NVIDIA unveils RTX Spark to push AI PCs deeper into Windows

NVIDIA is making a more direct play for the Windows PC

NVIDIA has introduced RTX Spark, a new system-on-a-chip aimed at Windows laptops and small desktops, in what looks like one of the company’s clearest attempts yet to shape the next generation of AI-capable personal computers. Unveiled at Computex, the chip is positioned against AMD’s Ryzen AI Max line and Qualcomm’s latest Snapdragon X-class systems, but NVIDIA is pitching it less as a conventional PC processor and more as the foundation for a new class of AI-first machines.

The company says RTX Spark can deliver 1 petaflop of AI computing power. It also says the chip combines 6,144 Blackwell RTX cores with 20 Arm CPU cores built with MediaTek, plus an NPU capable of meeting Microsoft’s Copilot+ threshold of 40 TOPS. In NVIDIA’s telling, however, the NPU is not the main story. The company is emphasizing the AI role of its Blackwell GPU tensor hardware and the benefits of unified memory for local AI workloads.

A Windows push built around Arm and unified memory

RTX Spark is notable not just for performance claims, but for how NVIDIA is assembling the package. The chip is designed for Windows systems and will appear in products including Microsoft’s Surface Laptop Ultra and Dell’s XPS 16, according to the supplied source text. NVIDIA also says systems are coming from every major OEM.

The architecture gives the GPU direct access to a large pool of unified memory that can range from 16 GB to 128 GB. That matters for local AI tasks, where moving large models and datasets efficiently can become as important as raw compute. The chip is also said to operate across a broad power envelope, from single-digit wattage up to 80 watts, suggesting NVIDIA wants it to scale from portable systems to more compact performance-oriented desktops.

Just as important is what RTX Spark is not. NVIDIA says there are no plans to pair it with a dedicated GPU. The company is clearly presenting it as a self-contained computing platform rather than a stopgap on the way to a modular notebook design.

Microsoft has been involved for years

The launch also highlights how closely the Windows ecosystem is being tuned around AI hardware. NVIDIA representatives said the company has been working with Microsoft for several years on the platform. Microsoft executive Pavan Davuluri, cited in the source text, said Windows 11 workload profile scheduling was optimized for RTX Spark.

That kind of coordination matters because Arm-based Windows systems still live under the shadow of software compatibility and workload balancing. NVIDIA said it is familiar with Microsoft’s Prism emulation layer for running older Windows applications, and also said it is working with major anti-cheat providers, a notable issue for gaming support on earlier Arm PCs.

The strategic goal goes beyond laptop refreshes

NVIDIA chief executive Jensen Huang framed RTX Spark in broader terms, arguing that PCs are evolving into devices increasingly shaped by AI agents rather than direct manual input alone. Whether or not that vision arrives on the timetable NVIDIA wants, the product makes the company’s ambition clearer. It is not merely supplying discrete graphics to Windows manufacturers anymore. It wants a central position in the AI PC stack, from silicon and memory architecture to software optimization and OEM rollout.

That ambition is arriving at a moment when the PC industry is searching for a post-refresh narrative strong enough to justify premium hardware. Copilot+ branding, on-device model execution, and AI-assisted workflows have become part of that effort, but many current systems still treat AI as an add-on rather than the organizing principle of the machine. RTX Spark is meant to argue for a more integrated model.

The open questions

For now, several key details remain unresolved. NVIDIA has not gone deep on pricing, real-world battery tradeoffs, or sustained performance under shipping thermal conditions. It has also not yet provided the kind of broad independent benchmark picture that would show how its 1 petaflop claim translates across creative tools, developer workflows, or local model inference tasks.

Still, the launch is strategically important even before those answers arrive. It puts NVIDIA directly into the conversation around Windows notebooks as complete computing systems, not just graphics accelerators. If RTX Spark performs as advertised, it could give OEMs a high-end alternative in the fast-forming AI PC category and tighten the competitive pressure on both AMD and Qualcomm.

The larger significance is that the fight over the AI PC is no longer just about who supplies the fastest accelerator. It is becoming a contest over who defines the machine. With RTX Spark, NVIDIA is making its bid to do exactly that on Windows.

This article is based on reporting by Engadget. Read the original article.