and RTX Spark laptops target the same goal: high‑end on‑device AI performance with plenty of memory headroom. AMD’s Strix Halo offers 16 Zen 5 CPU cores, 32 threads, a 40‑CU RDNA 3.5 GPU, and up to 128GB of unified memory, with as much as 96GB addressable as VRAM, in both laptops and mini PCs. NVIDIA’s RTX Spark SoC instead leans on Grace CPU cores and a Blackwell GPU with 6,144 CUDA cores, 1 PFLOPs of FP4 compute, and similar unified memory capacity. Both platforms promise to run sizable models and AI agents locally, while their vendors talk up future architectures such as Vera Rubin on NVIDIA’s side. In this NVIDIA vs AMD AI contest, raw specifications are no longer the only story; platform philosophy and how well each integrates CPU, GPU, and memory will shape user experience.
Software Stacks: CUDA Advantage vs AMD’s Catch-Up
The biggest difference between AMD’s early AI PCs and RTX Spark laptops is now software, not hardware. NVIDIA’s CUDA and related tools have long defined the default stack for AI developers, and that weight carries over to RTX Spark, which can tap into existing CUDA‑based frameworks and optimized models. AMD’s ROCm stack historically lagged in features and polish, but recent work has improved local AI support on its GPUs and APUs. XDA reports that older assumptions about AMD and local AI are “quickly going stale,” and AMD is investing to close the gap, including a Ryzen AI Halo developer mini‑PC built from Strix Halo that runs models up to 200 billion parameters and dual‑boots Windows 11 and Linux. In the next phase of NVIDIA vs AMD AI, the winner will be decided by developer adoption, toolchains, and how easily users can deploy models at home or at work.
OEM Backing and What It Means for Future AI Laptops
While AMD has the time advantage, RTX Spark laptops arrive with strong OEM momentum. At Computex, NVIDIA hardware appeared in designs from ASUS, Dell, HP, Microsoft, MSI, and Lenovo, ranging from creator‑focused notebooks like Dell’s XPS 16 Creator Edition to ASUS ProArt laptops bundled with local AI creative suites. This breadth signals that major PC brands are betting on NVIDIA’s client SoC as a key platform for AI‑centric Windows machines. AMD, for its part, can point to an established catalog of 35 Strix Halo systems already on sale, plus a turnkey Ryzen AI Halo developer mini‑PC that mirrors NVIDIA’s DGX Spark positioning. For buyers, the short‑term outlook is choice: RTX Spark laptops with deep CUDA roots against seasoned AMD AI processors with growing software support. Over time, the richer ecosystem and smoother tools are likely to matter more than who shipped first.
