NVIDIA's RTX Spark wants to reinvent the Windows PC

NVIDIA has declared war on the Windows PC processor market. At Computex in Taipei, chief executive Jensen Huang unveiled RTX Spark, an Arm-based superchip that pairs a 20-core NVIDIA Grace CPU, co-developed with MediaTek, with a Blackwell RTX GPU carrying 6,144 CUDA cores, joined by NVIDIA's NVLink chip-to-chip interconnect.

NVIDIA claims up to one petaflop of AI compute and up to 128GB of unified memory shared across CPU and GPU. One correction to the hype matters immediately: these machines are not shipping yet. NVIDIA says the first RTX Spark laptops and compact desktops arrive in fall 2026, from partners including ASUS, Dell, HP, Lenovo, Microsoft Surface, and MSI.

The pitch

The argument is about where AI runs. Today, serious AI work usually happens in a data centre, with subscriptions, latency, and privacy trade-offs attached. NVIDIA's bet is that more of that work should happen on your own machine. It claims RTX Spark can run large local models, handle demanding video workflows, and still play games well.

The differentiator is memory. No integrated neural processor in a conventional laptop chip competes with thousands of CUDA cores for local generative work, and 128GB of shared memory changes what a laptop can attempt.

Microsoft is part of the story too, with new operating-system security primitives and agent permission work designed so on-device AI agents touch only the data and tools a user permits. That is necessary because an AI agent with unrestricted access to your PC is not convenience. It is a security incident waiting for a calendar invite.

Why scepticism is fair

NVIDIA has been here before. Windows RT ran on NVIDIA Tegra chips in 2012 and collapsed because it could not run the x86 desktop software that defined Windows. Microsoft wrote down Surface RT heavily, and NVIDIA left the Windows PC processor fight for years.

Two things are different now. Microsoft's x86-to-Arm emulation is much better, and the software ecosystem has spent years adapting because of Qualcomm's Windows on Arm push. RTX Spark inherits a world that did not exist in 2012. But compatibility remains the issue to check before buying any Arm Windows machine.

Price and Kenya

NVIDIA has not announced pricing. Analyst notes have suggested premium configurations could sit far above mainstream laptop budgets. At that level, this is not a product for most Kenyan buyers in 2026.

Still, the direction matters. If local AI becomes the thing premium laptops compete on, that logic eventually flows downmarket. A machine that runs capable models locally needs less cloud subscription, less latency, and less reliable connectivity. In a market with expensive data and uneven connections, that is genuinely interesting.

The caution is equally real. The same memory demand fuelling local-AI machines is part of what is raising ordinary PC costs. The AI PC boom may make everyday laptops more expensive before it makes them smarter.

RTX Spark is not just another laptop chip. It is a test of whether Windows can survive a post-x86 identity crisis.