NVIDIA Just Built Its First Consumer CPU — And It's Trying to Turn Every PC Into an AI Supercomputer

I've covered a lot of hardware launches over the years, but when I saw NVIDIA's Computex 2026 keynote, I had to rewatch it twice just to make sure I understood what I was seeing. NVIDIA just built its first consumer CPU — and it's not a standalone processor. It's a chip that could fundamentally change what a personal computer is.

The RTX Spark: What It Actually Is

The RTX Spark is what NVIDIA calls a "superchip" — a single system-on-a-chip (SoC) that combines a powerful ARM-based CPU, an NVIDIA Blackwell GPU, and up to 128GB of unified memory. Built in partnership with MediaTek, this is NVIDIA's play to own the entire PC computing stack — not just the graphics card.

Here are the raw specs:

• CPU: 20 ARM cores (10x Cortex-X925 performance + 10x Cortex-A725 efficiency), peaking at 4.1GHz
• GPU: Blackwell architecture, 6,144 CUDA cores, fifth-gen Tensor Cores with FP4 precision
• AI Performance: 1 petaflop of FP4 AI compute
• Memory: Up to 128GB LPDDR5X with 300 GB/s bandwidth
• Manufacturing: 70 billion transistors on TSMC's 3nm process

Why This Changes Everything for AI on Your Laptop

Right now, running powerful AI models locally on a laptop is a compromised experience. You're either dealing with slow inference because you don't have enough VRAM, or you're using a beefy desktop setup that's impractical to carry. Cloud-based AI is fast but requires internet and costs money on every query.

The RTX Spark changes that equation. With 128GB of unified memory shared between the CPU and GPU — compared to 16GB or 32GB on most current laptops — you can run genuinely large AI models locally. NVIDIA calls it a "personal AI supercomputer," and for once, that's not just marketing speak.

NVIDIA has partnered with Microsoft to make Windows itself agentic with this chip. The vision: your laptop doesn't just run apps. It runs AI agents that complete tasks autonomously, using the RTX Spark as the local compute backbone so nothing has to touch the cloud unless you want it to.

NVIDIA vs. Apple Silicon: The Battle for the ARM PC

You can't talk about RTX Spark without mentioning Apple Silicon. Apple's M-series chips proved that ARM-based processors could compete with — and often beat — Intel and AMD x86 chips in performance per watt. The MacBook Air and MacBook Pro are consistently the best laptops on the market for creative professionals and developers.

NVIDIA has now entered that fight. And they're not bringing just a good CPU — they're bringing the best GPU brand in the world, combined with a CPU, in a single chip. ASUS, Dell, HP, Lenovo, Microsoft Surface, and MSI are all building systems around RTX Spark, with over 30 laptops and 10+ desktops planned in the first wave alone, arriving this fall.

The 1 Petaflop Number Nobody Is Talking About

The headline specs are impressive, but the one I keep coming back to is the AI compute figure: 1 petaflop of FP4 AI performance. A petaflop is a quadrillion floating-point operations per second. The RTX Spark delivers this in something that fits in a laptop bag — at a fraction of the power consumption of the server racks that used to be required.

The Bigger Picture: NVIDIA's Total Stack Ambition

Step back and look at what NVIDIA is doing. A few years ago, they were a GPU company. Then they became the backbone of the AI datacenter boom. Now they're entering the consumer PC market with their own CPU. They're also building robots, autonomous vehicles, and data center infrastructure.

NVIDIA isn't just building chips anymore. They're building computing platforms for every layer of the AI economy — from the cloud datacenter to the edge device in your bag. RTX Spark is the consumer edge piece of that strategy, arriving this fall from every major PC manufacturer.

What's your experience? Drop a comment below! 👇

Would you switch from a MacBook to an RTX Spark Windows laptop — or do you think Apple Silicon still has too big of an ecosystem advantage to close the gap?